Applied Digital APLD NASDAQ

The record shows a company that began with a modest equity base and has needed progressively larger and more structurally complex capital each year. The two 2006 entries (claims #21056 and #21058) appear to describe a different entity's technology programs and require explanation before they are read as part of this history. The record also contains no insider transactions, so that dimension cannot be assessed. The usable record begins with small convertible preferred tranches in 2021 (the $1.3M Series D, per claim #14501). It moves through an IPO that was marketed at up to $75M (claim #14041) but priced at $40M (claim #14007) and netted roughly $36M (claim #24799). That gap between the announced and the realized raise is the first data point on how much demand the market had for the equity.

Through 2022 and early 2023, the mix reads as coherent. Regional bank term loans were explicitly framed as non-dilutive (claim #21492) and tied to named facilities, such as the $20M Starion loan for the Ellendale expansion (claim #13326). The character changes from mid-2023. The $50M B. Riley facility at 9% (claims #12565 and #12517) was followed within weeks by a $125M ATM (claim #12397), which was reported complete by January 2024 (claim #21552), and the B. Riley loan was terminated early (claim #6291). In mid-2024 the layering accelerates. The company added a prepaid advance with YA II for $42.1M with monthly conversion limits (claims #6078 and #6111), a $50M Series E preferred (claim #6119), CIM borrowings carrying several warrant tranches (claims #6029 and #6031), a $50M Series F convertible preferred (claim #5833), a $160M PIPE (claim #5782), a $250M SEPA (claim #5780), and a renewed $125M ATM (claim #5754), all within roughly four months. Several of these instruments were amended, waived or conditioned on registration deadlines (claims #5837, #5760 and #6036). This pattern looks more like successive bridging than a single designed capital stack.

Use of proceeds is documented mainly at the facility level for the early bank loans. The Ellendale HPC construction financing was described as in advanced discussion in August 2024 (claims #24912 and #26120), but the record does not show it closing. Project-level financing for the 400 MW lease (claim #3237) is also still described as prospective.

A careful reader should watch for three things. First, whether the Ellendale project financing closes on non-dilutive terms. Second, how much of the SEPA and ATM capacity is actually drawn. Third, the cumulative share count once the preferred conversions, YA conversions and accordion-linked warrants settle. Many figures in the record appear unit-truncated (for example claims #3121 and #13552), and the dataset ends in September 2024, so this read covers only that window.

This record opens with two 2006 entries (claims 21056 and 21058) describing UNICORN and SOCRATES technology financing. They do not connect to the later data-center business and appear to belong to an unrelated or predecessor entity; their relevance requires explanation before they are read as part of this history. The record also ends mid-entry in September 2024, so the most recent period is incomplete.

The substantive history begins with small convertible preferred issuances in 2021 (the $660,000 Series C and $1.3M Series D, per claims 14500 and 14501). It then moves to an IPO whose stated size compressed from up to $75M (claim 14041) to $40M priced (claim 14007) and roughly $36M net (claim 24799), alongside a one-for-six reverse split (claim 26010). Through 2022 and early 2023 the company leaned on regional bank debt: the $7.5M Vantage loan (claim 21459) and the $15M and $20M Starion facilities (claims 13804 and 13310). Management framed this explicitly as non-dilutive (claim 23493).

That framing is no longer discussed after mid-2023. The $50M B. Riley term loan at 9% (claims 12565 and 12517) coincided with a $125M ATM (claim 12397), which was fully used by January 2024 (claim 21552). The B. Riley loan was then terminated early (claim 6291). From there the instruments become more structured and more equity-linked in quick succession: an unverified $20M AI Bridge note later prepaid in stock (claims 6292 and 6137), the $42.1M Yorkville prepaid advance with monthly conversion limits (claims 6106 and 6111), CIM borrowings carrying warrant tranches (claims 6029 and 6031), the Series E and Series F preferred (claims 6119 and 5757), a $160M PIPE (claim 5782), a $250M SEPA (claim 5780), and a second $125M ATM (claim 5754). The 2024 cadence reads less as a single capital plan than as successive layering, where each instrument bridged to the next.

On use of proceeds, the project-level financing for the 400 MW hyperscaler lease was cited in April 2024 (claim 3237). By August 2024 Ellendale construction financing was still described as advanced discussions (claims 24912 and 26120), which suggests that milestone was delayed relative to the equity-linked capital raised in the interim.

There are no insider transactions in the record, so no conclusions about insider behavior can be drawn. A careful reader should next watch four things: whether the Ellendale project financing closes; how much of the SEPA and the second ATM is actually drawn; the conversion of the roughly 7.6M Series F shares (claim 5834) and the outstanding warrants; and whether disclosures after September 2024 confirm or revise the $150M in funding reported as secured (claim 21568).

Applied Digital's record divides into three financing phases, each more dilutive and more structurally complex than the last. The earliest phase was small and conventional. It consisted of convertible preferred in the low single-digit millions (Series D at $1.3M, per claim 14501) and an IPO whose stated size was revised from $75M (claim 14370) to $40M at $5.00 per share (claim 14007), netting $36M (claim 24799) alongside a one-for-six reverse split (claim 26010). Through 2022 and early 2023 the company leaned on regional bank debt, which it framed as non-dilutive (claim 21492). That debt included the $7.5M Vantage loan (claim 21459) and the Starion facilities of $15M and $20M (claims 13804, 13310), and it was paired with a single round of share issuances to GMR, SparkPool and Valuefinder (claims 13751–13753). From mid-2023 the mix widened sharply. A $175M shelf (claim 12544), a $125M ATM (claim 12398) and a 9% B. Riley term loan of $50M (claims 12565, 12517) arrived within weeks of each other. The ATM was then drawn heavily, at $64.5M (claim 21542) and then $23.1M to completion (claim 21552), and the B. Riley loan was terminated early (claim 6291). The 2024 layer is where the pattern reads as patching rather than a single plan. Within roughly four months the company used an unsecured bridge later prepaid in stock (claims 6292, 6137) and a Yorkville prepaid advance of $42.1M (claim 6078) whose $16M monthly conversion cap (claim 6111) was subsequently eliminated (claim 5837). The same window included CIM debt with warrant tranches tied to accordion draws (claims 5973–5975, 6029, 6031), $50M of Series E (claim 6119), $50M of Series F convertible into 7,598,714 shares (claims 5757, 5834), a $250M SEPA (claim 5756) and a $160 private placement (claim 5783). Several instruments overlap in purpose and draw on the same counterparties. The stated use of proceeds tracks deployment only loosely. Ellendale expansion financing (claim 13326) is consistent with the land purchase (claim 12622). However, project-level financing for the 400 MW hyperscaler lease was first described in April 2024 (claim 3237) and was still characterized as advanced discussions in August (claim 24912), so it is delayed within this record. A sophisticated reader should watch four things: whether Ellendale HPC construction financing closes, the pace of Yorkville conversions now that the cap is gone, SEPA utilization, and whether CIM accordion draws trigger the contingent warrants. Three limits on the record also apply. It contains no insider transactions. It ends in September 2024. The 2006 SOCRATES and UNICORN entries (claims 21058, 21056) predate the current business and require explanation.

The record supports a reading of the 2022 to 2024 period. The two 2006 entries (claims 21056, 21058) describe unrelated technology programs and appear to belong to a different issuer, so they are set aside here. The list ends in September 2024 mid-entry, and it contains no insider transactions, so nothing can be said about insider selling.

The dilution pattern has two phases. Through early 2023 the company leaned on regional bank debt, which it explicitly framed as non-dilutive (claim 21492). Examples include the Vantage term loan (claim 21459), the $15M Starion facility (claim 13804) and the $20M Starion loan for Ellendale (claim 13326). That phase rested on modest IPO net proceeds of $36M (claim 24799). From mid-2023 the mix turns steadily toward equity and equity-linked paper. The $125M ATM (claim 12397) was reported complete by January 2024 (claim 21552). It was followed in quick succession by a $42.1M Yorkville prepaid advance (claim 6078), CIM borrowings carrying warrants (claim 6029), two preferred series, a $160M PIPE (claim 5782), a $250M standby equity purchase agreement (claim 5780) and a fresh $125M ATM (claim 5754). The pace of new instruments clearly accelerates through 2024.

Whether this is a coherent strategy is harder to establish than the cadence. Several features read as patching rather than design. The $50M B. Riley term loan (claim 12565) was terminated early (claim 6291). Yorkville received consent shares (claim 5988). A conversion limitation on the prepaid advance was later eliminated (claim 5837). CIM required a waiver to permit additional borrowing (claim 5760). Each of these is explainable on its own, but together they suggest terms being renegotiated as needs arose.

Use of proceeds tracks only partly. The Ellendale-directed bank debt and land purchase (claim 12622) line up with stated purpose. Project-level construction financing for Ellendale HPC, however, appears repeatedly as advanced discussions (claims 24912, 5867) without a closing on record. The $160M in asset sales and financing (claim 21556) and the $25M Garden City escrow release (claim 3242) are disclosed as sources, but the record does not show how they were deployed.

A careful reader should watch five things next. The first is how much of the SEPA and second ATM is actually drawn. The second is conversion of the Series F into roughly 7.6M shares (claim 5834). The third is accordion draws that trigger contingent warrants (claim 6031). The fourth is whether the Yorkville registration condition (claim 6036, not yet verifiable) was met. The fifth is whether the Ellendale construction financing is closed or is no longer discussed.

The record arrives truncated mid-entry at claim #5778 (September 2024), so this read covers only the period through that point. Two 2006 entries (claims #21056 and #21058, UNICORN and SOCRATES technology) appear to belong to a different issuer and are set aside, though their presence requires explanation. Several figures are also recorded at inconsistent scales, such as IPO proceeds of $36 in claim #3120 against $36,000,000 in claim #24799. The IPO size is likewise reported differently across claims, with $75M in claim #14370 and $40M at $5.00 per share in claim #14007. The underlying arc is still legible. After a modest 2022 listing netting roughly $36M (claim #24799), management leaned explicitly on what it called non-dilutive regional bank debt (claim #21492). That debt included the Vantage $7.5M term loan (claim #21459) and the Starion facilities of $15M and $20M (claims #13804, #13326). In 2023 the mix shifted toward higher-cost and equity-linked capital: the $50M B. Riley term loan at 9% (claims #12565, #12517), the $175M shelf (claim #12544), and a $125M ATM (claim #12397), which was run to completion by January 2024 (claim #21552). The first half of 2024 reads less as a single plan than as layered stopgaps. The B. Riley loan was terminated early (claim #6291) and replaced by an unsecured $20M AI Bridge loan (claim #6292), which was itself prepaid in stock (claim #6137). Alongside it came Series E preferred (claim #6119), a $42.1M Yorkville prepaid advance convertible at up to $16M per month (claims #6078, #6111), a second ATM with Roth (claim #6081), and the CIM note with a $75M accordion and tranche-linked warrants (claims #5975, #6031). By September 2024 the instruments had grown larger and more strategic in tone, including the $160M Goldman-backed PIPE (claim #5782), the $250M SEPA (claim #5780), and Series F preferred (claim #5833). Their cumulative overhang, however, points to dilution accelerating rather than stabilizing. On use of proceeds, the April 2024 reference to project-level financing for a 400 MW hyperscaler lease (claim #3237) had reverted by August to advanced discussions on Ellendale HPC construction financing (claim #24912). Deployment for that project therefore remained undisclosed in this record, and the 2022 loan-use claim (claim #13807) carries no specifics. The record contains no insider transactions, so nothing can be said about management selling. A careful reader should next watch for several things: whether the Ellendale construction financing closes on project-level, non-recourse terms; the pace of Yorkville conversions and SEPA draws; the triggering of CIM accordion warrants; and the handling of the Series F redemption window (claim #5839).

The record as supplied ends mid-entry in September 2024 and contains no insider transactions, so this read covers financing cadence only. The two 2006 entries (claims #21058, #21056) appear to belong to a predecessor or unrelated entity and are set aside. Several later figures are recorded in inconsistent units, for example the IPO net proceeds appear as $36,000,000 in claim #24799 and as $36 in claim #3120.

The trajectory falls into three phases. Through 2022 the company leaned on a modest IPO, the $40,000,000 raise per claim #14007, followed by regional bank debt it described as non-dilutive (claim #21492): the $7.5M Vantage loan (claim #21459) and the Starion facilities of $15M and $20M (claims #13804, #13310). 2023 marked a shift toward shelf capacity, with the $175M S-3 (claim #12544) and the $125M ATM (claim #12411), drawn by $64.5M (claim #21542) and reported complete with $23.1M more (claim #21552). It also brought costlier credit, the $50M B. Riley loan at 9% (claims #12565, #12517). Mid-2024 is where the pattern changes in character. Within roughly four months the company layered the $20M AI Bridge loan (claim #6292), the $50M Series E (claim #6119), the $42.1M Yorkville prepaid advance (claim #6078), the CIM note with warrants on 6,300,449 shares (claims #5973, #5972) and the $50M Series F (claim #5833). It then added the $160M PIPE (claim #5782) and the $250M SEPA (claim #5780). That density, together with the removal of the prepaid advance's conversion limitation (claim #5837) and a CIM waiver to permit further borrowing (claim #5760), reads more as successive bridging than as a pre-set capital plan.

Use of proceeds tracks clearly in the early phase, where the Starion loan was tied to the Ellendale expansion (claim #13326). The later and larger objective has not been resolved within the record. Project-level financing for the 400 MW hyperscaler lease (claim #3237) and construction financing for Ellendale HPC (claims #21574, #24912) are both described as in discussion, and neither is shown as closed by the end of the data.

A careful reader should watch whether that project financing closes, since it would explain the structured instruments as temporary. Further items to track are the pace of SEPA and ATM draws, the 7,598,714-share Series F conversion overhang (claim #5834), and the contingent accordion warrants on 5,559,220 shares (claim #6031).

The disclosed record runs from a small preferred-stock base in 2021 to a layered capital stack by September 2024, and the trajectory is one of accelerating, increasingly structured issuance. The early phase was conventional. It combined Series C and D convertible preferred (claims #14500, #14501), an IPO of 8,000,000 shares at $5.00 (claim #14007) netting roughly $36M (claim #24799), and regional bank term loans the company framed as non-dilutive (claims #21459, #13804, #23493). Through early 2023 the mix reads as coherent: equity funded the uplisting, and secured bank debt funded discrete build-outs, with the $20M Starion loan tied explicitly to Ellendale expansion (claim #13326).

From mid-2023 the pattern shifts. A $175M shelf (claim #12544) and a $125M ATM (claim #12397) arrived alongside a 9% B. Riley term loan (claims #12565, #12517) that was terminated early (claim #6291). Between May and September 2024 the company added a Yorkville prepaid advance with $40M and $10M notes (claims #6079, #6080), Series E and Series F preferred (claims #6119, #5757), a second ATM with Roth (claim #6081), a CIM senior secured facility carrying warrants on 6,300,449 shares (claim #6029), and a $250M SEPA (claim #5756). The density of instruments, the consent shares issued to Yorkville (claim #6032) and the waiver required for additional CIM borrowing (claim #5760) suggest capital sourced as it became available rather than against a fixed plan, although each instrument is individually disclosed.

Whether proceeds track deployment is only partly observable. The bank loans map to named facilities, and the $160M in asset sales and financing (claim #21556), along with the escrow release (claim #3242), correspond to the Garden City disposition. Ellendale HPC construction financing, however, is described repeatedly as being in advanced discussions (claims #24912, #5867) without a closed facility appearing in the record. Project-level financing for the 400 MW hyperscaler lease (claim #3237) is likewise not shown as closed.

Several features of the record require explanation. The 2006 claims (#21056, #21058) reference technology unrelated to the current business and may reflect a predecessor or a mismatched entity. The record also contains no insider transactions, and it ends mid-entry in September 2024, so later activity cannot be assessed. A sophisticated reader should watch four things: whether the Ellendale construction financing closes, how much SEPA and ATM capacity is drawn, the pace of Yorkville conversions against the $16M monthly limit (claim #6111), and whether Series E and F dividends are paid in cash or in stock.

Applied Digital's record shows a company that started with lender-funded expansion and moved, over about two years, toward layered equity-linked capital. The early years are modest. The IPO priced at $5.00 on 8,000,000 shares for $40,000,000 (claim 14007) and netted roughly $36,000,000 (claim 24799), alongside a one-for-six reverse split (claim 26010). In 2022 the company described its regional bank debt as non-dilutive (claims 21492, 23493). The Vantage $7.5M term loan (claim 21459) and Starion's $15M and $20M facilities (claims 13804, 13326) fit that framing. The two 2006 entries (claims 21058, 21056) appear to describe an unrelated technology business and should not be read as part of this company's financing history.

The trajectory changes in 2023. The $50M B. Riley facility at 9% (claims 12565, 12517) and the $125M Craig-Hallum ATM (claim 12397) brought in costlier and dilutive capital. The ATM was drawn quickly: $64.5M by October 2023 (claim 21542) and fully used by January 2024 (claim 21552). By mid-2024 the instruments overlap rather than follow each other. They include the $20M AI Bridge loan (claim 6292), $50M of Series E preferred (claim 6119), and the Yorkville prepaid advance of $42.1M with a monthly conversion cap (claims 6106, 6111). There is also the CIM facility with warrants on 6,300,449 shares (claim 6029), a $25M Roth ATM (claim 6081), Series F preferred (claim 5833), and a $250M standby equity purchase agreement (claim 5780). The Yorkville conversion limit was later removed (claim 5837), and CIM required a waiver to permit additional borrowing (claim 5760). Together these read less as a planned capital structure and more as bridging toward a larger project financing.

The record does not allow a clear comparison between the stated use of proceeds and actual deployment. Claim 13807 refers to a use of loan proceeds without detail, and the Ellendale HPC construction financing appears only as advanced discussions (claims 24912, 21574) as of the last entries. The record breaks off in September 2024, so later outcomes are not covered.

A careful reader should watch four things. The first is whether the Ellendale project-level financing closes, or keeps being described as under discussion. The second is the pace of Yorkville conversions now that the cap is gone. The third is how much of the SEPA is drawn. The fourth is whether CIM accordion draws trigger the contingent warrants of 5,559,220 shares (claim 6031). Inconsistent units across several claims, such as claims 3121 and 12517, also limit how precisely these figures can be reconciled.

This record is truncated mid-entry in September 2024, and it contains no insider transactions, so any read is limited to roughly two and a half years of financing activity. The two 2006 entries (claims #21058, #21056) describe unrelated technology programs and require explanation before they are treated as part of this company's history. Several amounts also appear unit-inconsistent, with share counts recorded as dollars (claim #14080) and millions recorded as single digits (claim #3139), so the figures below are those that reconcile across multiple claims.

The trajectory moves through three distinct phases. In 2022 the company paired a modest IPO, with net proceeds of $36M per claim #24799, with regional bank term debt such as the $15M Starion facility (claim #13804), and it explicitly framed that debt as non-dilutive (claim #21492). In 2023 the cost and scale of capital stepped up. The company took a $50M B. Riley term loan at 9% (claims #12565, #12517), filed a $175M shelf (claim #12544), and opened a $125M ATM (claim #12397), which it drew at least $64.5M from by October (claim #21542) and reported complete by January 2024 (claim #21552). By mid-2024 the stack had become layered and increasingly equity-linked. It included an unsecured AI Bridge loan later prepaid in stock (claim #6137), a $42.1M Yorkville prepaid advance (claim #6078) whose monthly conversion cap (claim #6111) was later eliminated (claim #5837), CIM borrowings carrying warrants on 6,300,449 shares (claim #6029), and 9% Series E preferred (claim #6136). In a single September cluster the company added a $160M PIPE (claim #5782), a $250M standby equity facility (claim #5780), a new $125M ATM (claim #5778), and a $125M CIM facility (claim #5753).

The mix reads less as a single plan than as successive instruments added to bridge construction timing. The early termination of the B. Riley loan (claim #6291) and the relaxation of the Yorkville conversion terms suggest that terms were renegotiated as needs shifted. Stated use of proceeds is thin. Ellendale is tied to specific debt (claim #13326), but project-level HPC construction financing appears only as advanced discussions as of August 2024 (claims #24912, #21574), and its resolution falls outside this record.

A careful reader should watch whether Ellendale construction financing closes, the pace of draws under the standby facility and the ATM, the conversion behaviour on the Yorkville notes, warrant exercise under the CIM accordion (claim #6031), and how the Series E and F preferred are redeemed or converted.

The record runs through 2024-09-10 and stops partway through an entry, so this reading covers the period up to that date. No insider transactions appear in it. The two 2006 entries (claims #21056, #21058) predate the current business by more than a decade and require explanation before they are treated as part of this lineage.

The trajectory of dilution is clear. Applied Digital moved from a conventional base to more layered instruments over roughly two years. It began with a small IPO of 8,000,000 shares at $5.00 (claim #14007), netting $36M (claim #24799), and with regional bank debt that management described as non-dilutive (claim #21492). That debt included the $7.5M Vantage loan (claim #21459) and successive Starion facilities of $15M and $20M (claims #19165, #13310). In 2023 the company added a $50M B. Riley term loan at 9% (claims #12514, #12517) and a $125M ATM (claim #12397). That ATM was drawn quickly, with $64.5M reported in October (claim #21542) and the program complete by January 2024 (claim #21552). The 2024 mix then broadens considerably. It includes the Yorkville prepaid advance of $42.1M (claim #6078) and CIM debt carrying warrants for 6,300,449 shares (claim #5972), with up to $110M in subsequent borrowings and a $75M accordion (claims #5974, #5975). It also includes Series E and Series F preferred (claims #6119, #5833), a $250M SEPA (claim #5756), and a fresh $125M ATM (claim #5778).

This mix reads less like a single plan than like layered bridging. The B. Riley loan was terminated early (claim #6291), less than a year after it was signed. The AI Bridge note was prepaid in common stock (claim #6137). Yorkville's conversion limit was later eliminated (claim #5837), and a prepayment obligation was tied to an S-1 effectiveness deadline (claim #6036, not yet verifiable). Each of these instruments was revised or replaced within a short window.

The stated use of proceeds has stayed consistent, centring on Ellendale (claim #13326). The project-level financing for the 400 MW hyperscaler lease (claim #3237) and the Ellendale HPC construction financing, described as being in advanced discussions (claim #24912), are not shown as closed anywhere in the record.

A careful reader should watch for three things next. The first is whether the Ellendale HPC construction financing closes on disclosed terms. The second is the pace of draws under the SEPA and the new ATM. The third is whether the CIM accordion is exercised, since that would trigger the additional warrants (claim #6031).

The record supplied ends mid-entry in September 2024, so this read covers roughly three years of activity and cannot speak to anything after the SEPA and ATM filings of that month. Two 2006 entries (claims #21056 and #21058, referencing UNICORN and SOCRATES technology) do not fit the data-center narrative that follows and appear to belong to a different predecessor or entity; their linkage requires explanation before they are read as part of this history.

The cadence moves from modest to continuous. The 2022 listing is itself recorded inconsistently: a $75M firm-commitment offering (claim #14370, #14041), an $8.57M figure (claim #14065), and an 8,000,000-share raise at $5.00 for $40M (claim #14007) that netted $36M (claim #24799). From there the company leaned first on regional bank debt, which it framed as non-dilutive (claims #23493, #21492), including the $15M Starion facility (claim #13804) and a $20M follow-on (claim #13310). By mid-2023 the mix shifted toward equity: a $175M shelf (claim #12544), a $125M ATM (claim #12397) that was exhausted by January 2024 (claim #21552), then 2024's dense layering of Series E and F preferred, the Yorkville prepaid advance of roughly $42.1M (claim #6106), CIM debt carrying warrants for 6,300,449 shares (claim #6029), a second $125M ATM (claim #5778), and a $250M SEPA (claim #5756).

That 2024 sequence reads less as a single financing plan than as successive bridges. The AI Bridge loan was prepaid with stock (claim #6137), the B. Riley term loan was terminated early (claim #6291), the Yorkville conversion cap was later eliminated (claim #5837), and CIM required a waiver to lend more (claim #5760). Each instrument appears to have been arranged partly to accommodate the one before it, and the warrant- and conversion-linked structures carry dilution that is deferred rather than avoided.

On use of proceeds, the data is thin. Ellendale construction is the consistent stated destination (claims #13326, #26120), but project-level financing for the hyperscaler lease was described as in advanced discussions in both April and August 2024 (claims #3237, #24912) without a recorded close in this file.

A careful reader should watch whether that project financing is actually executed, how much of the SEPA and second ATM is drawn, the conversion activity on Series F (claim #5834) and the Yorkville notes, and whether the 2022 IPO figures are reconciled.

The disclosed record traces a shift from small, lender-anchored financing toward a layered and increasingly equity-linked capital stack. The two 2006 entries (claims #21056, #21058) describe unrelated technology contracts and appear to belong to a different entity; they require explanation before being read as part of this history. Several 2022 entries also record share counts as dollar amounts (claims #14079, #14080, #14065), and the IPO is variously stated at $75M (claim #14370) and $40M (claim #14007), with $36M net (claim #24799). The record ends mid-entry in September 2024.

Through 2022 the posture was consistent. After the uplisting and one-for-six reverse split (claim #26010), management emphasized non-dilutive regional bank debt (claim #23493): the $7.5M Vantage loan (claim #21459) and the $15M Starion facility (claim #13804). The 2023 turn introduced a $175M shelf (claim #12544), the 9% B. Riley term loan (claims #12565, #12517), and a $125M ATM (claim #12397) that was used heavily, with $64.5M issued (claim #21542) and the program reported complete by January 2024 (claim #21552).

The 2024 pattern reads less as a planned sequence and more as successive layering. The AI Bridge loan (claim #6292) was settled partly in stock (claim #6137). The B. Riley loan was terminated early (claim #6291). The Series E dividend is stated at 8% (claim #13312) and later at 9% (claim #6136); that revision is not discussed. The YA II prepaid advance (claim #6078) originally carried a $16M monthly conversion limit (claim #6111), which was later eliminated (claim #5837). In a single quarter the company added CIM debt with warrants (claims #5973, #6029), Series F convertible preferred (claim #5833), a $160M private placement (claim #5783), a $250M SEPA (claim #5756), and a further $125M ATM (claim #5754).

On use of proceeds, debt was explicitly tied to Ellendale (claim #13326). Project-level financing for the hyperscaler lease, however, first appears in April 2024 (claim #3237) and is still described as "advanced discussions" in August (claim #24912), with no closing on record; that item is delayed.

A careful reader should watch several things next. The first is whether Ellendale project financing closes and what it replaces. The second is the pace of YA conversions now that the limit has been removed. Beyond that, the reader should track draws under the SEPA and the new ATM, the combined warrant and convertible-preferred overhang, and the fixed dividend burden on the Series E and Series F preferred.

Applied Digital's record describes a company that started with modest bank debt and moved, within about thirty months, to a layered stack of instruments that increasingly settle in common stock. The 2006 entries (claims #21056, #21058) appear to describe an unrelated technology business and require explanation before being read as part of this history. The 2021 Series C and D preferred raises were small, $660,000 and $1.3M per claims #14500 and #14501. The IPO was announced at up to $75M (claim #14041) but priced at $40M (claim #14007) and netted $36M (claim #24799), alongside a one-for-six reverse split (claim #26010). The opening raise therefore came in at roughly half its stated size. Through 2022 and early 2023 management described its funding as non-dilutive regional bank debt (claims #23493, #21492): the $7.5M Vantage loan (claim #21459), then Starion loans of $15M and $20M (claims #13804, #13326). That framing does not survive mid-2023. A $175M shelf (claim #12544) and a $125M ATM (claim #12397) followed the 9% B. Riley facility (claim #12517), and the ATM was fully used by January 2024 (claim #21552). The 2024 mix looks less like a sequenced plan and more like overlapping fixes. An AI Bridge note (claim #6292) was later prepaid in stock (claim #6137). The Yorkville prepaid advance of $42.1M (claim #6078) carried a monthly conversion cap (claim #6111) that was then removed (claim #5837). Within roughly four months the company also added Series E and Series F preferred (claims #6119, #5833), a CIM facility with warrants and an accordion (claims #5973, #5975), a $250M SEPA (claim #5756), a second $125M ATM (claim #5754), and a $160M private placement (claim #5783). Use of proceeds is thinly documented. Claim #13807 references loan uses without specifics, and project-level financing for the 400 MW hyperscaler lease and Ellendale HPC construction is described as in discussion from April through August 2024 (claims #3237, #24912, #5867) without a recorded close. The record as supplied ends in September 2024, so later deployment cannot be assessed here. A careful reader should watch for whether the Ellendale project financing closes, the pace of Yorkville conversions now that the cap is gone, draws on the SEPA and the new ATM, accordion-triggered warrants (claim #6031), and conversion of the Series F into common (claim #5834).

Applied Digital's financing record falls into three phases, and the instruments become more complex and more dilutive across them. Two 2006-dated claims describing SOCRATES and UNICORN technology (claims #21058, #21056) do not fit the company's later data-center profile and should be read as possibly belonging to a predecessor entity. The record supplied also ends partway through September 2024 and contains no insider transactions, so this read covers only the financing side. Several amounts appear normalized inconsistently: share counts are logged as dollars (claim #14080), and percentages are logged as dollar figures (claim #12517). Comparisons below rely on the clearly stated figures.

The 2022 phase was modest and leaned on debt. It consisted of an IPO netting $36M (claim #24799), the $7.5M Vantage loan (claim #21459), and Starion facilities of $15M and $20M (claims #13804, #13310). Management described this regional bank borrowing as non-dilutive (claim #21492). In 2023 the company moved to market-priced capital: a $175M shelf (claim #12544), the 9% B. Riley facility of $50M (claims #12514, #12517), and a $125M ATM (claim #12397) that was completed by January 2024 (claim #21552).

The 2024 layers point more toward patching than toward a single plan. In the space of months the company added the AI Bridge loan, later prepaid in stock (claim #6137); Series E and Series F preferred of $50M each (claims #6119, #5833); the Yorkville prepaid advance of $42.1M (claim #6078); the CIM facility with 6,300,449 verified warrants (claim #6029); a $250M SEPA (claim #5780); and a second $125M ATM (claim #5754). Yorkville's monthly conversion limit of $16M (claim #6111) was then eliminated (claim #5837), which loosened a dilution constraint. The B. Riley loan was reported terminated early (claim #6291) and then appears again as a loan received in June (claim #6049), which requires explanation.

On use of proceeds, the earlier debt ties directly to the Ellendale expansion (claim #13326). The larger stated goal, project-level financing for the 400 MW hyperscaler lease (claim #3237) and Ellendale HPC construction, is described as being in "advanced discussions" in August 2024 (claims #24912, #5867). No closing appears in the record.

A careful reader should watch four things: whether the Ellendale HPC construction financing actually closes; how much is drawn under the SEPA and the second ATM; whether CIM accordion draws trigger the contingent warrants (claim #6031); and how the combined Series E and Series F preferred dividends affect cash flow.

The record shows a company whose financing has grown faster and more complex than its operating base, and it should be read with two caveats. First, it ends mid-entry in September 2024, so nothing after that date can be assessed. Second, the 2006 entries on SOCRATES and UNICORN technology (claims #21058, #21056) do not connect to the later data-center business and require explanation before they are treated as part of this history.

Dilution moved through three phases. The first was a small, conventional base: the $40M IPO at $5.00 (claim #14007), netting $36M (claim #24799), alongside the one-for-six reverse split (claim #26010). The second, through 2023, leaned on regional bank debt, including the $15M and $20M Starion loans (claims #13804, #13326). Equity came through the $125M Craig-Hallum ATM, which was drawn down quickly: $64.5M by October 2023 (claim #21542) and completed by January 2024 (claim #21552). The third phase, in 2024, compressed many instruments into a few months. These include the Yorkville prepaid advance of $42.1M (claim #6078), CIM borrowings with warrants on 6,300,449 shares (claim #5972), Series E and Series F preferred (claims #6119, #5757), a $250M SEPA (claim #5756), a $160M PIPE (claim #5782) and a fresh $125M ATM (claim #5754).

The mix reads less like a planned capital stack than like successive layering. The B. Riley term loan at 9% (claims #12565, #12517) was terminated early within months (claim #6291). The Yorkville $16M monthly conversion cap (claim #6111) was later eliminated (claim #5837). The CIM note needed a waiver to permit $20M of further borrowing (claim #5760). Each step is individually explicable, but together they show terms being revised as needs arose. No insider transactions appear in the record, so that dimension cannot be assessed.

On use of proceeds, the early bank loans tie to the Ellendale facility (claims #13326, #13807). The larger thesis, project-level financing for the 400 MW hyperscaler lease (claim #3237) and Ellendale HPC construction, remained described as discussions in August 2024 (claims #24912, #5867) and is not shown as closed.

A careful reader should watch four things. The first is the pace of SEPA and ATM drawdowns. The second is conversion of the Series F into roughly 7.6M shares (claim #5834). The third is contingent warrants tied to accordion draws (claim #6031). The fourth is whether the Ellendale project financing actually closes, which would mark a shift from corporate-level patching to asset-level funding.

Applied Digital's record shows a company that has moved from cheap, narrow sources of capital to expensive, layered ones in about two and a half years, with dilution speeding up sharply in 2024. Two 2006 items about SOCRATES and UNICORN technology (claims #21058, #21056) do not fit the later business and need to be reconciled as to which entity they belong to. The record also ends in September 2024, so this reading covers only that window.

After the IPO priced at $40M (claim #14007) and netted roughly $36M (claim #24799), alongside a one-for-six reverse split (claim #26010), management leaned on regional bank debt. That included the $7.5M Vantage loan (claim #21459), the $15M Starion facility (claim #13804) and a further $20M from Starion tied to the Ellendale expansion (claim #13326). Management explicitly called these non-dilutive (claim #21492). That framing is no longer discussed after mid-2023. The $175M shelf (claim #12544), the $50M B. Riley term loan at 9% (claims #12514, #12517) and a $125M ATM that was fully used by January 2024 (claims #12397, #21552) mark the turn toward equity.

The 2024 mix reads more as sequential patching than a single coherent plan. It includes an unsecured $20M bridge (claim #6292) and early termination of the B. Riley loan (claim #6291). It also includes a Yorkville prepaid advance of about $42.1M (claim #6078), which was partly settled in common stock (claim #6137) and later had its conversion limits removed (claim #5837). On top of that sit CIM borrowings with warrants (claims #5973, #6029), Series E and Series F preferred (claims #6119, #5757), a $250M SEPA (claim #5756) and a $160M PIPE (claim #5782). Each instrument adds a separate path to dilution, priced on different terms.

The stated use of proceeds tracks reasonably well at Ellendale for the early bank debt. Project-level financing for the 400 MW hyperscaler lease (claim #3237) and construction financing for Ellendale HPC (claims #24912, #26120) are described as in discussion across several filings. The record does not show either one closing, so that gap requires explanation.

A careful reader should next watch four things. The first is whether Ellendale project financing actually closes, which would replace corporate-level equity. The second is the pace of SEPA draws and PIPE and Series F conversions. The third is the accordion warrants, which depend on further CIM draws (claim #6031). The fourth is insider transactions, which are absent from this record and so cannot be assessed against the dilution pattern.

The record covers the May 2022 uplisting through September 2024 and appears truncated mid-entry, so this read is limited to that window. Two 2006 entries, the SOCRATES and UNICORN technology items (claims 21058 and 21056), do not clearly belong to the current data-center business and require explanation before being treated as part of this history. No insider transactions appear in the record.

The IPO was modest, with $36M in net proceeds per claim 24799, accompanied by a one-for-six reverse split per claim 26010. The company then leaned first on regional bank debt. That included the $7.5M Vantage term loan (claim 21459), the $15M Starion facility (claim 13804) and a further $20M Starion loan (claim 13310), framed explicitly as non-dilutive (claim 21492). That posture changed in 2023. The company filed a $175M shelf (claim 12544), signed a $125M ATM (claim 12411) and took a $50M B. Riley loan at 9% (claims 12514 and 12517). The ATM was reported complete by January 2024 (claim 21552), about seven months after signing, which marks the point where equity issuance became a primary funding channel.

In 2024 both the pace and the number of instruments increased. Within roughly four months the company added a $20M unsecured bridge (claim 6292) that was later prepaid in stock (claim 6137) and a $50M Series E preferred program (claim 6119). It also took $40M and $10M Yorkville prepaid advances (claims 6079 and 6080), whose conversion limits were later eliminated (claim 5837). The CIM facility followed, with $15M initially borrowed and warrants on 6.3M shares (claims 5973 and 5972). Then came a $50M Series F convertible preferred (claim 5757), a $250M SEPA (claim 5756), a second $125M ATM (claim 5754) and a $160M Goldman-backed PIPE (claim 5782). Several of these carry equity-linked conversion features. The mix reads more as filling successive gaps than as one capital plan, although the stated anchor has stayed consistent: project-level financing for the 400 MW hyperscaler lease and Ellendale HPC (claims 3237 and 21574).

The record states intended uses, such as the Ellendale expansion in claim 13326, but says little about actual deployment. Ellendale construction financing was still described as in advanced discussions in August 2024 (claim 24912), about four months after project-level financing was first cited.

A careful reader should watch whether the Ellendale project financing closes. It is also worth tracking how much of the SEPA and the second ATM is drawn, and how quickly the Yorkville and Series F instruments convert into common stock.

The record visible here ends mid-entry in September 2024, and it contains no insider transactions, so this read covers financing behavior only and says nothing about insider selling. The two 2006 entries describing UNICORN and SOCRATES technology (claims #21056 and #21058) do not fit the data-center business that follows, and their attribution to this company requires explanation. The IPO record is also internally inconsistent. Claims #14041 and #14370 describe a $75M offering, claim #14007 records 8,000,000 shares at $5.00 for $40M, and claim #24799 gives net proceeds of $36M.

The trajectory moves in three distinct phases. Through 2022 and early 2023, management described its funding as non-dilutive regional bank debt (claims #21492 and #23493), drawing on Vantage's $7.5M (claim #21459) and Starion's $15M and $20M facilities (claims #13804 and #13310). Mid-2023 marks the turn. A $50M B. Riley term loan at 9% (claims #12514 and #12517) arrived alongside a $125M ATM (claim #12397), which had issued $64.5M by October (claim #21542) and was reported complete in January 2024 (claim #21552). In 2024, the mix widened sharply. It came to include an unsecured $20M AI Bridge loan (claim #6292, not yet verifiable) that was later prepaid in stock, a $42.1M Yorkville prepaid advance (claim #6078) whose monthly conversion limit was subsequently eliminated (claim #5837), and CIM debt with warrants for 6,300,449 shares (claim #5972). The same year added $50M Series E and Series F preferreds (claims #6119 and #5757), a $160M PIPE (claim #5782), a $250M SEPA (claim #5756), and a fresh $125M ATM (claim #5754). This layering reads less as a single capital plan than as successive instruments added as each prior source tightened, although the PIPE suggests improving institutional access.

Use of proceeds is only partly traceable. Ellendale expansion is tied to specific loans (claim #13326). Project-level financing for the 400 MW hyperscaler lease (claim #3237) and Ellendale HPC construction financing, described as in advanced discussions (claim #24912), are not shown as closed within this record.

A careful reader should watch for four things: whether the Ellendale construction financing closes on project-level, non-recourse terms; how fast the SEPA and the new ATM are drawn; the conversion of the 7,598,714 Series F-linked shares (claim #5834); and whether the contingent accordion warrants (claim #6031) are triggered, since each of these would add to dilution beyond what is currently outstanding.

The disclosed record moves through three phases, and the shift between them is its most informative feature. Through 2022 and early 2023, financing leaned on regional bank debt that management framed as non-dilutive (claims #23493, #21492). That debt included the $7.5M Vantage Bank Texas term loan (claim #21459), the $15M Starion facility (claim #13804) and a further $20M Starion loan for the Ellendale expansion (claim #13310), layered over IPO net proceeds of $36M (claim #24799). From mid-2023 the mix turned toward equity at increasing scale. A $125M ATM with Craig-Hallum (claim #12397) had produced $64.5M of issuance by October 2023 (claim #21542) and was reported complete by January 2024 (claim #21552). By mid-2024 the instruments became layered and more structured. They included Series E preferred (claim #6119), a $42.1M Yorkville prepaid advance with conversion terms (claim #6106), CIM borrowings carrying warrants (claims #5973, #6029), Series F convertible preferred (claim #5757), a $160M PIPE (claim #5782), a $250M SEPA (claim #5756) and a renewed $125M ATM (claim #5754). Dilution therefore accelerated rather than steadied. The 2024 period reads more as several equity-linked channels opened at once than as a single sequenced plan. Several instruments were reshaped soon after they were put in place. The B. Riley term loan was terminated early (claim #6291), the AI Bridge note was prepaid in common stock (claim #6137), and the Prepaid Advance conversion limitation was eliminated (claim #5837). Each change can be explained on its own, but together they suggest terms were adjusted to near-term liquidity needs. On use of proceeds, the record repeatedly ties capital to Ellendale. Project-level financing for a 400 MW hyperscaler lease was flagged in April 2024 (claim #3237), yet Ellendale HPC construction financing was still described as in discussion in August 2024 (claims #24912, #26120). Within this record, that financing is described but not closed. No insider transactions appear in the data provided. The record as supplied ends mid-entry in September 2024, so later events are not assessed. The 2006 claims (#21056, #21058) appear to relate to a predecessor or unrelated entity and require explanation before being read into this history. Several amounts are also recorded in inconsistent units or duplicated across claims. A sophisticated reader should watch four things next: whether the Ellendale construction financing closes and on what terms, the pace of draws on the SEPA and ATM, conversion activity on the Series F and Yorkville instruments, and whether the CIM accordion and its associated warrants (claims #5975, #6031) are used.

The record runs through September 2024 and is cut off mid-entry, and it contains no insider transactions, so this read covers financing structure only. The two 2006 claims (#21056, #21058) appear to relate to a predecessor business and require explanation before being read as part of this company's capital history. Several entries also carry figures whose units are inconsistent: share counts are recorded as dollar amounts (#14079, #14080), and dollar amounts appear truncated to $8, $15 or $36. The IPO sequence needs reconciling as well, since an offering of up to $75 million (#14041) priced at $40 million (#14007) and netted $36 million (#24799).

The trajectory runs from conservative to layered. Through early 2023 the company leaned on regional bank term debt, including the $7.5 million Vantage loan (#21459), the $15 million Starion facility (#13804) and a $20 million Starion expansion loan (#13310), and it described this borrowing as non-dilutive (#21492). From mid-2023 the mix shifted. It took a $50 million B. Riley loan at 9% (#12514, #12517), and a $125 million ATM (#12397) had produced $64.5 million (#21542) by October and was completed by January 2024 (#21552). The B. Riley loan was then terminated early (#6291).

The 2024 financing reads less like a sequenced plan than like successive patching. An unsecured $20 million AI Bridge loan (#6292) was prepaid in common stock (#6137). A $42.1 million Yorkville prepaid advance (#6078) followed, with a $16 million monthly conversion limit (#6111) that was later eliminated (#5837), plus shares issued to Yorkville for consent (#5988). The CIM borrowing carries tranched warrants (#6029, #6031). Within roughly four months the company also added Series E preferred (#6119), Series F convertible preferred covering 7,598,714 shares (#5834), a $250 million SEPA (#5756), a $160 million PIPE (#5782) and a fresh $125 million ATM (#5754). Each instrument is individually explainable, but together they show dilution accelerating and an increasing reliance on counterparties who are paid in equity.

On use of proceeds, the specific stated use of the 2022 Starion advances (#13807) is not detailed in the record. Ellendale HPC construction financing is described as "in advanced discussions" repeatedly (#24912, #26120, #5867), and project-level financing for the hyperscaler lease (#3237) is stated without a closing recorded here.

A careful reader should next watch for three things. The first is the pace of Yorkville and SEPA conversions now that the conversion cap has been removed. The second is whether the CIM accordion is drawn, since a draw triggers additional warrants. The third is whether Ellendale project financing actually closes, as distinct from being discussed.

Applied Digital's record moves through three distinct financing regimes, and the shift between them carries more information than any single transaction. The 2006 entries on UNICORN and SOCRATES technology (claims #21056, #21058) appear to describe a predecessor or unrelated entity, and their link to the current business requires explanation. The substantive history starts with small convertible preferred issues in 2021, and even these were later restated: the Series D figure moves from $1.3M (claim #14501) to $1.38M (claim #14356). The IPO went the same way. It was filed for up to $75M (claim #14370), priced at 8,000,000 shares at $5.00 for $40M (claim #14007), and netted $36M (claim #24799). It sat alongside a one-for-six reverse split (claim #26010). Through 2022 and early 2023 management described regional bank debt as deliberately non-dilutive (claim #23493). That debt included the $7.5M Vantage loan (claim #21459), the $15M Starion facility (claim #13804) and a further $20M Starion term loan (claim #13310). That framing became harder to hold from mid-2023. The $50M B. Riley loan at 9% (claims #12565, #12517) arrived alongside a $175M shelf (claim #12544) and a $125M ATM (claim #12397), which was fully drawn by January 2024 (claims #21542, #21552). The B. Riley loan was terminated early (claim #6291). From May to September 2024 the mix fragmented. In roughly four months the company used an unsecured bridge later repaid in stock (claim #6137), a $42.1M Yorkville prepaid advance (claim #6078), and Series E and Series F preferred at $50M each (claims #6119, #5833). It also added a CIM loan carrying 6.3M warrants (claim #6029), a $250M SEPA (claim #5756), a $160M PIPE (claim #5782) and a fresh $125M ATM (claim #5754). The elimination of the Yorkville monthly conversion limit (claim #5837) loosened a dilution control that had been negotiated only months earlier. This sequence reads as successive patching across instruments with different seniorities, not as a single capital plan. Project-level financing for Ellendale HPC was described as in discussion in April 2024 (claim #3237) and again in August and September (claims #24912, #5867). It is not shown as closed anywhere in the record, so stated and actual deployment cannot be reconciled from this data. The record also contains no insider transactions and is truncated after September 2024. A careful reader should next look for confirmation that the Ellendale construction financing closed. They should also track the pace of Yorkville and SEPA conversions, warrant issuance tied to CIM accordion draws (claim #6031), and the redemption mechanics of the Series E and Series F preferred.

The record as supplied ends mid-entry in September 2024, so this read covers roughly three years of activity and cannot speak to anything after that. Two 2006 entries (claims #21056 and #21058) describe technology programs unrelated to the current business and appear to belong to a predecessor or a different registrant; they are set aside here. Several figures are also recorded in inconsistent units, so the following relies on the claims that state full amounts.

The trajectory moves from modest to large and from simple to layered. After small convertible preferred issues in 2021, the company completed a $40M IPO (claim #14007), which netted about $36M (claim #24799). Through 2022 and early 2023 it leaned on regional bank debt, including the $7.5M Vantage loan (claim #21459) and Starion facilities of $15M and $20M (claims #13804, #13310). It then stepped up to a $50M B. Riley term loan at 9% (claims #12565, #12517) and a $125M ATM (claim #12397), which was fully used by January 2024 according to claims #21542 and #21552.

After that the mix becomes harder to read as a single plan. Within about four months of 2024 the record shows an early exit from the B. Riley loan (claim #6291), a $20M AI Bridge loan later repaid in common stock (claim #6137), Series E preferred (claim #6119), a $42.1M Yorkville prepaid advance (claim #6078), a CIM facility with warrants attached (claims #5973, #6029), a Roth sales agreement (claim #6081), Series F convertible preferred (claim #5757), a $160M PIPE (claim #5782), a $250M SEPA (claim #5756), and a renewed $125M ATM (claim #5754). The Yorkville monthly conversion cap of $16M (claim #6111) was later removed (claim #5837). That change shifts more control over dilution timing to the holder and requires explanation. Taken together, the sequence looks more like layered stopgap funding than a planned capital structure, with debt repeatedly retired or refinanced through equity-linked instruments.

The stated use of proceeds has stayed consistent: Ellendale and the HPC buildout. The deployment vehicle has not kept pace with it. Project-level financing was first raised in April 2024 (claim #3237) and was still described as advanced discussions in August 2024 (claims #24912, #26120). The record does not show it closing.

A careful reader should watch four things. First, whether the Ellendale project financing actually closes. Second, the size and pace of SEPA and ATM drawdowns. Third, Series F conversions into roughly 7.6M shares (claim #5834). Fourth, any draws on the CIM accordion, which would trigger the contingent second warrant tranche (claim #6031).

The record shows a company whose financing grew from small preferred placements to a layered capital stack within about three years. The Series C and D convertible preferreds of roughly $660,000 and $1.3 million (claims #14500, #14501) gave way to a $40 million IPO at $5.00 per share (claims #14007, #14000), alongside a one-for-six reverse split (claim #26010). After the IPO, the company relied first on regional bank debt: the Vantage Bank $7.5 million term loan (claim #21459) and the Starion facilities of $15 million and $20 million (claims #13804, #13310). Management described this as non-dilutive (claims #23493, #21492). That framing held only into mid-2023. From then on, dilution accelerated. The $125 million Craig-Hallum ATM (claim #12397) produced at least $64.5 million and then $23.1 million and was reported complete (claims #21542, #21552). In mid-2024 the instruments multiplied: Series E preferred (claim #6119), the Yorkville prepaid advance of about $42.1 million convertible into common (claim #6078), the CIM loan carrying warrants for about 6.3 million shares (claim #5972), Series F convertible preferred (claim #5757), a $160 million PIPE (claim #5782), and a $250 million SEPA (claim #5756). The overall direction is coherent, since each step tracks the capital intensity of the move toward HPC hosting. The timing is less so. The B. Riley loan was terminated early (claim #6291), the AI Bridge note was prepaid in common stock (claim #6137), and conversion limits on the prepaid advance were removed (claim #5837). These look like near-term patching rather than a planned sequence. Comparing announced uses with actual deployment is limited because the stated use of the Starion loan (claim #13807) is not detailed in the record. Ellendale construction financing was described as in advanced discussion (claims #24912, #26120), but no completed project-level facility appears in these claims, so that item is still pending. The record also requires explanation in places. The 2006 UNICORN and SOCRATES entries (claims #21056, #21058) appear to describe a different issuer. Several figures are inconsistently scaled, for example IPO share counts recorded as dollar amounts (claim #14079). No insider transactions are disclosed, and the claims end in September 2024. A careful reader should watch three things next: actual draws under the SEPA and ATM, conversion activity on the Series F and the Yorkville notes, and whether the Ellendale project debt closes or is no longer discussed.

Applied Digital's record shows dilution that started modestly and accelerated sharply in mid-2024. The two 2006 entries (claims #21056 and #21058) appear to describe a different business and require explanation before being read as part of this history. The record also stops in September 2024 and contains no insider transactions, so this read covers only the financing side through that date. The 2022 IPO netted $36M (claim #24799), well below the $75M offering amount first described (claim #14370). That year's follow-on capital came from regional bank debt, which management presented as non-dilutive (claim #23493). In 2023 the company moved to costlier and more equity-linked sources: a 9% B. Riley term loan of $50M (claims #12565, #12517), an 8% Series E preferred (claim #13312), and a $125M ATM (claim #12397). That ATM appears to have been fully used by January 2024 (claim #21552). The pace then picked up considerably. Within roughly five months the company added an unsecured bridge loan of $20M (claim #6292) that was repaid partly in common stock (claim #6137). It also took a Yorkville prepaid advance of $42.1M (claim #6106), whose monthly conversion cap (claim #6111) was later removed (claim #5837). Other additions were the CIM facility with warrants on 6,300,449 shares (claim #6029), a $50M Series F convertible into 7,598,714 shares (claim #5834), a $250M SEPA (claim #5756), a $160M PIPE (claim #5782), and a renewed $125M ATM (claim #5754). The Series E dividend is listed at 8% in claim #13312 and 9% in claim #6136, a revision that the record does not explain. Taken together, the mix looks less like a planned capital structure than a series of layered fixes, each priced by whoever could close fastest. Debt was refinanced or terminated early (claims #6291, #6050), and the conversion protections attached to the Yorkville instrument were removed. Use of proceeds can only partly be checked. Ellendale-directed debt (claim #13326) and the land purchase (claim #12622) line up with stated intent. Project-level financing for the 400 MW hyperscaler lease, however, was first cited in April 2024 (claim #3237) and was still described as being in discussion in August (claims #24912, #26120), so it should be treated as delayed pending a documented close. Going forward, a careful reader should track draws under the SEPA, conversions of the Series F, and use of the CIM accordion along with its contingent warrants on 5,559,220 shares (claim #6031). The most important item is whether Ellendale construction financing closes at the project level or ends up being funded by further corporate equity.

Applied Digital's record describes a company whose financing cadence accelerated sharply once its capital needs shifted from crypto-hosting buildout to high-performance compute. The early phase, from the $40 million IPO (claim #14007, net $36 million per claim #24799) through 2022, relied mostly on modest regional bank debt: the $7.5 million Vantage loan (claim #21459) and the $15 million and $20 million Starion facilities (claims #13804, #13310). Management framed this as non-dilutive (claim #21492). From mid-2023 the mix changed. The $50 million B. Riley term loan at 9% (claims #12514, #12517) was followed within weeks by a $125 million ATM (claim #12397), which was fully drawn by January 2024 (claim #21552). The B. Riley loan was then terminated early (claim #6291). By mid-2024 the company was layering instruments in quick succession: Series E preferred (claim #6119), the $42.1 million Yorkville prepaid advance (claim #6078), the CIM facility with $110 million in subsequent tranches and a $75 million accordion tied to warrants (claims #5974, #5975, #5972), Series F convertible preferred (claim #5757), the $160 million PIPE (claim #5782), a $250 million SEPA (claim #5780), and a fresh $125 million ATM (claim #5754). Several of these settle in common stock, including the AI Bridge note prepaid in shares (claim #6137). The elimination of the prepaid advance's conversion limitation (claim #5837) widens the dilution path rather than narrowing it. Taken together, the pattern reads less as a staged capital plan than as successive patching, with each instrument bridging to the next. On use of proceeds, the record is thin. The stated purpose of the Starion proceeds (claim #13807) is not itemized. The Ellendale HPC construction financing appears repeatedly as being in advanced discussion (claims #24912, #26120, #5867), but no closed facility is recorded in this dataset, so whether project-level debt replaced corporate-level equity cannot be confirmed. The record contains no insider transactions. The 2006 claims (#21056, #21058) appear unrelated to the current business and require explanation. The IPO figures conflict across claims (#14370, #14065, #14007). The data also ends in September 2024, so later events are outside this read. A reader should watch for whether the Ellendale construction financing closes, how much SEPA and ATM capacity is actually used, and how Series F conversions and CIM warrant exercises affect the share count.

Two caveats frame this read. The 2006 entries describing UNICORN and SOCRATES® technology (claims #21056, #21058) do not match Applied Digital's disclosed business and appear to belong to another entity, so they are set aside. The record also ends partway through September 2024 and contains no insider transactions, so this assessment covers only the period from the 2022 IPO through the September 2024 financing cluster. The IPO figures themselves require reconciliation. A $75,000,000 offering size was stated (claims #14370, #14041), but pricing came in at 8,000,000 shares for $40,000,000 (claim #14007) and net proceeds of $36,000,000 (claim #24799). In its first eighteen months the company leaned on regional bank debt, including the Vantage $7.5M term loan (claim #21459) and Starion facilities of $15M and $20M (claims #13804, #13326), and it described this approach explicitly as non-dilutive (claim #21492). That posture changed in mid-2023. A $125,000,000 ATM (claim #12397) was drawn for $64.5M (claim #21542) and was reported complete by January 2024 (claim #21552). It was followed by a dense sequence of instruments in mid-2024: the AI Bridge note, partly settled in common stock (claim #6137); the Yorkville prepaid advance of $42,131,579 (claim #6078); and CIM borrowing with 6,300,449 warrants attached (claim #6029). By September came a $250,000,000 SEPA (claim #5756), a $160,000,000 PIPE (claim #5782), Series F preferred convertible into 7,598,714 shares (claim #5834), and a renewed $125,000,000 ATM (claim #5754). The overlap of these facilities, and the early termination of the B. Riley loan (claim #6291), read as layered patching around a capital need rather than a single sequenced plan. Terms also shifted. The Series E dividend rose from 8% at designation (claim #13312) to 9% at offering (claim #6136), and the Yorkville monthly conversion limit of $16,000,000 (claim #6111) was later eliminated (claim #5837). On deployment, project-level financing for the 400 MW hyperscaler lease was first raised in April 2024 (claim #3237) and was still described as being in discussion in August (claims #24912, #5867). That item is delayed relative to the equity actually raised. A sophisticated reader should watch the pace of Yorkville conversions now that the cap has been removed, the size of SEPA draws, whether the second CIM warrant tranche is triggered by accordion borrowing (claim #6031), and whether the Ellendale construction financing closes or is no longer discussed.

The disclosed record runs from 2021 to September 2024 and is truncated mid-entry, so this read covers only that window. The two 2006 entries on UNICORN and SOCRATES technology (claims #21056, #21058) do not obviously belong to the current business and require explanation before being treated as part of its history. No insider transactions appear, so insider selling cannot be assessed. The early phase was modest in scale. The IPO was marketed at up to $75 million (claim #14041) but priced at $40 million (claim #14020), netting roughly $36 million (claim #24799). It was paired with regional bank term loans of $7.5 million (claim #21459), $15 million (claim #13804) and $20 million (claim #13310), which management framed as non-dilutive (claim #21492). From mid-2023 the instruments widened and the cadence accelerated. A $125 million ATM (claim #12397) was described as complete by January 2024 (claim #21552). The $50 million B. Riley loan at 9% (claims #12514, #12517) was terminated early (claim #6291). Between May and September 2024 the company then layered on several new instruments. These included a $42.1 million Yorkville prepaid advance (claim #6078), whose monthly conversion limit was later eliminated (claim #5837), and a CIM loan carrying warrants on 6.3 million shares (claim #6029). It also issued Series E and Series F preferred (claims #6119, #5757), sold a $160 million PIPE (claim #5782), signed a $250 million SEPA (claim #5756) and opened a fresh $125 million ATM (claim #5754). The dilution trajectory steepens sharply in that period. The 2022-23 mix reads as coherent, with bank debt funding site build-out. The 2024 sequence reads more as bridging. Warrant-bearing loans, consent shares issued to Yorkville (claim #6032) and an AI Bridge note prepaid in stock (claim #6137) sit alongside project-level financing that was repeatedly described as under discussion (claims #3237, #24912, #5867) but is not shown closing in this record. On use of proceeds, the Ellendale expansion loan (claim #13326) is consistent with the land purchase that followed (claim #12622). Beyond that the record does not map deployment against stated purposes. The Highland Digital fund (claim #3169) is no longer discussed after its launch. A sophisticated reader should watch four things. The first is whether Ellendale HPC construction financing actually closes. The second is the pace of draws on the SEPA and the new ATM. The third is conversion of the Series F into about 7.6 million shares (claim #5834) and of any remaining Yorkville balance. The fourth is whether the CIM accordion and its contingent warrant tranche (claims #5975, #6031) are triggered.

The record shows a company whose financing grew in size and complexity faster than any single instrument could support. It begins with small convertible preferred tranches (the $1.3M Series D and $660K Series C, per claims #14501 and #14500), then an uplisting IPO. The intended IPO size was $75M (claims #14370 and #14041), but it priced at $40M (claim #14007) and netted $36M (claim #24799). The IPO was paired with a one-for-six reverse split (claim #26010) and modest regional bank debt. Through 2022 and into early 2023, management described its funding as non-dilutive regional lending (claims #23493 and #21492), including the $15M and $20M Starion facilities (claims #13804 and #13310). That framing is no longer discussed after mid-2023. The $125M ATM with Craig-Hallum (claim #12397) was drawn quickly, with $64.5M reported by October 2023 (claim #21542) and completion noted by January 2024 (claim #21552). From that point, equity dilution became the primary funding channel. The 2024 mix reads more as layered patching than a sequenced plan. In roughly four months the company added Yorkville prepaid advances convertible into common (claim #6078), a CIM term loan carrying warrants (claims #6026 and #6029), Series E and Series F preferred (claims #6119 and #5833), a second ATM with Roth (claim #6081), a $250M SEPA (claim #5780) and a $160M PIPE (claim #5782). Two details suggest these terms were negotiated under pressure: shares were issued to Yorkville in exchange for consent (claim #6032), and the prepaid advance's conversion limitation was later eliminated (claim #5837). The $20M AI Bridge loan (claim #6292) was settled in stock (claim #6137) and the B. Riley loan was terminated early (claim #6291), so short-dated debt was routinely refinanced into equity. Use of proceeds is difficult to test. Claim #13807 references loan use without detail, and the record of Ellendale HPC construction financing stops at advanced discussions (claims #24912 and #26120), with no closing on record. The record contains no insider transactions, and the dataset ends in September 2024, so later events cannot be assessed. The 2006 SOCRATES and UNICORN entries (claims #21058 and #21056) appear to belong to a different entity and require explanation. A careful reader should watch for whether the Ellendale construction financing closes, how much of the SEPA and second ATM is actually drawn, and whether further CIM accordion draws trigger the contingent warrants (claim #6031).

The record runs through roughly September 2024 and breaks off mid-entry, so this read covers the company's first three years as a listed issuer and cannot speak to anything later. The two 2006 entries (claims 21056, 21058) appear to describe an unrelated predecessor and are set aside. Many 2024 events are logged two or three times across filings, so the claim count overstates the number of distinct transactions.

The trajectory is one of escalating scale and declining cost discipline. The company listed on a modest $40M IPO at $5.00 per share (claim 14007), netting about $36M (claim 24799). It then leaned on regional bank debt, the $7.5M Vantage loan (claim 21459) and Starion facilities of $15M and $20M (claims 13804, 13310), which management framed as non-dilutive (claim 21492). By mid-2023 the mix shifted to costlier capital: the $50M B. Riley loan at 9% (claims 12514, 12517) and a $125M ATM (claim 12397), from which $64.5M and then a further $23.1M were drawn before it was reported complete (claims 21542, 21552).

The first half of 2024 reads more as patching than as a coherent plan. The B. Riley loan was terminated early (claim 6291). The $20M AI Bridge loan (claim 6292, not yet verifiable) was partly settled in stock (claim 6137). Series E preferred (claim 6119), the $42.1M Yorkville prepaid advance with stock-settled conversion (claims 6078, 6111), and a CIM facility carrying warrants (claims 5973, 5972) followed within weeks. The later removal of the prepaid advance's conversion limit (claim 5837) and the $250M SEPA (claim 5780) point to continued reliance on structures that settle in shares. The $160M PIPE (claim 5782) and the $125M CIM senior secured note (claim 5753) mark a move toward larger, more institutional capital, though they sit alongside these instruments rather than replacing them.

On use of proceeds, the early bank loans track to Ellendale build-out (claims 13807, 13326). Project-level financing for the hyperscaler and Ellendale HPC program, however, was described in April 2024 (claim 3237) and remained in advanced discussions in August 2024 (claim 24912). Corporate-level instruments appear to have bridged that gap, and that sequence requires explanation.

A careful reader should look for three things. The first is whether the Ellendale project financing closed and on what terms. The second is how much of the SEPA, the prepaid advance and the Series F convertible (claim 5834) has been converted into common shares. The third is whether later filings show the warrant-bearing CIM structure refinanced or still outstanding.

The record runs from the 2022 uplisting to September 2024 and ends mid-entry at claim 5760, so this read covers that window only. Two 2006 technology items (claims 21056, 21058) and the 2021 Series C and D preferred issuances (claims 14500, 14501, 14355, 14356) appear to predate the current business and require explanation before they are counted as part of its history. No insider transactions appear in the record, so insider selling cannot be assessed. The IPO entries are also inconsistent. A $75 million firm commitment (claims 14370, 14041) gives way to a $40 million pricing (claims 14007, 14000) and $36 million net (claim 24799), and that gap is never reconciled.

The trajectory moves from restraint to acceleration. Through 2022 the company described its funding as non-dilutive regional bank debt (claims 21492, 23493), including the Vantage $7.5 million (claim 21459) and Starion $15 million (claim 13804) loans. In 2023 the mix shifted to costlier private credit, with B. Riley's $50 million at 9% (claims 12565, 12517), and to the first $125 million ATM (claim 12397), which was reported complete by January 2024 (claim 21552). In 2024 cadence compressed sharply. Within about four months the company added an unsecured $20 million bridge later repaid in stock (claims 6292, 6137), Series E preferred (claim 6119), and the $42.1 million Yorkville prepaid advance (claim 6078). It also took on CIM borrowings carrying warrants (claim 6029), a $25 million Roth ATM (claim 6081), $50 million of Series F convertible preferred (claim 5833), a $160 million private placement (claim 5783), a $250 million SEPA (claim 5784), and a fresh $125 million ATM (claim 5778). Several instruments convert at market-linked prices, and the Yorkville monthly conversion cap of $16 million (claim 6111) was later eliminated (claim 5837). That pattern reads more as layered patching to bridge construction than as a single planned capital structure.

On use of proceeds, the Ellendale term loans (claims 13326, 12622) map to stated facility spending. However, the 400 MW project-level financing announced in April 2024 (claim 3237) had become advanced discussions on Ellendale HPC construction financing by August (claims 24912, 5867), with no close on record. The Highland Digital fund (claim 3169) is no longer discussed.

A careful reader should watch for four things. The first is whether Ellendale project financing actually closes. The second is the share count drawn under the SEPA and ATMs, along with Series F conversions against the 7.6 million convertible shares (claim 5834). The third is whether any CIM accordion draws trigger the contingent warrant tranche (claim 6031). The fourth is how the record continues past September 2024.

The record supplied ends mid-entry in September 2024 at claim #5742, and it contains no insider transactions or separately tagged capital events. This read therefore covers financings only, through that date. The two 2006 entries (claims #21058 and #21056, referencing SOCRATES and UNICORN technology) do not fit the company's later profile and require explanation as a possible entity-matching issue. Several dollar fields appear to be in mixed units, such as the $8 letter of credit in claim #13552 and the $36 equity raise in claim #3120. Figures should be read against their parallel claims.

The trajectory runs from modest, mostly non-dilutive patching toward heavier, structurally dilutive capital. The IPO was framed at up to $75 million (claim #14041) but priced at $40 million (claim #14007), netting $36 million (claim #24799). It sat alongside a one-for-six reverse split (claim #26010). Through 2022 and early 2023, management described its funding as non-dilutive regional bank debt (claim #21492), and the record bears that out. It shows a $7.5 million Vantage term loan (claim #21459) and Starion loans of $15 million (claim #13804) and $20 million (claim #13310).

From mid-2023 the mix shifts. A $125 million ATM (claim #12397) produced $64.5 million in one period (claim #21542) and was described as complete by January 2024 (claim #21552). The 9% B. Riley loan (claim #12517) was terminated early (claim #6291). It was followed in quick succession by several instruments that reach common equity. These include an AI Bridge note partly prepaid in stock (claim #6137) and a $42.1 million Yorkville prepaid advance (claim #6078), whose $16 million monthly conversion cap (claim #6111) was later eliminated (claim #5837). Others are CIM debt carrying warrants (claim #5972), Series E and Series F preferred (claims #6119 and #5757), a $160 million PIPE (claim #5782) and a $250 million SEPA (claim #5784). The layering reads more as sequential gap-filling than as a single designed capital structure, though each instrument is individually disclosed.

On use of proceeds, the Starion loan was tied to Ellendale expansion (claim #13326). The broader Ellendale HPC project-level financing, however, remained in advanced discussions as of August 2024 (claim #24912), with no closing on record before the data ends.

A careful reader should watch several items: whether Ellendale construction financing actually closes, the pace of Yorkville conversions after the cap removal, and draws on the CIM accordion that trigger further warrants (claim #6031). Two more are conversion of the Series F into roughly 7.6 million shares (claim #5834) and utilization of the SEPA.

The record supplied ends mid-entry in October 2024 (claim 5618 is truncated), so this read covers about three years and cannot address anything later. Two 2006 entries, the UNICORN and SOCRATES® items (claims 21056 and 21058), do not fit the company's later history and may belong to a predecessor or another entity, so they are set aside. No insider transactions appear in the data.

The cadence moves from sparse to dense. Through 2022 the financing is modest and conventional: Series C and D convertible preferred (claims 14500, 14501), a $40,000,000 IPO at $5.00 per share (claims 14007, 14000), and a $15,000,000 Starion term loan (claim 13804). In 2023 the instruments multiply, with a second $20,000,000 Starion loan (claim 13326), a $50,000,000 B. Riley facility (claim 12565), and a $125,000,000 ATM (claim 12411). By mid-2024 raising becomes nearly continuous. Within roughly four months come the $42,131,579 Yorkville prepaid advance (claim 6078), the CIM note with warrant tranches (claims 5974, 6010), a $50,000,000 Series F convertible preferred (claim 5781), a $160,000,000 PIPE (claim 5782), and a $250,000,000 SEPA (claim 5780).

The mix reads more as layering than as a single plan. Several obligations were revised soon after they were put in place. The B. Riley loan was terminated early (claim 6291), and the AI Bridge note was prepaid in common stock (claim 6137). The Yorkville advance carried a prepayment trigger tied to registration timing (claim 6036), and its conversion limit was later eliminated (claim 5837). Each change moves repayment from cash toward equity, so dilution compounds beyond what the headline offering sizes suggest.

Use of proceeds is thinly documented. The Starion loan carries a stated use (claim 13807), and project-level financing for a 400 MW hyperscaler lease is referenced (claim 3237). The HPC Ellendale financing, however, appears three times only as discussions (claims 5867, 5892, 5907), with no closed facility in this record. How the 2024 capital was deployed relative to that project requires explanation.

A careful reader should watch for draws under the SEPA and ATM, conversions of the Series F and Yorkville notes, and exercises of the CIM warrants. The key open question is whether the Ellendale discussions become a disclosed, closed financing. Several figures are recorded in ambiguous units (claims 3240, 5783, 6026) and should be read with caution.

Applied Digital's record shows dilution that rose steadily and then sped up sharply. Before 2023 it was modest: small convertible preferred rounds (the $660,000 Series C and $1.3 million Series D, per claims #14500 and #14501) and an IPO that was revised downward from a $75 million framing (claim #14370) to $40 million at $5.00 per share (claim #14007). Two 2006 technology-financing entries (claims #21056 and #21058) appear unrelated to the current business and require explanation. Through 2023 the company leaned on debt, including Starion term loans of $15 million and $20 million (claims #13804 and #13310) and a $50 million B. Riley facility (claim #12565), which was terminated early (claim #6291). The $125 million ATM (claim #12398) began a turn toward equity, and by mid-2024 instruments were arriving within weeks of one another.

The 2024 mix looks more like opportunistic patching than a single plan. The AI Bridge note (claim #6292) was prepaid in common stock (claim #6137). The Yorkville prepaid advance of $42.1 million (claim #6078) originally carried a $16 million monthly conversion cap (claim #6111), which was later eliminated (claim #5837). The CIM facility came with warrants for roughly 6.3 million shares (claim #6010), a contingent second tranche (claim #6031), and a waiver to permit added borrowing (claim #5760). Consent payments in shares to Yorkville (claim #5988) suggest each new layer had to be negotiated around the last. The $160 million PIPE (claim #5782) and the expanded $200 million CIM facility (claim #5660) look more deliberate, but they sit alongside a $250 million SEPA (claim #5780) and Series E, E-1 and F preferred with 8 to 9 percent dividends (claims #6136 and #5835).

The record says little about how proceeds were actually deployed. Ellendale is the stated destination for the Starion loan (claim #13326), and HPC Ellendale financing is described as in discussion several times (claims #5867, #5892 and #5907). However, the record shows no project-level close, including for the 400 MW hyperscaler lease (claim #3237). The $25 million Garden City escrow release recurs across filings and is marked delayed (claims #3242 and #5625).

No insider transactions appear in the record provided, and the record ends in October 2024. A careful reader should watch for three things: whether project-level Ellendale financing actually closes, how much of the SEPA is drawn and on what terms, and how the preferred dividends and warrant overhang add to dilution beyond the common share count.

Two caveats frame this read. The record shown here breaks off in early October 2024, so nothing after that point can be assessed. The two 2006 entries, the SOCRATES backlog (claim #21058) and UNICORN financing (claim #21056), appear to belong to an unrelated predecessor entity, and their inclusion requires explanation rather than interpretation. No insider transactions appear in the record, so that dimension cannot be evaluated.

Within the 2021–2024 window, the cadence moves from episodic to near-continuous. The IPO figures are internally inconsistent. Claims #14370 and #14041 describe up to $75M, while claims #14007 and #14000 record a $40M close. Small Series D discrepancies, $1.3M per claim #14501 against $1.38M per claim #14356, suggest the record was revised. After a 2022–2023 phase anchored in bank debt, the Starion term loans of $15M (claim #13804) and $20M (claim #13326), the mix shifts in 2024 toward layered, structure-heavy instruments. These include the $42.1M Yorkville prepaid advance (claim #6106), CIM borrowings with attached warrants (claims #6026, #6010), Series E and F preferreds, the $160M PIPE (claim #5782), and the $250M SEPA (claim #5780). These sit alongside the $125M ATM (claim #12397).

The sequence looks more like successive refinancing than a single plan. The $50M B. Riley loan (claim #12565) was terminated early (claim #6291). Its replacement, the $20M AI Bridge loan (claim #6292), was itself retired partly through an $8.4M stock issuance (claim #6137). The Yorkville $16M monthly conversion cap (claim #6111) was later eliminated (claim #5837), with early redemption following (claim #5838). Each step converts nearer-term debt into equity or equity-linked claims. The $50M Series E offering (claim #6119) shows only $6.9M of completed sales (claim #5657) before the Series E-1 was layered on at $62.5M (claim #5656).

On use of proceeds, the Ellendale expansion is explicitly tied to the $20M Starion loan (claim #13326). However, the HPC Ellendale project-level financing is described only as in discussion across three claims (#5907, #5867, #5892) and is not shown as closed within the record. The CIM facility is cited at $125M (claim #5753) and later at $200M (claim #5660), a revision that is not explained.

A careful reader should watch four things. The first is whether the HPC Ellendale financing actually closes. The second is the pace of SEPA and ATM draws relative to the share count. The third is uptake on the E-1 preferred given the thin Series E sales. The fourth is whether further CIM tranches trigger the contingent warrant issuance (claim #6031).

The two 2006 claims (SOCRATES and UNICORN, #21058 and #21056) predate the current business and appear to belong to a predecessor or unrelated entity. They require explanation before being read as part of this history. The record also ends partway through October 2024, so what follows covers the IPO through that point. The cadence is one of accelerating layering rather than episodic raises. The IPO was marketed at up to $75M (#14041) and closed at $40M on 8,000,000 shares at $5.00 (#14007, #14000). The company then relied on modest bank debt, the $15M and $20M Starion loans (#13804, #13310), before the 2023 pivot to a $175M shelf (#12544), a $125M ATM (#12397) and the 9% B. Riley facility (#12517). By mid-2024 the mix had widened to an unsecured AI Bridge loan (#6292), later prepaid in stock (#6137), and the Yorkville prepaid advance of $42.1M (#6078), whose $16M monthly conversion cap (#6111) was subsequently eliminated (#5837). It also added CIM senior debt with warrants on 6,300,449 shares (#6010), the $50M Series F convertible into 7,598,714 shares (#5834), the $160M PIPE (#5782) and a $250M SEPA (#5780), all within roughly four months. Several of these instruments convert to equity or carry warrants, so the effective dilution path is steeper than the headline equity raises suggest. The sequencing reads more as successive patching than a single designed capital stack. The B. Riley loan was terminated early (#6291), and Yorkville terms were revised within months. The Series E program, sized at $50M with $46M in stated net proceeds (#6119, #6120), had completed only $6.9M of sales by September (#5657) before a $62.5M Series E-1 offering appeared (#5656). The CIM facility also grew from $15M plus $110M plus a $75M accordion (#5974, #5975) to a $200M tranche structure (#5660), with a waiver enabling additional borrowing (#5760). On deployment, the project-level financing for the 400 MW hyperscaler lease (#3237) was stated in April 2024. By August, the Ellendale HPC financing was still described as in discussion (#5867, #5907), and the record shows no close. A sophisticated reader should watch four things: whether that project-level financing closes and on what terms, actual draws under the SEPA and ATM, the conversion pace of Series F and the remaining Yorkville balances, and whether Series E-1 replaces or supplements the undersold Series E. The data also contains unit inconsistencies, such as $75 versus $75,000,000, which require normalization before quantitative comparison.

The record is truncated after the October 2024 entries and contains no insider transactions, so this read covers financing only and stops at that point. The two 2006 entries, claims #21058 and #21056, describe SOCRATES and UNICORN technology programs that do not match the later data-center business; they appear to belong to a predecessor or a different entity and require explanation before being counted. Several amounts are recorded in inconsistent units (for example, $75 in claim #14041 against $75,000,000 in claim #14370), so figures here are cited only where the scale is clear.

Dilution and leverage were modest through 2023 and accelerated sharply in 2024. The early structure was conventional: the $40M IPO closing per claim #14000, then Starion bank term loans of $15M and $20M per claims #13804 and #13310. Mid-2023 marked a step up, with a $175M shelf per claim #12544, the $50M B. Riley facility per claim #12514, and a $125M ATM per claim #12397.

The 2024 mix reads less as a single plan than as sequential patching. The AI Bridge note per claim #6292 was partly retired with stock, $8.4M per claim #6137. The B. Riley loan was terminated early per claim #6291. The Yorkville prepaid advance of $42.1M per claim #6106 carried a conditional prepayment tied to registration timing per claim #6036, and its conversion limit was later eliminated per claim #5837. The CIM facility also came with warrants on 6.3M shares per claim #6010. September 2024 brought a more coherent package: the $160M Goldman-backed PIPE per claim #5782, the $50M Series F per claim #5781, a $200M tranched CIM facility per claim #5660, and a $250M SEPA per claim #5780. The Series E dividend rate was revised from 8% per claim #13312 to 9% per claim #6136. Completed Series E sales of $6.9M per claim #5657 lag the $50M offering per claim #6119.

On use of proceeds, the record names Ellendale expansion per claim #13326 and project-level financing for a 400 MW hyperscaler lease per claim #3237. However, HPC Ellendale financing appears three times as discussions only, per claims #5867, #5907 and #5892, with no closing recorded. A careful reader should watch for four things: whether that project financing closes, the pace of SEPA and ATM draws, the CIM tranche draws and the warrants they trigger, and whether the Series E-1 program per claim #5656 sells better than Series E did.

The record spans two distinct eras. The 2006 entries (claims #21058 and #21056, referencing SOCRATES and UNICORN technology) appear to belong to an unrelated predecessor that used the Applied Digital name, so they are set aside here as requiring explanation rather than treated as part of the current pattern. The substantive history begins with small preferred issuances in 2021, including the $660,000 Series C and $1.3M Series D (claims #14500 and #14501). The IPO was then revised down from up to $75M (claim #14370) to $40M priced at $5.00 per share (claims #14007 and #14000).

From there the cadence accelerates. Bank debt came first and was modest: the $15M Starion term loan (claim #13804) and the $20M Starion loan for Ellendale (claim #13310). By mid-2023 the company had added a $50M B. Riley facility at 9% (claims #12565 and #12517), a $175M shelf (claim #12544) and a $125M ATM (claim #12398).

In 2024 the company moved from conventional instruments toward structured, dilution-linked capital. That year brought the $20M AI Bridge note, later prepaid in stock (claims #6292 and #6137), and the $42.1M Yorkville prepaid advance (claim #6078). It also brought warrant-bearing CIM debt (claims #6010 and #5974), Series E, E-1 and F preferreds (claims #6119, #5656 and #5781), a $250M SEPA (claim #5780) and the $160M PIPE (claim #5782).

The mix reads less as a single plan than as successive layers, each added as the prior source was drawn down or constrained. Two items suggest terms were renegotiated mid-course: the early termination of the B. Riley loan (claim #6291) and the removal of the prepaid-advance conversion limit (claim #5837).

The stated use of proceeds consistently points to the Ellendale and HPC buildout (claims #13326 and #5907). The record shows repeated financing discussions for the HPC facility (claims #5867 and #5892) rather than closed project-level financing. The 400 MW hyperscaler project financing (claim #3237) is stated without a figure. Actual deployment cannot be verified from these claims.

The data has limits. The list is truncated in October 2024, no insider transactions appear, and many amounts are recorded in inconsistent units ($75 versus $75,000,000), so the figures should not be summed. A careful reader should watch for three things: whether project-level debt for the HPC campus closes, how much of the SEPA and ATM capacity is actually drawn, and how the conversion and redemption terms on the Series F and Yorkville instruments play out, since several are marked delayed.

The record runs from an April 2022 IPO through early October 2024 and appears to end mid-entry, so this read covers only that window. Two caveats come first. The 2006 entries (claims #21058 and #21056) describe technology programs that do not obviously belong to the current data-center operator and are set aside here. Many figures are also recorded in inconsistent units, for example the $75 offering in claim #14041 against $75,000,000 in claim #14370, which reflects extraction rather than disclosure.

The trajectory is one of accelerating layering rather than steady dilution. The first year relied on a $40,000,000 IPO (claim #14007) and modest bank debt from Starion, including $15,000,000 (claim #13804) and $20,000,000 (claim #13310). From mid-2023 the mix shifted toward instruments that settle in equity: a $125,000,000 ATM (claim #12397), the Yorkville prepaid advance of $42,131,579 (claim #6078), and Series E preferred. By September 2024, a single month carried a $160,000,000 PIPE (claim #5782), a $50,000,000 Series F convertible preferred (claim #5781), a $250,000,000 SEPA (claim #5780), and expanded CIM borrowing with attached warrants (claims #6010 and #5661).

The mix reads less as a planned capital structure than as sequential bridging. A $50,000,000 B. Riley facility (claim #12565) was terminated early (claim #6291). An AI Bridge loan was prepaid in stock, $8,421,146 per claim #6137. Yorkville consent was obtained through a share issuance (claim #5988), and a prepayment obligation was tied to an S-1 effectiveness deadline (claim #6036). Each step appears to resolve a near-term constraint while adding a new conversion or warrant overhang.

Use of proceeds is difficult to test from this record. Ellendale expansion is cited for the Starion loan (claim #13326), and project-level financing for a 400 MW hyperscaler lease is referenced (claim #3237). However, the HPC Ellendale financing appears across several claims as being in discussion (claims #5867, #5907, #5892) without a closing on record. That gap requires explanation.

Going forward, a reader should watch whether project-level debt closes and replaces corporate-level equity-linked funding. Other points to track are the draw pace on the SEPA and ATM, the conversion and redemption of Series F (claim #5839), and whether the many items marked delayed are completed, revised, or no longer discussed.

The record as provided ends in early October 2024, and it contains no insider transactions, so this read covers only the company-side financing pattern through that date. Two 2006 entries, the UNICORN and SOCRATES items per claims #21056 and #21058, appear to belong to a predecessor or unrelated entity and require explanation before they are read as part of Applied Digital's history.

The cadence moves from modest to compressed. The IPO was described as up to $75M (claims #14370 and #14041) and priced and closed at $40M (claims #14007 and #14000), a revised outcome that set a pattern of capital arriving in smaller pieces than first framed. Through 2022 and early 2023 the mix stayed bank-centric, with the $15M and $20M Starion term loans (claims #13804 and #13310). From mid-2023 the instruments multiply. A $125M ATM (claim #12397) and a $50M B. Riley facility (claim #12565) were followed within a year by an unsecured bridge loan (claim #6292), a $42.1M prepaid advance with monthly conversion mechanics (claims #6106 and #6111), warrant-bearing CIM debt (claims #6010 and #5753), Series E, E-1 and F preferred (claims #6119, #5656 and #5781), a $250M SEPA (claim #5780) and a $160M PIPE (claim #5782). The bridge note was prepaid in stock (claim #6137) and the B. Riley loan was terminated early (claim #6291). The Series E-1 and F terms are not fully disclosed in this record, but taken together the instruments shift the dilution trajectory from discrete issuances toward continuous issuance. The September 2024 cluster reads as a deliberate consolidation, but the twelve months before it look more like sequential patching, with consents bought in shares (claim #6032) and conversion limits later eliminated (claim #5837).

Stated use of proceeds is thin. The second Starion loan was tied to Ellendale expansion (claim #13326), and the 2024 materials repeatedly reference financing discussions for HPC Ellendale (claims #5867 and #5907) and project-level financing for a 400 MW hyperscaler lease (claim #3237). The record does not show deployment, so whether proceeds tracked these purposes cannot be confirmed here. Many claims carry a delayed status, including the Series F proceeds (claim #5616) and the CIM borrowing (claim #5973). Several figures also appear in inconsistent units.

A careful reader should watch for four things next: how much of the SEPA and ATM capacity is actually drawn, the conversion behavior of the prepaid advance and Series F, whether the Ellendale project-level financing closes or is no longer discussed, and whether later filings reconcile the delayed items.

The record divides into phases with very different capital signatures. The 2006 entries (claims #21056, #21058) reference UNICORN and SOCRATES technology programs and appear to relate to a predecessor or different registrant. They are not continuous with the later data-center business and are set aside here. From 2021 through early 2023 the financing is modest and conventional: small Series C and D convertible preferred issues (claims #14500, #14501), an IPO announced at $75M (claim #14370) but priced and closed at $40M (claims #14007, #14000), and Starion Bank term loans of $15M and $20M (claims #13804, #13326). The IPO downsizing is the first instance of a stated figure being revised downward.

From mid-2023 the cadence accelerates sharply. A $175M shelf (claim #12544) and a $125M ATM (claim #12397) arrive alongside a $50M B. Riley term loan (claim #12565), which was terminated early less than a year later (claim #6291). The first half of 2024 then layers several instruments in quick succession. These include an unsecured bridge loan (claim #6292) prepaid in common stock (claim #6137) and a Yorkville prepaid advance of roughly $42M with a $16M monthly conversion limit (claims #6106, #6111), a limit that was later eliminated (claim #5837). They also include retail Series E preferred and a CIM facility carrying warrants and an accordion (claims #6010, #5975). Each instrument appears to address the gap left by the previous one rather than forming part of a pre-planned structure. The recurring consent share issuances and lender waivers (claims #5988, #5760) are consistent with that reading.

September 2024 marks a change in counterparty profile. A $160M PIPE described as Goldman-backed (claim #5782), $50M of Series F convertible preferred (claim #5781), and a $250M SEPA (claim #5780) were announced together. The combined dilution capacity from the SEPA, ATM, Series F conversion and outstanding warrants is large relative to earlier years, though the record does not provide shares outstanding with which to size it.

On use of proceeds, the Ellendale facility and a 400 MW hyperscaler lease are cited repeatedly (claims #13326, #3237). HPC Ellendale financing is described as under discussion across three claims (#5867, #5892, #5907), but no closed project-level facility appears in the record, and that gap requires explanation in later disclosures. No insider transactions appear in the data, and the record is truncated in early October 2024. A careful reader should watch for closure of project financing, the draw rate on the SEPA and ATM, and Series F conversion activity.

The record supplied here ends abruptly in early October 2024, so this read covers roughly three years of Applied Digital's capital history, not its full disclosed history. Two 2006 entries, the UNICORN and SOCRATES items (claims #21056 and #21058), describe technology contract activity that does not fit the company's later profile. They likely belong to a predecessor or a different registrant and require explanation before being read as part of this pattern. Several dollar figures also appear to be rates or values stated in millions, such as the 8.0% Series E dividend in claim #13312, so only figures with clear scale are relied on here.

The trajectory runs from modest bank debt to layered, structurally complex financing. After the IPO of 8,000,000 shares at $5.00 for $40,000,000 (claim #14007), the company leaned first on Starion term loans of $15,000,000 and $20,000,000 (claims #13804 and #13326). It then added a $50,000,000 B. Riley facility (claim #12565) and a $125,000,000 ATM (claim #12397). From early 2024 the pace accelerates sharply. Within about five months the company used an AI Bridge loan later prepaid in stock (claim #6137) and a Yorkville prepaid advance of $42,131,579 (claim #6078). It also drew on the CIM secured facility with attached warrants for 6,300,449 shares (claim #6010), Series E and F preferred, and a $250,000,000 SEPA (claim #5780). In the same window it closed a $160,000,000 PIPE (claim #5782). The B. Riley loan was terminated early (claim #6291), and the Yorkville conversion limit was later eliminated (claim #5837). Taken together, this looks less like a planned capital stack than successive instruments arranged to bridge toward larger project-level funding. Each new layer added potential dilution through conversion features, warrants, or equity lines.

On deployment, the record states a use of proceeds for the Starion loan (claim #13807) but does not show its content. The project-level financing for a 400 MW hyperscaler lease (claim #3237) and the Ellendale HPC financing discussions (claims #5867, #5892 and #5907) appear repeatedly as discussions, without a closing in the data shown. The $25,000,000 Garden City escrow release is flagged delayed in claim #3242 and is stated as released in claim #5779. The supplied data contains no insider transactions, so nothing can be said about insider selling.

Going forward, a reader should watch for three things. The first is whether the Ellendale project financing actually closes or is no longer discussed. The second is how much of the SEPA and ATM capacity is drawn, since that sets realised dilution. The third is whether the CIM accordion of $75,000,000 (claim #5975) is used, because each draw carries contingent warrants (claim #6031).

Applied Digital's record describes a company that moved from an equity-funded start-up to a layered, increasingly structured capital stack within about thirty months. After the 2022 IPO, which priced at $40M against an earlier 'up to $75M' framing (claims 14007, 14000, 14041), funding shifted first to bank debt, with the $15M and $20M Starion term loans (claims 13804, 13310). From mid-2023 it widened to the $50M B. Riley facility (claim 12514), a $125M ATM (claim 12397) and a $175M shelf (claim 12544). The cadence accelerated sharply in 2024. Between May and October the record shows the $42.1M Yorkville prepaid advance (claim 6078), the CIM secured note with warrants and a $75M accordion (claims 5975, 6010), Series E and E-1 preferred (claims 6119, 5656), the $50M Series F convertible preferred (claim 5781), the $160M PIPE (claim 5782) and the $250M SEPA (claim 5780). Dilution in this period comes less from any single raise than from stacked convertible, warrant-bearing and standby channels that settle into common stock over time. The AI Bridge note prepaid in shares (claim 6137) is one instance of debt resolving into equity. The mix reads as a strategy that evolved under pressure rather than one set in advance. Bridge debt was taken and retired within months (claims 6292, 6291, 6050), consent shares went to Yorkville (claim 6032), and conversion limits were later eliminated (claim 5837), which suggests terms were renegotiated as needs shifted. Several items carry delayed status, including the Yorkville registration deadline (claims 6108, 6036) and the Series F registration rights (claim 5836). On use of proceeds, the Starion borrowing is tied to the Ellendale expansion (claim 13326), and project-level financing is referenced for a 400 MW hyperscaler lease (claim 3237). HPC Ellendale financing appears only as discussions (claims 5867, 5907), with no recorded close, so deployment against stated aims cannot be confirmed from this data. A careful reader should watch whether the HPC Ellendale financing closes, how much SEPA and ATM capacity is drawn, and how much conversion occurs on the Series F and CIM warrants. Parts of the dataset also require explanation. The 2006 technology-contract entries (claims 21056, 21058) and some IPO figures (claims 14065, 14079) do not clearly fit this issuer. Several amounts appear to be recorded in millions without units. The list is truncated in October 2024, and no insider transactions are shown.

The record shows a company that moved from small convertible preferred rounds into a layered, high-frequency capital stack within roughly three years. The two 2006 entries (claims #21058 and #21056) describe unrelated technology programs and appear to belong to a predecessor or similarly named entity, so this read starts in 2021. The IPO was first described as up to $75M (claim #14041) and closed as 8,000,000 shares at $5.00 for $40M (claims #14007, #14000). That revision set the tone: early capital came in below the initial target, and debt filled the gap quickly. Starion term loans of $15M and $20M (claims #13804, #13326) were followed by a $50M B. Riley facility at 9% (claims #12565, #12517), which was terminated early less than a year later (claim #6291) and partly replaced by a $20M unsecured bridge (claim #6292). That bridge was later settled in common stock (claim #6137). This is the first clear case of debt converting into dilution. By mid-2024 the mix had broadened to include the Yorkville prepaid advance of $42.1M with a $16M monthly conversion cap (claims #6106, #6111), the CIM facility with warrants on 6,300,449 shares (claims #6026, #6029), Series E and F preferred, a $160M PIPE (claim #5782), a $250M SEPA (claim #5780), and the standing $125M ATM (claim #5778). Each instrument is individually explicable for a capital-intensive buildout. Their overlap within weeks, including a CIM waiver to permit additional borrowing (claim #5760), looks more like sequential patching than a pre-planned structure. Use of proceeds is only partly traceable. The Starion loan is tied to Ellendale expansion (claim #13326), yet HPC Ellendale financing appears three times as being in discussion rather than closed (claims #5892, #5867, #5907). The project-level financing for the 400 MW hyperscaler lease (claim #3237) is announced without a recorded close. The Series E offering targeted $50M (claim #6119), but completed sales reached $6.9M (claim #6657 is not on record; per claim #5657), after which a $62.5M Series E-1 was introduced (claim #5656). That gap requires explanation. No insider transactions appear in the record, and the data ends in early October 2024, so later developments are not assessed. A careful reader should watch whether the Ellendale and hyperscaler project financings close, how much of the SEPA and ATM capacity is drawn, the pace of Yorkville conversions, and whether the CIM accordion and contingent warrants (claim #6031) are triggered.

Applied Digital's financing record shows a company that moved from modest, bank-led funding into a dense, multi-instrument capital stack within roughly eighteen months. The 2022 IPO, priced at $5.00 for 8,000,000 shares and $40,000,000 gross (claims #14007 and #14000), was well below the $75,000,000 figure first registered (claim #14370). Through 2022 and early 2023, financing came mainly from Starion Bank term loans of $15,000,000 and $20,000,000 (claims #13804 and #13326), the second tied to the Ellendale expansion. The B. Riley $50,000,000 facility (claim #12514) and the $125,000,000 ATM (claim #12397) in mid-2023 widened the toolkit. The real inflection is 2024. That year the record adds an unsecured AI Bridge loan (claim #6292), a Yorkville prepaid advance of $42,131,579 convertible into common stock (claim #6106), a CIM secured facility with tranche warrants (claims #5974 and #6010), Series E, E-1 and F preferred programs (claims #6119, #5656 and #5781), a $160,000,000 PIPE (claim #5782), and a $250,000,000 SEPA (claim #5780). Dilution sources multiplied over that period, through conversions, warrants, preferred stock, standby equity and the ATM. The direction of dilution is therefore clearly upward, though the record does not quantify realized share issuance.

The mix reads more as sequential patching than a planned capital structure. The B. Riley loan was terminated early (claim #6291). The AI Bridge note was retired with stock (claim #6137). The Yorkville conversion limit, originally $16,000,000 monthly (claim #6111), was later eliminated (claim #5837). CIM required a waiver to permit added borrowing (claim #5760). Each of these is an amendment to terms set only months earlier. The Series E program was sized at $50,000,000 (claim #6119), yet completed sales are recorded at $6,900,000 (claim #5657), which suggests limited retail uptake.

The use-of-proceeds trail is thin. Ellendale is the only consistently named destination. The HPC Ellendale financing appears three times as discussions (claims #5867, #5892 and #5907), and project-level financing for the 400 MW hyperscaler lease is stated but not shown as closed (claim #3237). The record contains no insider transactions.

Several items in the record require explanation. The 2006 UNICORN and SOCRATES entries (claims #21056 and #21058) appear unrelated to the current business. Many figures carry inconsistent units, such as "$75" against "$75,000,000". The supplied record also ends abruptly in early October 2024. Readers should watch whether the hyperscaler and HPC project financing actually closes, how much of the SEPA and ATM capacity is drawn, and whether the preferred series and CIM warrants convert into material share count.

The record as provided ends abruptly in October 2024 and contains no insider transactions, so this read covers financing behavior only and cannot speak to insider selling. Two 2006 entries describing UNICORN and SOCRATES technology (claims #21056 and #21058) do not match this issuer's later business and require explanation before being treated as part of its history. Several amounts also appear to be share counts or rates logged as dollars, such as the 3,236,245-share IPO entry per claim #14079, so they should be read with that caveat.

The trajectory is one of accelerating cadence. After small convertible preferred issues and a $40M IPO (claim #14007), the company relied on modest bank debt: the $15M Starion term loan (claim #13804) and a further $20M (claim #13326). In 2023 the instruments widened to a $50M B. Riley term loan (claim #12565) and a $125M ATM (claim #12397). By mid-2024 the pace became nearly continuous. Within roughly four months the record shows a $20M unsecured bridge (claim #6292), a $42.1M Yorkville prepaid advance (claim #6078), $50M of Series E preferred (claim #6119), a CIM facility with warrants attached (claims #6010 and #5974), $50M of Series F convertible preferred (claim #5781), a $160M PIPE (claim #5782), a $250M SEPA (claim #5780), and $62.5M of Series E-1 (claim #5656).

The mix reads less like a planned capital stack than sequential bridging toward project-level financing. The bridge note was prepaid partly in stock (claim #6137). Yorkville received consent shares (claim #6032), and its conversion limits were later removed (claim #5837). A prepayment obligation was tied to a registration deadline (claim #6036). The B. Riley loan was terminated early (claim #6291). The Goldman-backed PIPE stands out as the first large conventional equity raise in the record.

On deployment, the stated purposes center on the Ellendale build-out. Financing for the HPC Ellendale facility appears in repeated statements of advanced discussions (claims #5867, #5892 and #5907) without a recorded close. The project-level financing for a 400 MW hyperscaler lease (claim #3237) likewise has no recorded resolution. The $25M Garden City escrow release moved from delayed (claim #3242) to stated (claim #5779).

A careful reader should track several items: whether Ellendale project debt actually closes, how much of the SEPA and ATM capacity is drawn, contingent warrants tied to CIM accordion draws (claim #6031), the 7.6M shares underlying Series F (claim #5834), and the cumulative preferred dividend load across the E, E-1 and F series.

The record is dominated by a compressed burst of activity between May and October 2024, and the list provided ends mid-entry in October 2024, so this read covers only that span. The two 2006 entries (claims #21056 and #21058) predate the current business and are not referenced again. Early financing was modest and conventional: small convertible preferred issuances in 2021 (claims #14500, #14501), an IPO initially described at $75M (claim #14370) that priced and closed at $40M (claims #14007, #14000), and asset-backed Starion term loans tied to Ellendale (claims #13804, #13326). The IPO sizing was revised downward, which is the first instance of stated capacity exceeding realized proceeds.

From mid-2023 the mix widens rapidly. A $50M B. Riley term loan at 9% (claims #12565, #12517) was terminated early within a year (claim #6291). A $20M AI Bridge loan (claim #6292) was prepaid in common stock (claim #6137). The Yorkville prepaid advance of roughly $42.1M (claim #6078) carried a monthly conversion limit (claim #6111) and a registration deadline tied to a prepayment obligation (claim #6036), and the conversion limitation was later eliminated (claim #5837). In practice, a sequence of debt instruments was retired through equity. That reads more as successive refinancing than a pre-set capital plan.

September 2024 layers a $160M PIPE (claim #5782), $50M Series F convertible preferred (claim #5781), a $250M SEPA (claim #5780), a $125M ATM (claim #5778), and a CIM facility described at $200M (claim #5660) with warrants attached (claim #6010). Stated capacity here substantially exceeds what the record shows as drawn or sold. Series E, for example, was registered at $50M (claim #6119), but completed sales total $6.9M (claim #5657).

Use of proceeds is specific only for the Starion Ellendale loans. The HPC Ellendale project financing appears in repeated "discussions" entries (claims #5867, #5892, #5907) without a closed facility on record. Actual deployment cannot be verified from this data. No insider transactions are included.

A careful reader should watch for four things. First, how much of the SEPA and ATM capacity is actually drawn, and at what prices. Second, whether the CIM warrant tranches are triggered through accordion draws (claim #6031). Third, the conversion activity on the Series F preferred (claim #5834). Fourth, whether the Ellendale project financing moves from discussion to a closed agreement.

The record needs two caveats before any reading. The 2006 entries covering UNICORN and SOCRATES technology financing (claims #21056 and #21058) do not fit Applied Digital's disclosed history and require explanation, possibly as predecessor or misattributed filings. Several figures also appear to be unit-truncated or share counts recorded as dollars, and the IPO is described inconsistently: $75M (claim #14370), 8,000,000 shares at $5.00 for $40M (claim #14007), and $8,571,428 (claim #14065). The record ends in October 2024, and no insider transactions are on file.

Within those limits, the cadence is clear and accelerating. The 2022 IPO and a $15M Starion term loan (claim #13804) were followed in 2023 by a $20M Starion loan (claim #13310), a $175M shelf (claim #12544), a $50M B. Riley facility (claim #12565) and a $125M ATM (claim #12411). In 2024 the pace compressed into months. It included Series E preferred, a Yorkville prepaid advance of $42.1M (claim #6078), CIM secured notes with warrants, Series F convertible preferred, a $250M SEPA (claim #5780), a $160M PIPE (claim #5782) and Series E-1 (claim #5656).

The mix reads as layered rather than sequenced. Senior secured credit, variable-price equity lines and convertible preferred were added in parallel. Several involved consent shares, waivers or modified conversion limits (claims #5988, #5760, #5837), which suggests terms were renegotiated to accommodate the next instrument. The B. Riley loan was terminated within a year (claim #6291). Demand for the retail preferred also appears thin: the Series E offering was sized at $50M (claim #6119), but completed sales total $6.9M (claim #5657).

Stated purposes centre on Ellendale expansion (claim #13326) and project-level financing for a 400 MW hyperscaler lease (claim #3237). The HPC Ellendale financing is described three times as being in discussion (claims #5867, #5892, #5907), but no closing is recorded. Deployment against that stated aim is therefore not yet evidenced in this data.

The items to watch are the structural dilution already in place. That includes the Series F conversion into roughly 7.6M shares (claim #5834), CIM warrants of 6.3M shares (claim #6010) plus a contingent 5.56M (claim #6031), and SEPA drawdowns. Two further questions are whether the CIM balance of $105M (claim #5621) moves into its accordion and whether Ellendale project financing is reported as closed or is no longer discussed.

The record opens with two 2006 items on SOCRATES and UNICORN technology (claims #21056, #21058) that do not connect to the later data-center business and require explanation before they are read as part of this company's history. The substantive record begins with small convertible preferred issuances in 2021, such as the $1.3M Series D (claim #14501). The IPO is recorded as a $75M target (claims #14370, #14041) that priced at $40M (claims #14007, #14000), so the company entered public markets with roughly half of its stated ambition.

Financing then shifts from bank debt to increasingly layered, equity-linked capital. The 2022–2023 phase relied on conventional secured loans from Starion, of $15M and $20M (claims #13804, #13310), then a $50M B. Riley facility (claim #12565), alongside a $125M ATM (claim #12411). By 2024 the mix becomes dense and overlapping. It includes a $20M unsecured AI Bridge loan (claim #6292) that was later retired through common stock issuance (claim #6137), a Yorkville prepaid advance of $42.1M (claim #6106), a $250M SEPA (claim #5780), $50M of Series F convertible preferred (claim #5781), a $160M PIPE (claim #5782), a $62.5M Series E-1 program (claim #5656), and a CIM facility of up to $200M carrying warrants for 6,300,449 shares (claims #5660, #6010). Several of these signings fall within weeks of each other. Debt settled in stock, a waived conversion limitation on the prepaid advance (claim #5837), and consent shares issued to Yorkville (claim #5988) read more as responsive patching than as a pre-planned capital structure, though the PIPE and CIM facility suggest institutional backing did arrive.

The record also shows execution gaps. The Series E offering was sized at $50M (claim #6119), yet completed sales total $6.9M (claim #5657) before it was re-launched as E-1. The stated use of loan proceeds (claim #13807) carries no figure, and the dataset contains no deployment reporting to test against it. Advanced financing discussions for HPC Ellendale recur across three claims (#5867, #5892, #5907) without a closed transaction on record. Project-level financing for the 400 MW hyperscaler lease (claim #3237) is likewise stated but not yet evidenced.

No insider transactions appear in the record, so that dimension cannot be assessed. A careful reader should watch the pace of SEPA draws and Yorkville conversions, further draws on the CIM accordion beyond the $105 balance noted (claim #5621, units unstated), and whether the Ellendale and hyperscaler project financings close or are no longer discussed.

The record covers late 2021 through early October 2024 and is cut off mid-list, so this read addresses that window only. The two 2006 entries for SOCRATES and UNICORN technology (claims #21058 and #21056) appear to belong to a different issuer and are set aside. No insider transactions are present, so that dimension cannot be assessed. Several amounts are stated in inconsistent units (for example '$75' in claim #14041 and '$8' in claim #13312), which limits precise comparison.

Dilution starts modestly and accelerates. The IPO was pitched at up to $75,000,000 (claim #14370) but priced at $40,000,000 for 8,000,000 shares at $5.00 (claim #14007). That is a revised outcome worth noting. Early growth was then financed mainly with secured debt: Starion term loans of $15,000,000 (claim #13804) and $20,000,000 (claim #13310), then a $50,000,000 B. Riley facility at 9% (claims #12565 and #12517). A $125,000,000 ATM (claim #12397) opened a standing equity channel in mid-2023.

The 2024 mix reads less as a planned capital stack and more as layered patching. The B. Riley loan was terminated early (claim #6291) and replaced by an unsecured $20,000,000 AI Bridge loan (claim #6292), which was later settled in stock (claim #6137). Yorkville's $42,131,579 prepaid advance (claim #6078) initially had a monthly conversion cap (claim #6111), but that limitation was later eliminated (claim #5837). The CIM facility came with warrants covering 6,300,449 shares (claim #6010) and a subsequent waiver for an extra $20,000,000 (claim #5760). The Series E preferred was sized at $50,000,000 (claim #6119), yet completed sales total only $6,900,000 (claim #5657), and a Series E-1 then followed (claim #5656). September 2024 brought a more institutional reset: the $160,000,000 Goldman-backed PIPE (claim #5782), the $50,000,000 Series F (claim #5781) and a $250,000,000 SEPA (claim #5780). These sat alongside the instruments above rather than replacing them.

Use of proceeds tracks early on. The Starion loan was tied to the Ellendale expansion (claim #13326). Later, project-level financing for a 400 MW hyperscaler lease (claim #3237) and 'advanced financing discussions' for HPC Ellendale (claims #5907, #5892 and #5867) were repeated without a disclosed close in this record, which requires explanation.

A careful reader should watch four things: whether the Ellendale project financing is announced or quietly no longer discussed, the pace of SEPA and ATM draws, conversion of the Series F into roughly 7,598,714 shares (claim #5834), and whether CIM accordion draws trigger the contingent warrants (claim #6031).

The record shows dilution that starts small and speeds up sharply. The pre-IPO period is modest, with Series C and D preferred of $660,000 and $1.3M (claims #14500, #14501). The IPO was described as a firm-commitment offering of up to $75M (claim #14041) but priced and closed at $40M, 8,000,000 shares at $5.00 (claims #14007, #14000). That is a revised outcome, and the record does not explain it. Through 2022 and 2023 the company relied on conventional secured bank debt, the Starion loans of $15M and $20M (claims #13804, #13310), backed by a $175M shelf and a $125M ATM (claims #12544, #12397). In 2024 the pace changes. Within about four months the company layered an AI Bridge unsecured loan of $20M (claim #6292), later prepaid in stock (claim #6137); a Yorkville prepaid advance of $42.1M (claim #6078); a CIM facility with warrants on 6,300,449 shares (claim #6010), grown to $200M (claim #5660); Series E, E-1 and F preferred; a $160M PIPE (claim #5782); and a $250M SEPA (claim #5780).

The mix from early 2024 onward looks more reactive than designed. The B. Riley $50M facility was terminated early (claims #12514, #6291). Consent shares were issued to Yorkville (claim #6032). A CIM waiver was needed to borrow more (claim #5760). The $16M monthly conversion cap on the Yorkville note (claim #6111) was later removed (claim #5837). Each instrument appears to have required concessions to earlier ones. The September 2024 PIPE and SEPA read as an attempt to consolidate this. The Series E offering also requires explanation: it was sized at $50M (claim #6119), but only $6.9M of completed sales is recorded (claim #5657).

Use of proceeds can only be partly checked. The Starion loans were tied to Ellendale expansion (claim #13326). Project-level financing for the 400 MW hyperscaler lease (claim #3237) and the HPC Ellendale facility appears repeatedly as discussion rather than closure (claims #5907, #5867, #5892). The record contains no deployment data to confirm spending against those stated purposes. The record also contains no insider transactions. The 2006 claims (#21056, #21058) sit outside this pattern and require explanation. The 2022 SparkPool forfeiture (claim #13738) is marked delayed.

The record ends in October 2024 and appears truncated. A careful reader should watch four things. First, conversion of the Series F into 7,598,714 shares (claim #5834). Second, the roughly $30M of potential warrant proceeds (claim #5725). Third, how heavily the SEPA is drawn. Fourth, whether HPC project financing moves from discussion to a closed transaction.

The record is dense from mid-2022 onward and effectively silent before it. The two 2006 entries (claims #21058 and #21056) reference SOCRATES and UNICORN technology programs that do not connect to the data-center business described elsewhere, and they require explanation before being read as part of this issuer's history. The listing itself is recorded inconsistently. A $75M offering size (claims #14370, #14041) sits against an 8,000,000-share, $40M pricing and closing (claims #14007, #14000), which reads as a downsized IPO rather than execution of the original plan.

The cadence of financing accelerates sharply. Through 2023 the mix is conventional for an infrastructure build: Starion term loans of $15M and $20M tied to Ellendale (claims #13804, #13326), a $50M B. Riley facility (claim #12565), a $175M shelf (claim #12544) and a $125M ATM (claim #12397). From February 2024 the pattern shifts toward shorter-dated, layered instruments. These include the $20M AI Bridge loan (claim #6292), later prepaid in stock (claim #6137), and Yorkville prepaid advance notes of roughly $42.1M with a $16M monthly conversion limit (claims #6106, #6111). They also include warrant-bearing CIM debt (claims #6010, #6029), Series E, E-1 and F preferred (claims #6119, #5656, #5781), a $250M SEPA (claim #5780) and a $160M PIPE (claim #5782). Many of these land within weeks of one another, and the Yorkville conversion limit was later eliminated (claim #5837). Taken together, this reads less as a sequenced capital plan than as successive bridging toward a larger project-level financing. That financing is repeatedly described as in discussion for HPC Ellendale (claims #5867, #5907) and the 400 MW hyperscaler lease (claim #3237), but it is not recorded as closed here.

Use of proceeds can only be partially checked. Early debt is tied explicitly to Ellendale (claim #13326), and the land purchase (claim #12622) is consistent with that purpose. The later convertible, SEPA and preferred instruments carry no stated deployment in this record. No insider transactions are included, so that dimension cannot be assessed.

A reader should watch whether the project-level financing closes and on what terms, and how much of the SEPA and ATM capacity is drawn. Other items to track are the pace of Yorkville conversions (claim #5624) and whether cumulative preferred dividends (claims #6136, #5835) begin to compete with operating needs. The record ends mid-entry in October 2024, so anything later falls outside what this read can support.

The supplied record ends mid-entry in early October 2024, so this read covers roughly thirty months of activity and nothing later. Two 2006 entries describing SOCRATES and UNICORN technology programs (claims #21058, #21056) do not connect to the later data-center business and require explanation, possibly as misattribution to another entity. Several other entries carry values without a clear unit (the "$75" offering in claim #14041, the "$160" in claim #3240), and this read relies on them only where another claim corroborates them. The IPO itself was revised downward. An offering of up to $75M (claims #14370, #14041) closed as 8,000,000 shares at $5.00 for $40M (claim #14007). Through 2022 and 2023 the company leaned on secured debt tied to the Ellendale site: the Starion loans of $15M and $20M (claims #13804, #13326) and the $50M B. Riley facility (claim #12565). The $175M shelf (claim #12544) and the $125M ATM (claim #12397) sat alongside that debt as equity capacity. In 2024 both the tempo and the complexity rose sharply. The record shows a $50M Series E preferred program (claim #6119) with only $6.9M of completed sales by September (claim #5657), succeeded by a $62.5M Series E-1 (claim #5656). It also shows a $42.1M Yorkville prepaid advance (claim #6078) whose monthly conversion limit was later eliminated (claim #5837), a CIM facility that grew to $200M with warrant tranches attached (claims #5660, #6010), a $160M PIPE (claim #5782), a $50M Series F (claim #5781), and a $250M SEPA (claim #5780). The 2022–2023 pattern reads as asset-matched borrowing. The 2024 pattern reads more like sequential bridging. A $20M unsecured AI Bridge loan (claim #6292) was partly settled in stock (claim #6137), the B. Riley loan was terminated early (claim #6291), consent shares went to Yorkville (claim #6032), and a prepayment obligation was triggered if a registration statement was not effective by July 8 (claim #6036). The Starion proceeds were tied explicitly to Ellendale (claims #13326, #13807). For most 2024 instruments the record does not state a use of proceeds, so deployment cannot be tested against intent. Project-level financing for the 400 MW hyperscaler lease (claim #3237) appears repeatedly as "in discussions" (claims #5892, #5907) without a recorded close. A careful reader should watch four things: whether that project financing closes and retires the short-dated instruments; the pace of draws across the ATM, the SEPA, and the Yorkville conversions, which run at the same time; the warrant issuance contingent on CIM accordion draws (claim #6031); and the treatment of the Series E and E-1 redeemable preferred. The supplied record contains no insider transactions.

Applied Digital's disclosed record shows a company whose financing cadence accelerated sharply as its ambitions scaled. Through 2022 and early 2023, capital came in modest, conventional increments: a $40M IPO (claim #14007), a $15M Starion term loan (claim #13804), and a $20M Starion facility earmarked for the Ellendale expansion (claim #13326). The 2023 additions of a $50M B. Riley term loan (claim #12565) and a $125M at-market agreement (claim #12397) marked a shift to larger, more flexible instruments. By mid-2024 the pace had changed character. Within roughly four months the company layered a $42.1M Yorkville prepaid advance (claim #6078), a CIM credit facility described at $200M in total (claim #5660), a $25M Roth sales agreement (claim #6081), $50M of Series F convertible preferred (claim #5781), a $160M PIPE (claim #5782), a $250M standby equity purchase agreement (claim #5780), and a $62.5M Series E-1 program (claim #5656). The dilution trajectory is therefore not linear but concentrated in the second half of 2024. The mix reads more as sequential patching than a single designed structure. Short-dated, conversion-heavy Yorkville paper sat alongside secured CIM debt that carried warrants, and the record shows the Yorkville monthly conversion cap being eliminated (claim #5837) and a waiver needed from CIM to borrow further (claim #5760). The early termination of the B. Riley loan (claim #6291) shows prior instruments being retired rather than extended. On use of proceeds, the Ellendale facility is the consistent stated destination, but the record repeatedly describes financing for the HPC Ellendale build as being in discussion (claims #5867, #5892, #5907) and the project-level financing for the 400 MW hyperscaler lease (claim #3237) is announced without a disclosed closing in this dataset, so deployment against stated purpose cannot be confirmed from what is here. No insider transactions appear in the record provided, and the claim list is truncated in October 2024, so later events are not assessed. A careful reader should watch whether the Ellendale project-level financing closes on disclosed terms, how much of the SEPA and ATM capacity is actually drawn, and whether the Series E, E-1 and F preferred layers are redeemed or converted, since that sequence will determine realized dilution.

The record shows two distinct phases. Through mid-2023, Applied Digital funded itself mostly with conventional instruments: an IPO that closed at $40M (claims #14007, #14000) against an earlier-stated $75M target (claims #14370, #14041), followed by bank term loans from Starion of $15M and $20M (claims #13804, #13310) and a $50M B. Riley facility (claim #12565), later terminated early (claim #6291). A $125M Craig-Hallum ATM (claim #12397) added a standing equity channel. The 2022 share issuances to GMR, SparkPool and Valuefinder, and the SparkPool forfeiture (claims #13751, #13752, #13753, #13738), are marked delayed and require explanation.

From early 2024 the cadence changes sharply. Within roughly eight months the company layered a $20M AI Bridge loan (claim #6292), later prepaid in common stock (claim #6137); a $42.1M Yorkville prepaid advance convertible at a discount-based price (claims #6106, #6112), whose $16M monthly conversion cap (claim #6111) was later eliminated (claim #5837); a CIM secured note with warrants for 6,300,449 shares (claim #6010); a $25M Roth ATM (claim #6081); and, in September, a $160M PIPE (claim #5782), $50M of Series F convertible preferred (claim #5781), a $250M SEPA (claim #5780) and $62.5M of Series E-1 preferred (claim #5656). This reads less as a sequenced capital plan than as parallel channels opened to bridge a build-out, with several instruments carrying conversion or warrant features that transfer dilution to later periods.

Use of proceeds is only partly traceable. The $20M Starion loan was tied to the Ellendale expansion (claim #13326), but later disclosures describe Ellendale HPC financing as still in discussion (claims #5867, #5907), and the project-level financing for the 400 MW lease (claim #3237) has no closing on record. The $50M Series E program (claim #6119) had completed only $6.9M of sales by September 2024 (claim #5657), shortly before the E-1 offering appeared. The $25M Garden City escrow release is marked delayed (claim #5625).

The record contains no insider transactions, so insider alignment cannot be assessed, and it is truncated in October 2024. A reader should watch the pace of SEPA draws and Yorkville conversions now that the monthly cap is gone, whether the E-1 preferred sells better than Series E did, and whether Ellendale project financing actually closes or is no longer discussed.

Applied Digital's financing record moves from small preferred issuances before listing (the $660,000 Series C and $1.3M Series D, per claims #14500 and #14501) to a $40M IPO at $5.00 per share (claim #14007), and then into a steadily widening set of instruments. Through 2022 and early 2023 the mix leaned on site-specific bank debt, notably the $15M Starion term loan (claim #13804) and a $20M Starion loan for the Ellendale expansion (claim #13326). A $175M shelf (claim #12544) and a $125M ATM with Craig-Hallum (claim #12397) sat behind that debt. That phase reads as project finance with an equity backstop.

From early 2024 the cadence changes. Within roughly eight months the record shows a $20M unsecured AI Bridge loan (claim #6292), later prepaid in common stock (claim #6137). It also shows a $50M Series E preferred (claim #6119) and a $42.1M Yorkville prepaid advance (claim #6078), whose $16M monthly conversion cap (claim #6111) was subsequently eliminated (claim #5837). The same period adds a CIM facility with attached warrants (claims #6010, #6030), a $50M Series F convertible preferred (claim #5781), a $160M PIPE (claim #5782) and a $250M SEPA (claim #5780). Many of these convert or settle in stock, so dilution is layered and partly deferred rather than taken in a single issuance. The pattern resembles serial access to whatever structure was available more than a staged capital plan. Some instruments were also restructured or retired soon after issuance, including the early termination of the B. Riley loan (claim #6291).

On use of proceeds, the record ties the Starion loans to Ellendale. It also references project-level financing for a 400 MW hyperscaler lease (claim #3237) and repeated financing discussions for HPC Ellendale (claims #5867, #5907). The claims shown do not document a closed project-level facility, so deployment against those stated aims cannot be confirmed from this record.

Several limits apply. The data provided ends on 2024-10-03 and appears truncated. Several figures carry inconsistent units, such as $150 in claim #3241 and $110 in claim #6027, which likely denote millions but should be treated as unverified. No insider transactions appear in the data. Going forward, a reader should watch three things: whether the HPC Ellendale project financing closes on the terms discussed, how much SEPA and ATM capacity is actually drawn, and conversion activity under the Series F and Yorkville instruments, which will determine the realized share count.

The record runs from 2021 pre-IPO convertible preferred (claims 14500, 14501) to early October 2024 and then cuts off mid-entry. Nothing after that date can be assessed, and no insider transactions appear. The pattern falls into three phases, each more complex than the last. The IPO itself was revised downward: a firm-commitment offering of up to $75M (claim 14041) priced and closed at $40M (claims 14007, 14000). The next eighteen months relied mostly on conventional secured debt, namely the Starion term loans of $15M (claim 13804) and $20M (claim 13310), and the $50M B. Riley facility (claim 12565), of which only $36.5M was advanced initially (claim 12515). A $125M at-the-market agreement followed (claim 12397), which marked the shift toward continuous equity issuance. From February 2024 the cadence changes sharply. The company layered an unsecured $20M bridge (claim 6292), later partly settled in stock (claim 6137), onto the Yorkville prepaid advance of $42.1M (claim 6078). That advance's $16M monthly conversion cap (claim 6111) was subsequently eliminated (claim 5837). Next came the CIM secured facility with warrants attached (claims 5973, 6010), later described as a $200M tranche structure (claim 5660). Series E, Series E-1 and Series F preferred followed (claims 6119, 5656, 5781), along with a $160M PIPE (claim 5782) and a $250M SEPA (claim 5780). These instruments differ in seniority, dilution mechanics and cost, and several were amended or waived within weeks (claim 5760). The sequence reads more as successive patching than as a capital plan fixed in advance. Several plans were not executed as announced. Against the $50M Series E program, completed sales were $6.9M (claim 5657). The Yorkville registration deadline and its prepayment consequence (claim 6036) are recorded as delayed. Project-level financing for the 400 MW hyperscaler lease (claim 3237) and the HPC Ellendale facility (claims 5867, 5892, 5907) is described repeatedly as being in discussion, with no closing in the record. Use-of-proceeds language (claim 13807) is too thin to test against actual deployment. Several dollar figures are recorded in inconsistent units, which limits precise reconciliation. A careful reader should watch four things. First, whether the Ellendale project financing closes. Second, how much of the SEPA and ATM capacity is actually drawn. Third, the conversion activity on the Series F and the Yorkville instruments. Fourth, whether the CIM warrants and accordion draws (claims 6031, 5975) enlarge the share count further.

The record shows two distinct financing regimes. From the 2022 IPO through mid-2023, Applied Digital raised capital in modest, conventional increments. The IPO itself was revised from a $75M firm-commitment framing (claim 14041) to a $40M raise at $5.00 per share (claim 14007). Two Starion term loans of $15M and $20M followed (claims 13804, 13310), then a $125M ATM (claim 12397). From early 2024 the cadence changes sharply. Within roughly four months the company layered a $20M AI Bridge loan (claim 6292), a $42.1M Yorkville prepaid advance (claim 6078), a CIM facility that grew to $200M with attached warrants (claims 5660, 6010), a $160M PIPE (claim 5782), a $250M SEPA (claim 5780), $50M of Series F convertible preferred (claim 5781), and a $62.5M Series E-1 program (claim 5656).

The mix reads more as sequential bridging than a pre-planned capital stack. Debt was repeatedly retired with equity. The B. Riley loan was terminated early (claim 6291). The AI Bridge note was prepaid through $8.4M of common stock (claim 6137). The Yorkville monthly conversion cap of $16M (claim 6111) was later eliminated (claim 5837). Each of these moves shifts obligations from cash toward dilution, and the dilution arrives through several channels at once: warrants, convertible preferred, a SEPA, and an ATM.

On use of proceeds, the record is thin. Project-level financing for a 400 MW hyperscaler lease (claim 3237) and financing for the HPC Ellendale facility (claims 5867, 5892, 5907) are described as discussions rather than closed transactions. The data cannot confirm how capital was actually deployed. One gap requires explanation: the Series E program is framed as a $50M offering with $46M of net proceeds (claims 6119, 6120), yet completed sales total $6.9M (claim 5657).

The disclosed history ends in early October 2024 and contains no insider transactions, so any read of subsequent periods or insider alignment is not supported here. A careful reader should watch four things. First, whether the Ellendale and hyperscaler project financings close at the asset level rather than the parent. Second, the pace of SEPA draws and Yorkville conversions (claim 5624). Third, CIM accordion draws, which would trigger the contingent warrant issuance of 5,559,220 shares (claim 6031). Fourth, conversion of the Series F into its 7,598,714 underlying shares (claim 5834).

Applied Digital's disclosed financing history shows dilution accelerating sharply after a measured start. Between 2021 and 2023 the company relied on small preferred placements, an IPO that priced at 8,000,000 shares for $40,000,000 (claim #14007) against a stated capacity of up to $75 million (claim #14041), and secured bank debt such as the $15,000,000 Starion term loan (claim #13804) and the $20,000,000 Ellendale expansion loan (claim #13326). The $125,000,000 Craig-Hallum ATM (claim #12411) in mid-2023 was the first standing channel for continuous equity issuance. The pace changed in 2024. Within roughly five months the record adds the Series E preferred at $50,000,000 (claim #6119), the Yorkville prepaid advance at $42,131,579 (claim #6078), the CIM facility and its warrant tranches (claims #6010 and #6031), the Series F convertible preferred convertible into 7,598,714 shares (claim #5834), a $160,000,000 PIPE (claim #5782), a $250,000,000 SEPA (claim #5780), and a $62,500,000 Series E-1 program (claim #5656).

The 2024 mix reads more as sequential patching than a single plan. The AI Bridge loan was prepaid in common stock at $8,421,146 (claim #6137). The B. Riley facility was terminated early (claim #6291). Yorkville received consent shares (claim #6032). The Yorkville advance also carried a prepayment obligation if a registration statement did not go effective on time (claim #6036), and its monthly conversion limit of $16,000,000 (claim #6111) was later removed (claim #5837). Each of these moves solved an immediate constraint, and each widened the path to additional share issuance.

Use of proceeds is only partly traceable. The Ellendale expansion loan names a specific site, but project-level financing for the 400 MW hyperscaler lease (claim #3237) was still described as financing discussions in August 2024 (claims #5892, #5907). The Garden City escrow release is marked delayed in some filings (claims #3242, #5625) and stated in another (claim #5779). Both of these require explanation.

The record contains no insider transactions, and the supplied data ends in early October 2024. A careful reader should watch four things. The first is how quickly the SEPA and ATM are drawn relative to the share count. The second is the pace of Series F and YA note conversions (claim #5624). The third is whether accordion draws on the CIM facility trigger the contingent second warrant tranche. The fourth is whether the HPC Ellendale project financing closes on terms that reduce reliance on these equity-linked instruments.

The record shows a company that moved from small pre-IPO convertible preferred rounds to a layered, multi-counterparty capital stack within roughly two and a half years. Dilution was modest through 2023. The April 2022 IPO raised $40M on 8,000,000 shares at $5.00 (claim 14007), below the up to $75 million originally framed (claim 14041). Growth capital was then sourced mainly through secured debt: the $15M Starion term loan (claim 13804), a further $20M from Starion for Ellendale (claim 13310), and the $50M B. Riley facility (claim 12565). The inflection comes in mid-2024, when equity-linked instruments arrive in rapid succession. These include the $42.1M Yorkville prepaid advance (claim 6106), CIM borrowing with an initial $15M draw and 6,300,449 warrants attached (claims 5973, 6010), Series F convertible preferred at $50M (claim 5781), a $160M PIPE (claim 5782), and a $250M standby equity purchase agreement (claim 5780), alongside the $125M ATM (claim 5778).

The mix reads less as a single designed structure than as successive layering. The 2024 instruments overlap in timing and several carry terms that were later revised. The monthly conversion cap on the Yorkville advance (claim 6111) was subsequently eliminated (claim 5837). The CIM facility required a waiver to permit an additional $20M (claims 5760, 5661) before being described as a $200M tranche structure (claim 5660). Registration-dependent obligations tied to Yorkville are marked delayed (claims 6108, 6036). The $50M Series E preferred program (claim 6119) had completed only $6.9M in sales (claim 5657) when a Series E-1 offering of $62.5M was announced (claim 5656). That gap requires explanation.

On use of proceeds, the record ties the Starion debt to the Ellendale expansion (claim 13326). It also references project-level financing for a 400 MW hyperscaler lease (claim 3237) and advanced financing discussions for the HPC Ellendale facility (claims 5907, 5892). The data does not show a closed project-level facility or a deployment reconciliation, so whether announced purposes tracked actual spending cannot be assessed from these claims. No insider transactions appear in the supplied record, and the claim list is truncated in October 2024.

A careful reader should watch several items. The first is whether the discussed project-level HPC financing closes and displaces the corporate-level equity lines. The second is how heavily the SEPA and ATM are drawn. The third is whether the CIM accordion is exercised, since it triggers the additional 5,559,220 warrants (claim 6031). The fourth is whether Series E-1 sells through where Series E did not.

The record supplied runs from August 2021 to early October 2024 and ends mid-entry, so this read covers that window only. No insider transactions appear in the data. Several figures are recorded in inconsistent units, such as the $75 underwritten offering in claim #14041 and the $150 funding in claim #3241, which limits precise sizing of the later capital stack. Through 2022 and most of 2023, financing was conventional and modest in scale. The IPO raised $40,000,000 at $5.00 per share, per claim #14007. It was followed by Starion Bank term loans of $15,000,000 and $20,000,000 (claims #13804 and #13326), a $175,000,000 shelf (claim #12544), a $50,000,000 B. Riley term loan at 9% (claims #12565 and #12517), and a $125,000,000 at-market agreement (claim #12397). The cadence changes sharply in 2024. Within roughly five months the company added an unsecured $20,000,000 AI Bridge loan (claim #6292) and a $42,131,579 Yorkville convertible note with a $16,000,000 monthly conversion limit (claims #6106 and #6111). It also added a CIM facility carrying a warrant figure of 6,300,449 (claim #6010), a $50,000,000 Series F convertible preferred (claim #5781), a $160,000,000 PIPE (claim #5782), a $250,000,000 standby equity purchase agreement (claim #5780), and a $62,500,000 Series E-1 preferred program (claim #5656). The mix migrated from senior bank debt toward equity-linked and discount-convertible structures. Several individual terms suggest that financings were adjusted after the fact rather than following a pre-set plan. The conversion limitation on the prepaid advance was later eliminated (claim #5837). Consent shares were issued to Yorkville (claim #6032). A prepayment obligation was tied to a registration deadline (claim #6036). Stated use of proceeds is thinly documented. The 2022 loan use in claim #13807 carries no detail, and HPC Ellendale financing is repeatedly described as under discussion (claims #5907, #5892, #5867) without a closed project-level facility appearing in the window. The project-level financing for the 400 MW lease in claim #3237 likewise has no recorded close. The Garden City escrow release appears more than once flagged as delayed (claims #3242 and #5625). A careful reader should watch four things: whether Ellendale project financing actually closes or is no longer discussed; how much of the at-market and standby equity capacity is drawn concurrently; the cumulative dividend load from preferred series paying 8% to 9% (claims #6136 and #5835); and whether CIM accordion draws trigger the contingent warrants described in claim #6031. Data after October 2024 would be required to judge how these commitments resolved.

The record shows a company whose financing grew in size, speed and structural complexity over three years, with a sharp change in 2024. The early years were modest. Convertible preferred issues ran at $660,000 and $1.3M (claims #14500, #14501). The IPO priced at $40M (claim #14007), below the $75M first described (claim #14370). The 2022–2023 period used conventional secured bank debt from Starion, a $15M loan (claim #13804) and a $20M loan (claim #13326). Shelf and ATM capacity was put in place at $175M (claim #12544) and $125M (claim #12397). The $50M B. Riley facility at 9% (claims #12514, #12517) marked the first step toward costlier capital.

In 2024 the approach changes from planned layering to a run of bridge instruments, each quickly replaced by the next. A $20M unsecured AI Bridge loan (claim #6292) was later prepaid in stock (claim #6137). The B. Riley loan was terminated early (claim #6291). The Yorkville prepaid advance of $42.1M (claim #6078) came with several added terms: a consent share issuance (claim #6032), a prepayment trigger tied to S-1 effectiveness (claim #6036), and later removal of its conversion limit (claim #5837). Taken together, these read as opportunistic patching rather than a set sequence. The Series E preferred illustrates the gap between intent and uptake. It was launched at $50M (claim #6119) but had completed $6.9M in sales by September (claim #5657), and a $62.5M Series E-1 followed (claim #5656). That shortfall requires explanation. The September 2024 package appears more coherent. It combined a $160M PIPE (claim #5782), $50M Series F convertible preferred (claim #5781), a $250M SEPA (claim #5780) and an expanded CIM facility of $200M (claim #5660).

The stated uses of proceeds point steadily to Ellendale expansion (claim #13326) and HPC buildout. However, repeated references to 'financing discussions' for the HPC Ellendale facility (claims #5907, #5892, #5867) and to project-level financing for a 400 MW lease (claim #3237) are not matched by a disclosed close in this record. Actual deployment is not documented. The $25M Garden City escrow release appears in both delayed and stated forms (claims #3242, #5779).

The record contains no insider transactions, and it ends in October 2024 with claim #5619 cut off. From here, readers should track four things: SEPA draws, conversion of the Series F into 7,598,714 shares (claim #5834), warrant exercises estimated at $30M (claim #5725), and whether the project-level financing is ever confirmed or simply stops being mentioned.

The record shows two distinct phases. Through 2023 the company financed itself conventionally: a $40M IPO at $5.00 per share (claim #14007), Starion Bank term loans of $15M and $20M (claims #13804, #13310), a $50M B. Riley facility (claim #12514), and a $125M at-the-market agreement (claim #12411) sitting beside a $175M shelf (claim #12544). From early 2024 the cadence quickens and the instruments grow more equity-linked. Within roughly eight months the filings add a $20M unsecured bridge later prepaid in stock (claims #6292, #6137), a $42.1M Yorkville prepaid advance (claim #6078), a CIM note with warrants attached (claims #6026, #6010), Series F convertible preferred (claim #5781), a $160M PIPE (claim #5782), a $250M standby equity purchase agreement (claim #5780), and Series E-1 preferred at $62.5M (claim #5656). Dilution in this period comes less from any single offering than from overlapping conversion, warrant and drawdown channels that can operate at the same time.

The mix reads as layered rather than sequenced. Several changes suggest terms were renegotiated under timing pressure. Yorkville received shares in exchange for consent (claim #6032), a prepayment obligation depended on a registration deadline (claim #6036), and the $16M monthly conversion cap (claim #6111) was later removed (claim #5837). Series E is a related case. It launched as a $50M offering (claim #6119), had recorded completed sales of $6.9M (claim #5657), and was then followed by a new E-1 series. That gap requires explanation.

On deployment, early proceeds are explicitly tied to Ellendale expansion (claim #13326). Later, financing for the HPC Ellendale facility is described only as in discussion, across three filings on the same date (claims #5867, #5892, #5907). Project-level financing for a 400 MW hyperscaler lease (claim #3237) is announced, but the record does not show it closing. Several dependent items, including the Garden City escrow release (claims #3242, #5625), carry a delayed status.

A careful reader should track three things next. The first is actual conversion volume under the uncapped Yorkville note and draws under the SEPA. The second is whether the Series F registration rights (claim #5836, delayed) are satisfied. The third is whether HPC or project-level debt closes, which would reduce reliance on equity-linked patching. No insider transactions appear in the data provided, and the record is truncated in October 2024, so nothing here speaks to later developments.

The record as supplied ends mid-entry in October 2024 and contains no insider transactions, so this read covers financing behavior only, through that date. The cadence is steady but thin through 2023 and then accelerates sharply. The early years are modest: small convertible preferred tranches in 2021 (claims #14501, #14500), then an IPO first sized at up to $75M (claim #14370) that priced and closed at $40M (claims #14020, #14000), a downward revision that set a smaller equity base than planned. Through 2023 the mix reads as coherent. Bank term debt from Starion ($15M and $20M, claims #13804, #13326), the second tied explicitly to the Ellendale expansion, was followed by a $175M shelf (claim #12544) and a $125M ATM (claim #12397). That is a conventional layering of secured debt and registered equity capacity. The 9% B. Riley facility (claims #12514, #12517) marks the first move toward costlier capital. The 2024 record looks more like opportunistic patching. Inside roughly five months the company terminated the B. Riley loan early (claim #6291) and replaced it with an unsecured $20M bridge (claim #6292), which was partly prepaid in stock (claim #6137). It also took a $42.1M Yorkville prepaid advance with a $16M monthly conversion cap (claims #6078, #6111), a cap later eliminated (claim #5837). Other layers include a CIM facility carrying warrants on 6,300,449 shares (claim #6010), $50M of Series F convertible preferred (claim #5781), a $160M PIPE (claim #5782) and a $250M SEPA (claim #5780). Several entries show consent-fee share issuances to Yorkville (claim #5988). The Series E preferred is instructive: it was offered at $50M (claim #6119), only $6.9M was sold (claim #5657), and a $62.5M Series E-1 followed (claim #5656). That sequence requires explanation. On use of proceeds, the stated purposes (Ellendale expansion, claim #13326, and project-level financing for a 400 MW hyperscaler lease, claim #3237) are consistent across filings. The record, however, contains no deployment data against which to test them. The repeated "financing discussions" for HPC Ellendale (claims #5867, #5907, #5892) indicate that the project-level capital had not closed as of these filings. A careful reader should watch four things next. The first is whether that project-level financing closes, and on what terms relative to the CIM facility. The second is how much of the SEPA and ATM capacity is actually drawn. The third is the conversion pace on the Yorkville notes and Series F now that limits have been relaxed. The fourth is whether the preferred series continue to be re-launched rather than filled.

Applied Digital's record moves from micro-scale convertible preferred issuance to a layered capital stack. The 2021 Series C and Series D preferred rounds were small, at $660,000 and roughly $1.3 million (claims 14500 and 14501). The IPO was described as an underwritten offering of up to $75 million (claim 14041) but closed as 8,000,000 shares at $5.00 for $40 million (claims 14007 and 14000). That is a revision downward at the first public test. Through 2022 and 2023 the company leaned on project-adjacent debt: the $15 million and $20 million Starion term loans (claims 13804 and 13310) and the $50 million B. Riley facility (claim 12514). A $125 million ATM (claim 12397) sat behind them as a dilution backstop rather than a primary source.

The 2024 pattern is materially different in cadence and instrument mix. Within roughly four months the record shows a $20 million unsecured bridge (claim 6292), later partly retired through $8.4 million of common stock (claim 6137). It also shows the $42.1 million Yorkville prepaid advance (claim 6078), a CIM facility with a $15 million initial draw plus $110 million of subsequent availability and a $75 million accordion (claims 5973, 5974 and 5975), and warrants attached to that facility (claim 6010). The same stretch adds the $50 million Series F convertible preferred (claim 5781), the $160 million PIPE (claim 5782) and the $250 million SEPA (claim 5780). Several of these carry registration-deadline or consent mechanics, including the conditional Yorkville prepayment trigger (claim 6036) and consent shares (claim 5988). Together this reads more as sequential gap-filling around a funding need than as a single pre-planned structure, although the PIPE marks a shift toward larger, cleaner equity. The Series E program warrants note: a $50 million offering (claim 6119) had produced about $6.9 million in completed sales by September (claim 5657), and was followed by a Series E-1 offering (claim 5656).

On use of proceeds, the record repeatedly references project-level financing for a 400 MW hyperscaler lease (claim 3237) and advanced financing discussions for the HPC Ellendale facility (claims 5867 and 5907). No closed project financing appears in the data provided, so whether proceeds are deployed as described cannot be confirmed from this record. No insider transactions are included.

A careful reader should watch four things. The first is whether the Ellendale project financing closes or stops being discussed. The second is conversion activity once the Prepaid Advance conversion limitation is eliminated (claim 5837). The third is SEPA drawdown pace. The fourth is whether CIM accordion draws trigger the contingent warrant tranche (claim 6031). The record provided ends in October 2024, so later events are not reflected here.

The record shows Applied Digital's capital formation moving from small and occasional to continuous. The early period is modest: a $40M IPO (claims 14007, 14000), followed by conventional bank debt from Starion, $15M (claim 13804) and then $20M tied to the Ellendale expansion (claim 13326). From mid-2023 the instruments become larger and more frequent. A $50M B. Riley term loan (claim 12565) and a $125M ATM (claim 12411) come first. By the summer and early autumn of 2024, new facilities appear almost every few weeks: the $42.1M Yorkville prepaid advance (claim 6106), the CIM note with $110M of subsequent borrowing and a $75M accordion (claims 5974, 5975), a $50M Series F convertible preferred (claim 5781), a $160M PIPE (claim 5782), a $250M SEPA (claim 5780), and a $62.5M Series E-1 offering (claim 5656). Dilution risk grew accordingly. Much of it is contingent rather than realized, sitting in warrants, conversion features and standby equity.

The mix reads as layered rather than sequenced. Several instruments were retired or amended soon after they were put in place. The B. Riley loan was terminated early (claim 6291). The AI Bridge note was prepaid in common stock (claim 6137). The Yorkville conversion limitation was later eliminated (claim 5837). Consent shares were issued to Yorkville (claim 5988). Warrant tranches were attached to CIM borrowing (claims 6010, 5968). Taken together, this looks more like managing near-term liquidity against a growing build program than executing a fixed capital plan. The preferred series also add fixed dividend obligations in the 8–9% range (claims 13312, 6136, 5835).

Use of proceeds is only partly traceable. The Ellendale term loan is specifically tied to that facility (claim 13326), and project-level financing is cited for a 400 MW hyperscaler lease (claim 3237). Financing for the HPC Ellendale facility, however, appears only as discussions across three claims (5867, 5892, 5907), with no closed facility in the record; that requires explanation. The $25M Garden City escrow release is recorded as delayed more than once (claims 3242, 5625). The Series E program shows a gap between the offering and actual uptake: $50M offered (claim 6119) against $6.9M completed (claim 5657).

The record contains no insider transactions, and it ends in October 2024, so any read of later deployment is not supported here. A sophisticated reader should watch four things: actual draws under the SEPA and ATM, exercise of the CIM accordion and the related warrants, whether HPC Ellendale financing moves from discussion to a closed facility, and how the preferred dividend load is serviced.

Applied Digital's record reads as two different companies. Through early 2023 the financing is modest and sequential: small convertible preferred tranches, an IPO that closed at $40,000,000 (claim 14000) against an earlier stated ambition of $75,000,000 (claim 14370), and bank debt from Starion sized at $15,000,000 and then $20,000,000 (claims 13804, 13310). Mid-2023 marks the shift toward shelf capacity, with a $175,000,000 S-3 (claim 12544) and a $125,000,000 at-the-market program (claim 12411) alongside a 9% B. Riley term loan of $50,000,000 (claims 12565, 12517). By 2024 the cadence compresses sharply. Within roughly five months the company layers an unsecured bridge of $20,000,000 (claim 6292), Series E preferred of $50,000,000 (claim 6119), a Yorkville prepaid advance of $42,131,579 (claim 6078), a warrant-bearing CIM facility (claims 5973, 5974, 5975, 6010), Series F convertible preferred of $50,000,000 (claim 5781), a $160,000,000 PIPE (claim 5782), a $250,000,000 standby equity line (claim 5780), and Series E-1 at $62,500,000 (claim 5656).

The mix looks less like a single capital plan than a sequence of instruments, each addressing the obligations of the one before. The bridge note was retired partly through common stock (claim 6137). The Yorkville note carried a registration deadline that triggered a conditional prepayment obligation (claims 6036, 6108). A monthly conversion cap (claim 6111) was later eliminated (claim 5837). The B. Riley loan was terminated early (claim 6291). Each step is individually explicable, but together they shift the burden steadily toward common holders through conversions, warrants, and equity lines.

On use of proceeds, the Starion loans are tied explicitly to Ellendale (claim 13326). The larger 2024 raises are framed around HPC Ellendale, yet three claims on the same date describe financing only as being in discussion (claims 5867, 5892, 5907), and the project-level financing for the 400 MW lease (claim 3237) is not yet documented as closed. The Garden City escrow release of $25,000,000 (claim 5779) appears as delayed in several claims.

The record provided ends in October 2024 and contains no insider transactions, so no view on insider behavior is possible. A reader should watch for three things: whether the Ellendale project financing is documented as closed, the pace of SEPA and ATM draws, and whether CIM accordion draws trigger the contingent warrants (claim 6031).

The cadence compresses sharply. From seed-scale preferred issuances in 2021 — the $660,000 Series C and $1.3M Series D, per claims 14500 and 14501 — through the 2022 listing, capital events arrive roughly quarterly; by mid-2024 they arrive weekly. The week of September 10-12, 2024 alone carries a $160M private placement (claim 5783), a $250M standby equity purchase agreement (claim 5780), a $50M Series F (claim 5781), a $125M ATM (claim 5778) and a $20M incremental CIM draw (claim 5724). Four dilution channels opened in parallel rather than in sequence is the defining structural fact here. Worth noting earlier: the IPO marketed at $75M (claims 14370, 14041) priced at $40M on 8,000,000 shares at $5.00 (claim 14007) — revised downward before it ever reached scale.

The mix shows a coherent asset thesis financed through an increasingly improvised instrument stack. The Starion facilities, $15M and then $20M for Ellendale expansion (claims 13804, 13326), are ordinary matched-asset lending. The inflection is the AI Bridge $20M unsecured note (claim 6292) being prepaid in common stock (claim 6137) rather than retired in cash. After that point the structures turn progressively equity-linked: the Yorkville prepaid advance of $42,131,579 (claim 6106) carried a $16M monthly conversion cap (claim 6111) that was subsequently eliminated (claim 5837), and CIM warrants layer at 6,300,449, 2,964,917 and a further 5,559,220 contingent on accordion draws (claims 6029, 6030, 6031). The $100,000 share issuance to Yorkville purely to obtain consent (claim 6032) indicates covenant friction between concurrently live financings.

Stated use of proceeds stays consistent around Ellendale, but deployment lags disclosure. Project-level financing for the 400 MW hyperscaler lease announced April 2024 (claim 3237) is still characterized as discussions four months later (claims 5867, 5892, 5907), and the corporate-level stack that followed suggests project capital did not arrive on the announced timeline. The $25M Garden City escrow release recurs across August, September and October 2024 (claims 3242, 5779, 5625) without a completion marker — requires explanation.

Watch SEPA drawdown pace and discount relative to the ATM, the stacked 8%/9%/8% preferred dividends (claims 13312, 6136, 5835) ranking ahead of common, and accordion warrant triggers. No insider transactions appear on record, so that dimension cannot be assessed. The record also terminates in October 2024.

The cadence here is not a ramp but a step function. From the 2021 preferred rounds (per claims #14500 and #14501, $660,000 and $1,300,000) through the April 2022 listing, capital arrived in small, conventional increments, and the equity story was already revised once at the gate: the $75,000,000 contemplated offering (per claims #14370 and #14041) priced at $40,000,000 (per claims #14020 and #14000). For roughly eighteen months thereafter the company financed itself through senior bank debt at Starion (per claims #13804 and #13310), which is the cheapest capital on this record and the last of it. Everything after the May 2023 shelf (per claim #12544) is structured.

The mix shifts from a coherent asset-backed strategy to something closer to continuous patching. The B. Riley term loan at 9% (per claims #12565 and #12517) was terminated early (per claim #6291), replaced by an unsecured bridge (per claim #6292) that was itself prepaid in common stock (per claim #6137). By mid-2024 the instruments are stacked rather than sequenced: a Yorkville prepaid advance of $42,131,579 with a $16,000,000 monthly conversion ceiling (per claims #6078 and #6111), CIM borrowings of $15,000,000 with $110,000,000 subsequent and a $75,000,000 accordion (per claims #6026, #6027, #6028), each tranche carrying warrants (per claims #6029, #6030, #6031). That the company issued $100,000 in shares simply to obtain Yorkville's consent (per claim #6032) is a cost of capital datapoint, not a financing. In September 2024 a $160,000,000 PIPE, $50,000,000 Series F, $250,000,000 SEPA and $125,000,000 ATM were live concurrently (per claims #5782, #5781, #5780, #5778) — authorized dilution capacity well above anything drawn.

Use of proceeds tracks where it is specific: Ellendale expansion (per claim #13326), the Garden City escrow released after an initial delay (per claims #3242 and #5779). Where it does not track is the 400 MW hyperscaler project financing (per claim #3237), still described as advanced discussions months later (per claims #5892, #5907, #5867) — no longer discussed in closed form on this record.

Watch SEPA and ATM utilization against authorization, accordion draws triggering further warrants, and accruing 8–9% preferred dividends (per claims #5835, #6136). The record contains no insider transactions and terminates in October 2024; conclusions past that date are unsupported.

The trajectory here is not gradual. From the Series C and D preferred issuances of $660,000 and $1,300,000 (claims 14500, 14501) through an IPO that priced at $40,000,000 against an initially contemplated $75,000,000 (claims 14007, 14370, 14041), the company raised in its first two years roughly what it would later raise in a single week. By September 2024 the cadence had compressed to the point that five distinct vehicles — a $160,000,000 private placement, a $50,000,000 Series F, a $250,000,000 standby equity purchase agreement, a $125,000,000 ATM, and $20,000,000 of additional CIM borrowing (claims 5724 through 5729) — carry the same date. That is not a financing plan sequenced against milestones; it is capacity assembled simultaneously.

The mix reads as opportunistic patching layered over a genuine project-finance ambition. Bank debt from Starion was explicitly tied to Ellendale expansion (claims 13804, 13326), which is coherent. What followed is less so: the B. Riley $50,000,000 facility was terminated early (claims 12565, 6291), then a further B. Riley loan was taken and repaid early (claims 6049, 6050); the $20,000,000 AI Bridge note was prepaid in common stock (claims 6292, 6137); and the CIM structure attached three separate warrant tranches totaling roughly $14,800,000 in underlying stock (claims 6029, 6030, 6031). Equity-linked cost was accepted repeatedly to keep debt available.

On deployment, the gap that most requires explanation is the Series E. A $50,000,000 offering with $46,000,000 of stated net proceeds (claims 6119, 6120) shows completed sales of $6,900,000 by September (claim 5657), after which a Series E-1 was introduced at $62,500,000 and then at $2,500,000 (claims 5656, 5651). Separately, the HPC Ellendale facility financing appears three times on a single date as still in discussions (claims 5867, 5892, 5907), and the eventual September package was corporate and dilutive rather than project-level, despite the earlier framing of project-level financing for the 400 MW hyperscaler lease (claim 3237). The $25,000,000 Garden City escrow release recurs from August into October (claims 3242, 5779, 5625).

What to watch: actual drawdown against the $250,000,000 SEPA and $125,000,000 ATM, whether Series E-1 repeats Series E's subscription shortfall, and whether project-level capital ever displaces holdco dilution. The record here ends in early October 2024; nothing after that is observed.

The record divides cleanly into two regimes. Through 2022 and early 2023 the capital stack was small and conventional: modest Series C and D preferred rounds (claims 14500, 14501, 14355, 14356), an IPO that priced at $40M (claim 14007) against a firm-commitment target of up to $75M (claim 14041), and secured bank debt from Starion tied explicitly to Ellendale expansion (claims 13804, 13326). That is a coherent sequence — project debt against project assets, with stated use of proceeds (claim 13807) tracking the land and facility spend (claim 12622).

The second regime begins with the $175M shelf (claim 12544) and the first $125M ATM (claim 12397) in mid-2023 and accelerates sharply. By mid-2024 the company was running at least six financing channels concurrently: the Yorkville prepaid advance of $42.1M (claim 6078), the CIM facility with $15M initial, $110M subsequent and a $75M accordion (claims 6026, 6027, 6028), Series E at $50M (claim 6119), Series E-1 at $62.5M (claim 5656), Series F at $50M (claim 5658), a $160M PIPE (claim 5782), a $250M standby equity purchase agreement (claim 5784) and a second $125M ATM (claim 5778) — most of it within a four-month window. Cadence compressed from roughly two events a year to a dozen in a quarter.

Several features require explanation rather than alarm. Warrants were attached at each CIM tranche (claims 6029, 6030, 6031), and $100,000 in shares was issued to Yorkville purely for consent (claim 6032), which indicates covenant friction between overlapping instruments. Against $62.5M of Series E-1 authorization, only $6.9M of Series E sales were recorded as completed (claim 5657). The Garden City escrow release of $25M was flagged delayed in April (claim 3242) and restated in September (claim 5779). Multiple items cluster around registration-statement effectiveness (claims 6108, 6036).

Watch whether the accordion is drawn and the contingent warrants issue, SEPA versus ATM utilization, and the cash cost of the stacked 8–9% preferred dividends (claims 13312, 6136, 5835). The record contains no insider transaction data and ends in October 2024; nothing here supports conclusions about the subsequent two years.

The cadence here compresses in a way that is itself the story. Between 2021 and mid-2022 the company raised in discrete, spaced events: small preferred rounds (the $1.3M Series D and $660K Series C, per claims 14501 and 14500), then a public listing that priced at $40M on 8,000,000 shares at $5.00 (claim 14007) against the $75M firm-commitment size contemplated earlier (claim 14041). That gap between contemplated and realized size is the first instance of a pattern that recurs. Through 2023 the rhythm was quarterly and mostly secured: Starion term loans of $15M and $20M (claims 13804, 13326), a $175M shelf (claim 12544), a $50M B. Riley facility (claim 12565, marked delayed), and a $125M ATM (claim 12397). By mid-2024 the interval collapses to weeks, and the instruments change character.

The 2024 mix does not read as a single financing strategy executed in stages. The Yorkville prepaid advance of $42,131,579 (claim 6078), with monthly conversion limits of $16M (claim 6111) and consent shares issued to the same counterparty (claims 6032 and 5988), sits alongside a CIM structure that grew from a $15M initial borrowing to $110M of subsequent availability plus a $75M accordion (claims 6026, 6027, 6028), each tranche carrying its own warrant issuance (claims 6029, 6030, 6031). Then, within a single September window, a $160M private placement, a $250M SEPA, a $125M ATM, a $50M Series F and a $62.5M Series E-1 (claims 5783, 5784, 5778, 5781, 5656) were all live simultaneously. Capacity was being assembled from every available channel at once rather than sequenced.

On use of proceeds, the most instructive datapoint is absorption rather than announcement: the Series E offering sized at $50M (claim 6119) shows completed sales of $6,900,000 (claim 5657). Separately, project-level financing for the 400 MW hyperscaler lease (claim 3237) and the HPC Ellendale facility remained in "advanced discussions" across claims 5867, 5892 and 5907 without a recorded closing in this record; the dilutive stack appears to have filled that interval.

What merits watching: whether the Ellendale project financing converts from discussion to executed terms, the realized draw against the $250M SEPA and $125M ATM authorizations, and the warrant overhang from the CIM tranches. The record here ends in October 2024 and contains no insider transactions at all, which limits any read on alignment.

The trajectory here is not a series of discrete raises but a widening funnel. The first three years are conventional and small: two tranches of Series C/D preferred in 2021 totaling under $4M (claims 14500, 14501, 14355, 14356), then an IPO that priced at $40M on 8,000,000 shares at $5.00 (claim 14007) against a firm-commitment structure described at up to $75 million (claim 14041) and an earlier $75M figure (claim 14370). That gap between announced capacity and executed size is the earliest instance of a pattern that repeats.

From 2022 through early 2023 the mix is coherent and asset-matched: Starion Bank term loans of $15M and $20M (claims 13804, 13310) with stated use of proceeds (claim 13807) and an explicit tie to Ellendale expansion (claim 13326). Debt sized to a facility, secured by a facility, is a legible strategy.

Mid-2023 marks the inflection. The $175M shelf (claim 12544) and the $125M ATM with Craig-Hallum (claim 12397) establish standing dilution capacity, and by 2024 nearly every available instrument is running concurrently: the $42,131,579 Yorkville prepaid advance (claim 6078), the CIM structure of $15M initial plus $110M subsequent plus a $75M accordion with three separate warrant tranches (claims 6026, 6027, 6028, 6029, 6030, 6031), Series F at $50M (claim 5728), a $160M private placement (claim 5729), a $250M SEPA (claim 5727), a second $125M ATM (claim 5726), and Series E-1 at $62.5M (claim 5656). Preferred dividend rates ratchet from 8.0% (claim 13312) to 9% (claim 6136) and back to 8% on Series F (claim 5835). Repaying the AI Bridge note in common stock (claim 6137) and issuing shares to obtain Yorkville consent (claim 6032) indicate equity used as transactional currency rather than as funding.

On use of proceeds, the recurring reference to advanced project-level financing for the HPC Ellendale facility and the 400 MW hyperscaler lease (claims 3237, 5867, 5892, 5907) sits uneasily beside the corporate-level dilutive instruments actually funding the period; the substitution requires explanation.

The record here is truncated in October 2024 and contains no insider transaction data, so no read on insider behavior is possible. Watch actual draw rates against SEPA and ATM authorizations, warrant overhang from the accordion tranches, cumulative preferred dividend accrual, and whether project-level financing closes.

The trajectory divides cleanly into two regimes. Through 2021 and 2022 the company raised in small, discrete increments — the $1.3M Series D and $660K Series C (claims 14501, 14500), followed by an IPO marketed as a firm commitment of up to $75 million (claim 14041) that priced at $40 million (claims 14007, 14020, 14000). That shortfall between announced capacity and executed size is the earliest instance of a pattern that recurs. Financing then shifted to asset-level bank debt at Starion, $15M and $20M (claims 13804, 13326), which is coherent with a build-out thesis and is the one stretch where the instrument matches the use — claim 13807 is the only explicit use-of-proceeds statement in the record.

From mid-2023 the cadence compresses. The $175M shelf (claim 12544) and the $125M ATM (claims 12397, 12398, 12411) established standing dilution capacity before a specific need was articulated. In 2024 the spacing collapses to weeks: the Yorkville prepaid advance of $42.1M with a $16M monthly conversion cap (claims 6078, 6106, 6111), the CIM structure of $15M initial plus $110M subsequent plus a $75M accordion with three tranches of attached warrants (claims 6026, 6027, 6028, 6029, 6030, 6031), then within September a $160M PIPE, a $250M SEPA, Series F at $50M, Series E-1 at $62.5M and a renewed $125M ATM (claims 5783, 5784, 5781, 5656, 5754). Repaying the AI Bridge note in common stock (claim 6137) and issuing $100,000 in shares purely for Yorkville's consent (claims 6032, 5988) indicate terms renegotiated under constraint rather than a sequenced plan.

The mix reads as opportunistic patching layered over a genuine project need: Ellendale HPC is named consistently (claims 13326, 5867, 5892, 5907), but the instruments funding it are overwhelmingly equity-linked rather than project-level. Several items remain marked delayed, including the B. Riley $50M facility (claims 12514, 12565) and the 2022 SparkPool, GMR and Valuefinder issuances (claims 13751, 13752, 13753), which require explanation.

Two gaps limit this read. No insider transactions appear on the record at all, so no alignment inference is available. The record as supplied terminates in early October 2024 and is truncated mid-entry at claim 5619; nothing here supports conclusions about the subsequent two years. A sophisticated reader should watch whether SEPA and ATM capacity is actually drawn, and whether the delayed items are confirmed, revised, or no longer discussed.

The record runs from August 2021 to early October 2024 and appears truncated after that point. It contains no insider transactions, so this read covers only the financing sequence. The cadence has a clear shape. Small convertible preferred issuances in 2021 (claims 14500 and 14501) gave way to an April 2022 IPO of 8,000,000 shares at $5.00 for $40M (claim 14007). The company then leaned on bank debt: the $15M Starion loan (claim 13804), a second $20M Starion loan for Ellendale (claim 13326), and the $50M B. Riley facility at 9% (claims 12565 and 12517). The mid-2023 $175M shelf (claim 12544) and $125M ATM (claim 12397) marked the turn toward equity capacity. From May 2024 the pace accelerates sharply. That period brings the Yorkville prepaid advance of $42.1M with a $16M monthly conversion cap (claims 6078 and 6111), a cap later eliminated (claim 5837). It also brings the CIM note with $15M initial, $110M subsequent and a $75M accordion (claims 5973, 5974 and 5975), carrying 6,300,449 warrants (claim 6010). By September 2024 the list adds Series E, E-1 and F preferred, the latter convertible into 7,598,714 shares (claim 5834), a $160M PIPE (claim 5782) and a $250M SEPA (claim 5780). Dilution potential therefore steepens through 2024 rather than levelling off. The mix reads less as a single planned capital structure and more as sequential bridging. Conventional bank debt was replaced by structured, equity-linked instruments, and the B. Riley loan was terminated early (claim 6291). The AI Bridge note was prepaid in stock, $8,421,146 per claim 6137. Several instruments carry conversion or warrant features that shift cost onto common holders. Use-of-proceeds tracking is limited by the data. The Starion loans are tied to Ellendale (claims 13807 and 13326). However, the project-level financing for the 400 MW hyperscaler lease (claim 3237) is still described only as 'financing discussions' as of August 2024 (claims 5892 and 5907). The Series E offering was sized at $50M (claim 6119), yet completed sales total $6.9M (claim 5657), a gap that requires explanation. A reader should next watch four things: whether the HPC Ellendale project financing closes; the pace of SEPA draws and Yorkville conversions; whether CIM accordion draws trigger the contingent warrants (claim 6031); and whether the E-1 and F preferred are followed by further series, which would indicate the bridge has not yet reached permanent financing.

The record as provided ends in early October 2024 and contains no insider transactions, so this read covers only the company-level financing pattern through that date. The early history is modest. Small convertible preferred issuances (claims #14500, #14501, #14355, #14356) preceded an IPO first described as up to $75 million (claims #14041, #14370) and then priced and closed at $40 million (claims #14007, #14000). That revision set the pattern: the company has repeatedly raised less per instrument than first described and made up the gap by adding instruments. Through 2022 and 2023 the mix was coherent and conventional. Starion Bank term loans of $15 million and $20 million (claims #13804, #13310) were tied explicitly to the Ellendale expansion (claim #13326). A $175 million shelf (claim #12544) and a $125 million ATM (claim #12397) sat behind a $50 million B. Riley facility at 9% (claims #12514, #12517). The 2024 record reads differently. The B. Riley loan was terminated early (claim #6291). A $20 million unsecured bridge (claim #6292) was partly retired in stock (claim #6137). A Yorkville prepaid advance of $42.1 million (claim #6106) first carried a $16 million monthly conversion limit (claim #6111), which was later eliminated (claim #5837). Consent shares (claim #6032) and a contingent prepayment tied to registration timing (claim #6036) accompanied it. The CIM note stacked a $15 million initial draw, $110 million of subsequent capacity and a $75 million accordion (claims #6007, #6008, #6009) with warrants (claim #6010). By September 2024 the company had layered a $160 million PIPE, $50 million of Series F, a $250 million SEPA and a $62.5 million Series E-1 on top (claims #5782, #5781, #5780, #5656). Several of these funded near-term obligations more than a single capital plan, which looks more like sequential patching than design. On use of proceeds, the 2022–2023 debt maps cleanly to Ellendale, including the land purchase (claim #12622). The later equity and convertible layers carry no comparable deployment disclosure in this record. Project-level financing for the HPC Ellendale build remains described as in discussion (claims #5892, #5907), not closed. The Series E offering also requires explanation: $50 million was registered (claim #6119), but only $6.9 million of completed sales were reported (claim #5657). A careful reader should look for four things in filings after this record ends: whether Ellendale project financing closed, how much of the SEPA was drawn, how Yorkville conversions proceeded after the cap was removed, and whether further CIM draws triggered the contingent warrant tranches (claim #6031).

The record for Applied Digital describes a financing profile that shifted in kind, not just in scale, over roughly three years. The early phase was modest and conventional: small convertible preferred rounds, such as the $1.3M Series D per claim 14501, followed by an IPO of 8,000,000 shares at $5.00 for $40M per claim 14007. From 2022 into 2023 the company leaned mainly on secured bank and specialty debt. That includes the $15M Starion term loan (claim 13804), a further $20M from Starion tied to Ellendale expansion (claim 13326), and the $50M B. Riley facility with a $36.5M initial advance (claims 12514 and 12515). Equity capacity was put in place alongside it through a $175M shelf (claim 12544) and a $125M ATM (claim 12397).

In 2024 both the pace and the mix changed. Within about five months the record shows several new instruments. There was a $20M unsecured bridge loan (claim 6292), later partly settled in stock at $8.4M (claim 6137). There was a $42.1M Yorkville prepaid advance (claim 6106), whose $16M monthly conversion cap (claim 6111) was subsequently eliminated (claim 5837). The CIM note carries warrants for 6,300,449 shares (claim 6010) and further draws and an accordion of $110M and $75M (claims 5974 and 5975). Series F preferred was issued for $50M (claim 5781), alongside a $160M PIPE (claim 5782) and a $250M standby equity purchase agreement (claim 5780). The B. Riley loan was terminated early (claim 6291). Taken together, this looks less like a planned capital structure and more like successive layering of instruments that each carry their own conversion or warrant overhang. The Series E program also requires explanation: it was sized at $50M (claim 6119), but completed sales are recorded at $6.9M (claim 5657), and it was followed by a new E-1 series (claim 5656).

On use of proceeds, the only consistent stated purpose is Ellendale and the planned HPC buildout. As of August 2024, financing for the HPC Ellendale facility was still described as discussions (claims 5867 and 5907), and project-level financing for the 400 MW lease was stated without terms (claim 3237). The record contains no deployment data, so whether announced purposes match actual spending cannot be assessed from these claims.

The record includes no insider transactions and appears to end in early October 2024. Several claims record amounts without scale, for example $75 in claim 14041, so only fully stated figures are cited here. Going forward, a careful reader should track the pace of draws under the SEPA, conversion of the Series F into roughly 7.6M shares (claim 5834), CIM accordion and warrant activity, and whether the HPC project financing moves from discussion to a closed, disclosed facility.

Applied Digital's disclosed record moves from small-scale preferred issuance to a dense, layered capital stack in about three years. The early period is modest: Series C and D convertible preferred of $660,000 and $1.3M (claims #14500, #14501), then an IPO that closed at $40M for 8,000,000 shares at $5.00 (claims #14007, #14000), below the up-to-$75M framing in claims #14370 and #14041. Through early 2023 the mix leans on site-linked bank debt, namely the $15M Starion term loan (claim #13804) and a $20M Starion loan for the Ellendale expansion (claim #13326). That pattern reads as project-matched financing.

The cadence changes from mid-2023, when a $175M shelf (claim #12544), a $125M ATM (claim #12397) and a 9% $50M B. Riley facility (claims #12514, #12517) arrive within weeks. The 2024 sequence overlaps heavily. The B. Riley loan was terminated early (claim #6291), and an AI Bridge loan was prepaid in common stock (claim #6137). The company also added a $42.1M Yorkville prepaid advance with a $16M monthly conversion limit (claims #6078, #6111), Series E and E-1 preferred (claims #6119, #5656), CIM debt carrying a 6.3M-share warrant (claim #6010), a $50M Series F convertible (claim #5781), a $160M PIPE (claim #5782) and a $250M SEPA (claim #5780). The elimination of the Yorkville conversion limitation (claim #5837) and prepayment triggers tied to registration timing (claim #6036) point to dilution that is increasingly equity-settled even where the headline instrument is debt. Taken together, the mix reads less as a single plan than as successive layers, added while project-level financing for the 400 MW hyperscaler lease (claim #3237) and the HPC Ellendale facility (claims #5867, #5907) remained in discussion.

Use of proceeds is difficult to test. The 2022 Starion loan's stated use (claim #13807) and the Ellendale-linked loan are the clearest ties. For the 2024 facilities the record lists amounts rather than deployment, and several carry a delayed status (claims #5753, #5756, #5741), which requires explanation rather than inference. Amounts are also recorded inconsistently, with some millions-scale figures stored as bare units (claims #5727, #5620), so aggregates should be read with care. No insider transactions appear in the supplied record, and the list is truncated in October 2024.

A careful reader should watch four items. The first is whether Ellendale project financing closes, and on what terms. The second is how much SEPA and ATM capacity is drawn. The third is whether CIM accordion draws trigger the contingent 5.56M-share warrant (claim #6031). The fourth is whether the Series F registration and redemption commitments (claims #5836, #5839) are met on schedule.

The record supplied here is truncated partway through October 2024 and contains no insider transactions, so this read covers financing structure only. It also cannot say whether later capital events changed the picture. Several amounts appear to be recorded in millions or as share counts rather than dollars, so figures are cited only where the scale is unambiguous.

Dilution began modestly and accelerated sharply. The early years consist of small convertible preferred issuances and an IPO that closed at $40M (claim #14000). That was below the $75M firm-commitment size described beforehand (claims #14370, #14041), so the offering was revised downward rather than completed as first framed. Through 2022 and 2023 the company leaned on secured bank debt, a $15M and then a $20M Starion term loan (claims #13804, #13310), while putting equity capacity in place through a $175M shelf (claim #12544) and a $125M ATM (claim #12397).

In 2024 the mix shifts from sequenced to layered. A $50M B. Riley facility (claim #12565) was terminated early (claim #6291). It was followed within months by several other instruments:

* an unsecured $20M bridge loan (claim #6292), later prepaid in common stock (claim #6137) * a $42.1M Yorkville prepaid advance (claim #6078), whose $16M monthly conversion limit (claim #6111) was later eliminated (claim #5837) * a CIM facility carrying tranche warrants (claims #6010, #6031) * by September, a $160M PIPE, a $50M Series F convertible preferred and a $250M SEPA announced together (claims #5782, #5781, #5780)

Taken together, the stack reads more as successive patching to fund construction than as a single capital plan. The pattern is consistent: debt repaid with stock, warrants attached to loans, and consent shares issued to an existing lender (claim #6032).

Uptake of retail-style preferred was slow. Only $6.9M of the $50M Series E had sold by September 2024 (claims #5657, #6119), yet a $62.5M Series E-1 was opened anyway (claim #5656). That gap requires explanation.

On use of proceeds, the stated purpose is consistently the Ellendale build-out (claims #13326, #13807). However, HPC Ellendale financing appears three times as "discussions" (claims #5907, #5892, #5867). The 400 MW project-level financing (claim #3237) does not show a closed transaction within this window.

A careful reader should watch four things next. First, the actual pace of SEPA and Yorkville conversions now that the monthly limit is gone. Second, whether CIM accordion draws trigger the contingent warrants. Third, the delayed Series F registration commitment (claim #5836). Fourth, whether project-level debt closes and displaces equity as the marginal funding source.

The record as supplied runs from the April 2022 IPO through early October 2024 and contains no insider transactions, so this read covers only the company-side capital stack. Several entries also appear to carry dollar amounts in millions without scaling, so those figures should be read with care. The trajectory falls into two phases. Through mid-2023 the company funded itself conventionally. It raised a $40M IPO (claim 14007), then took Starion bank debt of $15M and $20M, with the latter tied to the Ellendale expansion (claims 13804, 13326). A $50M B. Riley term loan at 9% followed (claims 12565, 12517), and a $125M ATM capped the period (claim 12397). From February 2024 the mix changed in character. An unsecured $20M AI Bridge loan (claim 6292) and early termination of the B. Riley facility (claim 6291) were followed by a Yorkville prepaid advance of $42.1M (claim 6078) with conversion-price mechanics (claim 6112). Share issuances to Yorkville in exchange for consents came next (claims 6032, 5988), along with a prepayment obligation tied to registration effectiveness (claim 6036). The later elimination of the prepaid advance's conversion limitation (claim 5837) warrants attention, because it widens the dilution path rather than narrowing it. Instruments were layered over short intervals: the CIM secured note with warrants for 6,300,449 shares (claim 6010), a $250M SEPA (claim 5780), $50M of Series F convertible preferred (claim 5781), and a $160M PIPE (claim 5782). Together these read less as a designed capital plan and more as successive patching until the larger PIPE and CIM capacity arrived. The retail preferred channel illustrates the gap between capacity and execution. The Series E offering was sized at $50M (claim 6119), but completed sales were $6.9M (claim 5657), and a Series E-1 offering of $62.5M was then layered on (claim 5656). On use of proceeds, the bank debt ties clearly to Ellendale. The larger 2024 raises coincide with project-level financing for a 400 MW hyperscaler lease that remained at the discussion stage in the record (claims 3237, 5907), so deployment cannot yet be matched to proceeds. A sophisticated reader should watch four things. The first is actual SEPA and ATM draw volumes against their headline capacity. The second is CIM accordion draws, since those trigger additional warrants (claim 6031). The third is conversion activity on the Yorkville notes and the Series F. The fourth is whether the Ellendale project financing closes on terms that let the company retire the higher-cost bridge instruments. Several facilities are recorded as delayed, and whether they were completed, revised or no longer discussed requires explanation in later filings beyond this record.

The record shows a funding approach that escalated sharply and grew more structurally complex. It also carries two limits. The claims provided end in October 2024, and no insider transactions appear in them. Several figures are recorded without consistent units, such as the $160 private placement in claim #5783 and the $105,000 borrowing in claim #5591, so magnitudes should be read with care.

Through 2022 and early 2023 the capital base was conventional. The $40M IPO (claim #14000) was followed by senior bank debt from Starion, $15M (claim #13804) and then $20M for Ellendale (claim #13326). The 2023 shelf of $175M (claim #12544) and the $125M ATM (claim #12397) added optional equity capacity. The B. Riley $50M term loan (claim #12565) brought more expensive debt at a 9% rate (claim #12517).

The 2024 sequence reads less like a plan set in advance and more like successive patching. The B. Riley loan was terminated early (claim #6291). The $20M AI Bridge unsecured loan (claim #6292) was partly retired in stock (claim #6137). The Yorkville prepaid advance of $42.1M (claim #6078) carried a $16M monthly conversion ceiling (claim #6111) and required consent shares (claim #6032). The CIM facility attached warrants for 6,300,449 shares (claim #6029), with more tied to accordion draws (claim #6031). Series E preferred was marketed at $50M (claim #6119), but completed sales total $6.9M (claim #5657). The September 2024 package was the first coordinated recapitalization in the record. It combined the Goldman-backed $160M PIPE (claim #5782), $50M of Series F convertible into 7,598,714 shares (claim #5834), the $250M SEPA (claim #5780) and a CIM waiver (claim #5760).

On use of proceeds, the record supports only a partial comparison. Early loan proceeds were tied to Ellendale (claims #13807, #13326). The intended long-term solution, project-level financing for the 400 MW hyperscaler lease (claim #3237), still appears only as financing discussions as of August 2024 (claims #5867, #5907). The Garden City escrow release is marked delayed (claims #3242, #5625). The record does not show actual deployment, so whether capital went where it was announced cannot be confirmed from this data.

The items to watch follow from that gap. The first is whether project-level financing closes and replaces the structured instruments. The others are how heavily the SEPA is drawn, when the Series F converts, whether the CIM accordion and its second warrant tranche are used, and any disclosure after October 2024 that would complete the insider picture.

The record supplied ends mid-entry in October 2024 and contains no insider transactions, so this read covers financing behavior only through that point and cannot speak to insider selling. Within that window, the trajectory changes sharply. Through 2022 and 2023 the company relied mainly on conventional secured debt: the $15,000,000 Starion term loan (claim 13804), the $20,000,000 Starion expansion loan (claim 13326), and the $50,000,000 B. Riley facility (claim 12514). Standing equity capacity was built alongside it through the $175,000,000 shelf (claim 12544) and the $125,000,000 ATM (claim 12397). In 2024 the pace and complexity rise materially. The $20,000,000 unsecured AI Bridge loan (claim 6292) and early termination of the B. Riley loan (claim 6291) came first. They were followed within roughly four months by the Yorkville prepaid advance of $42,131,579 (claim 6078), the CIM facility later described at $200,000,000 in tranches (claim 5660), the $50,000,000 Series F convertible preferred (claim 5781), the $250,000,000 SEPA (claim 5780), the $160,000,000 private placement (claim 5782), and the $62,500,000 Series E-1 program (claim 5656). The mix moved away from bank debt toward instruments that settle in shares. The AI Bridge note was prepaid through a common stock issuance (claim 6137), $11,392,978 of debt was converted to equity (claim 5593), and the Yorkville conversion limitation was removed (claim 5837). CIM lending also carried warrants (claim 6010). Taken together, this reads less as a single designed capital structure than as successive layers, each added while earlier ones were still being amended, waived, or refinanced (claim 5760). Use of proceeds is specific in the early loans, tied to Ellendale expansion (claim 13326). The later stated purposes, project-level financing for a 400 MW hyperscaler lease (claim 3237) and financing for the HPC Ellendale facility (claims 5867, 5907), appear in the record only as discussions. The dataset contains no deployment figures against which they can be checked. A sophisticated reader should watch four things. The first is actual draw-down against the SEPA and ATM relative to capacity. The second is conversion of the Series F preferred into the 7,598,714 shares referenced (claim 5834). The third is the Yorkville note maturity, which is marked delayed (claim 6107). The fourth is whether the Ellendale HPC financing moves beyond discussion. Several items marked delayed, including the Garden City escrow release (claim 5625), require explanation in subsequent filings.

Applied Digital's financing record describes a company whose capital needs outran its instruments, and whose mix grew more layered and more expensive over roughly eighteen months. The 2023 structure was conventional: the $20,000,000 Starion term loan tied to the Ellendale expansion (claim #13326), the $175,000,000 shelf (claim #12544), the $50,000,000 B. Riley facility at 9% (claims #12514, #12517), and a $125,000,000 at-the-market program (claim #12411). By early 2024 the B. Riley loan had been terminated early (claim #6291). It was replaced by the $20,000,000 AI Bridge unsecured loan (claim #6292), which was later settled in common stock (claim #6137). That shift from lender-funded debt to equity-settled obligations is the defining trajectory of the record.

The mid-2024 cluster reads more as sequential patching than as a planned stack. Within four months the company added the $50,000,000 Series E preferred (claim #6119), the $42,131,579 Yorkville prepaid advance with a $16,000,000 monthly conversion limit (claims #6078, #6111), and warrant-bearing CIM borrowings (claims #6010, #6030, #6031). It then added the $50,000,000 Series F convertible preferred (claim #5781), the $160,000,000 PIPE (claim #5782), the $250,000,000 standby equity purchase agreement (claim #5780), and the $62,500,000 Series E-1 (claim #5656). Several of these carried consideration beyond cash: consent shares to Yorkville (claim #6032), elimination of the prepaid advance conversion limitation (claim #5837), and a CIM waiver to allow further borrowing (claim #5760). Registration and redemption terms were repeatedly marked delayed (claims #6108, #5836, #5839).

Use of proceeds is only partly traceable. Early Ellendale deployment is concrete (claims #12622, #13326). The project-level financing tied to the 400 MW hyperscaler lease (claim #3237), however, still appears only as financing discussions four months later (claims #5867, #5892, #5907). The record does not show it closing, and that gap requires explanation.

A careful reader should track actual draws under the SEPA, conversion of the Series F into its stated 7,598,714 shares (claim #5834), further accordion-linked warrant issuance, and whether the HPC Ellendale financing is completed or no longer discussed. The record has limits. It ends in October 2024 with a truncated entry (claim #5545), it contains no insider transactions, and several figures appear in inconsistent units (claims #5621, #5550, #5593). Dilution totals therefore cannot be computed reliably from this data alone.

The record as provided ends in late October 2024 and contains no insider transactions, so this read covers financing only and cannot address insider selling. Within that window the pattern moves in two distinct phases. Through 2023 the company relied mainly on secured, asset-linked debt: the $20M Starion term loan for the Ellendale expansion (claim #13326), the $50M B. Riley facility at 9% (claims #12565, #12517), and a $125M at-the-market program under a $175M shelf (claims #12397, #12544). That mix reads as a conventional build-out plan. From February 2024 the cadence speeds up and the instruments change. The B. Riley loan was terminated early (claim #6291) and replaced by an unsecured $20M bridge (claim #6292). In the four months that followed came a $42.1M Yorkville prepaid advance (claim #6078), a $50M Series E preferred (claim #6119), a $25M Roth sales agreement (claim #6081), and the CIM facility with attached warrants (claims #5973, #6010). The September round added a $160M PIPE (claim #5782), $50M of Series F convertible preferred (claim #5781), a $250M SEPA (claim #5780), and a $62.5M Series E-1 (claim #5656). Stacking convertible notes, a standby equity line, two preferred series, warrants, and an ATM within six months looks more like layered patching around near-term needs than a single capital plan. The later arrival of a large institutional PIPE partly offsets that reading. Execution against announced capacity is uneven. Series E completed sales of $6.9M (claim #5657) compare with the $50M offering (claim #6119). The CIM facility grew by waiver to a stated $200M (claims #5760, #5660). Use of proceeds is harder to test. The Ellendale expansion (claim #13326) and the 400 MW hyperscaler lease (claim #3237) are the stated destinations, and HPC Ellendale financing was described as being in advanced discussions in August 2024 (claim #5907). The record through October does not show that project-level financing closed, so actual deployment cannot be matched to intent from this data. A careful reader should watch four things. First, the pace of Yorkville note conversions (claims #5624, #5550) and their effect on share count. Second, how much of the SEPA and ATM is drawn. Third, the Series F registration rights and PIPE items, which are marked delayed (claims #5836, #5548). Fourth, whether the Ellendale project debt is ever documented as closed, which would allow the equity-linked layers to shrink.

Applied Digital's disclosed financing record runs from February 2023 to October 2024 and contains no insider transactions, so any read of insider alignment is not possible from this dataset. The record also ends nearly two years before the present date. Several claims appear to carry figures in thousands or share counts rather than dollars, such as the Series F issuance of $53,191 in claim #5590 and the CIM warrants of 6,300,449 in claim #6029, so those values should not be compared directly.

The cadence accelerates sharply. In 2023 the company relied mainly on conventional instruments: the $20M Starion Bank term loan for Ellendale (claim #13326), the $50M B. Riley facility at 9% (claims #12565, #12517), and a $125M ATM (claim #12397). From May 2024 the capital stack becomes more layered and more equity-linked. It adds the $42.1M Yorkville prepaid advance with a $16M monthly conversion limit (claims #6078, #6111), the CIM facility with warrants attached (claims #6026, #6030), and Series F convertible preferred (claim #5781). By September 2024 the record shows the $160M PIPE (claim #5782) and the $250M SEPA (claim #5780) arriving together. Over that period the company moves from debt it would repay toward instruments that settle in common stock. Examples include the elimination of the prepaid advance conversion limitation (claim #5837) and $11.4M of debt converted to equity (claim #5593).

The mix reads as sequential patching more than a pre-planned structure. Consent-share issuances to Yorkville (claim #6032), a CIM waiver needed to borrow an additional $20M (claim #5760), and a conditional prepayment tied to a registration deadline (claim #6036) each suggest that new instruments had to be negotiated around existing ones. The September 2024 package may represent an attempt to consolidate that arrangement.

On deployment, the stated purposes are Ellendale expansion (claim #13326) and project-level financing for a 400 MW hyperscaler lease (claim #3237). HPC Ellendale financing was described three times as being in discussion (claims #5867, #5892, #5907), and the record contains no closed project financing. Several expected inflows were delayed before being confirmed, including the Garden City escrow release (claims #3242, #5625, then #5546) and PIPE proceeds (claims #5527, #5548).

A careful reader should watch four things. The first is how much of the SEPA and ATM capacity is actually drawn. The second is the path of the remaining YA note balance (claim #5550). The third is shareholder authorization of Series F conversion (claim #5512). The fourth is any accordion draw that triggers the contingent warrants (claim #6031). Most important, it would require explanation if project-level Ellendale financing is still not documented.

The record supplied here ends mid-entry in late October 2024 and contains no insider transactions, so this read covers roughly eighteen months of financing activity and cannot speak to insider selling. Within that window, the cadence accelerates markedly. The 2023 base was conventional: a $175,000,000 shelf (claim 12544), a $50,000,000 B. Riley term loan (claim 12565), and a $125,000,000 at-the-market agreement (claim 12397). From February 2024 the company layered on a new instrument almost every few weeks. These included an unsecured AI Bridge loan (claim 6292), $50,000,000 of Series E preferred (claim 6119), a $42,131,579 Yorkville prepaid advance convertible at a discount with a $16,000,000 monthly conversion cap (claims 6078, 6111), a warrant-bearing CIM secured note (claims 6010, 5660), $50,000,000 of convertible Series F preferred (claim 5781), a $250,000,000 standby equity purchase agreement (claim 5780), a $160,000,000 PIPE (claim 5782), and a $62,500,000 Series E-1 offering (claim 5656). Dilution is therefore arriving through several channels at once: registered sales, preferred conversions, note conversions (the $11,392,978 debt-to-equity figure, per claim 5593), warrants, and consent shares issued to Yorkville (claim 6032).

The mix reads more as sequential patching than a single designed structure. Each facility appears to have required amendments or waivers to accommodate the next one. Examples are the conversion-limit elimination on the prepaid advance (claim 5837), the CIM waiver permitting an additional $20,000,000 (claim 5760), and early repayment of the B. Riley loan (claim 6291). The September 2024 PIPE and SEPA mark a shift toward larger, cleaner equity, but they sit alongside rather than replace the higher-cost instruments.

On use of proceeds, the record states intent more than deployment. Project-level financing for a 400 MW hyperscaler lease (claim 3237) and advanced financing discussions for the HPC Ellendale facility (claims 5867, 5907) are repeatedly referenced, but no claim in this set confirms a closed project financing or ties specific proceeds to construction spend. The $25,000,000 Garden City escrow release moved from delayed (claim 3242) to stated (claim 5546), which is the clearest instance of a flagged item being resolved.

A careful reader should watch whether the Ellendale project financing actually closes. They should also watch whether the YA notes and Series F shares are fully retired or converted, how much of the SEPA and ATM capacity is drawn, and whether later filings consolidate these facilities or add further amendments.

Applied Digital completed a capital campaign from February through October 2024 that layered preferred equity, convertible debt, and secured lending in a pattern marked by accelerating delays in documentation and registration. The Series E preferred offering closed at $50M with $46M net proceeds (claim 6120, May), followed by Series E-1 ($62.5M, claim 5656, September) and Series F ($50M, claim 5658, September). A convertible promissory note to Yorkville Advisors ($42.1M, claim 6106, May) was conditioned on a July 8 registration deadline (claim 6108); that deadline elapsed, and by October the note remained in conversion discussions ($86M outstanding, claim 5550) rather than redemption closure. A CIM secured credit facility initiated with $15M in June (claim 6026), expanded via accordion features to $125M total by October (claim 5549), and issued warrants in sequential tranches ($6.3M first tranche, claim 6010, June). A Roth Capital ATM facility ($25M, claim 5986, announced May) and Series E redeemable offering ($2M) remained delayed through October. In September, the company announced a $250M Yorkville SEPA (claim 5780) and Goldman-backed PIPE ($160M, claim 5782) simultaneously; the PIPE appeared delayed in October (claim 5548). A $125M ATM facility (claim 5778, September) and CIM accordion availability ($110M, claim 5619) created layered equity buffers. The October $300M convertible senior note offering (claim 5481) disclosed specific use: $53M prepaid forward, $43M capped call, $31M share repurchase (claims 5477/5476/5478). Closed equity tranches (Series E/E-1/F) contrast with mounting delays in debt and derivative instruments. Warrant issuances to lenders (CIM $6.3M, AI Bridge $3M delayed, claim 5985) represent dilution independent of equity tranches, suggesting sweeteners deployed to secure debt certainty. The 18-month accumulation of delayed registrations, conversions, and accordion draws indicates reliance on contingent capital paths rather than linear execution. Stated proceeds deployment for the convertible—hedges plus buyback totaling ~$127M—implies management anticipated stock volatility and sought downside protection. Whether delayed tranches (Roth ATM, PIPE, escrow releases) ultimately closed or required amendment remains the critical open question, as their activation would indicate capital demand exceeded closed tranches and signal further structural revision.

Applied Digital's 2024 financing profile reveals a company managing liquidity through layered, increasingly aggressive capital markets access rather than executing a unified strategy. From February through May, the company accumulated conditional debt and preferred equity: an $20 million unsecured loan from AI Bridge (claim 6292, delayed), a $42.1 million convertible note from Yorkville (claim 6106), and prepaid advances of $40 million and $10 million from YA II (claims 6079, 6080, both delayed). These instruments carried structural features—monthly conversion caps on the Yorkville note (claim 6111), accordion facilities on the CIM term loan (claims 6028, 6009), and registration contingencies (claim 6108)—that suggest disciplined dilution management. Yet the May-June period accumulated five separate delayed filings for the Roth ATM, YA promissory notes, and Series E redeemable preferred stock, indicating registration or market friction.

By September, the company pivoted to aggressive capital aggregation: a $250 million Standby Equity Purchase Agreement with Yorkville Asset Management (claim 5784), a $160 million Goldman-backed PIPE (claim 5782), $50 million Series F preferred (claim 5757), and $125 million ATM capacity (claim 5778)—all announced within two days. October crystallized this trajectory with a $300 million convertible senior notes offering (claim 5481) plus $45 million greenshoe, generating $362 million in net proceeds (claim 5475). Yet the capital redeployment was defensive: $31 million to share repurchase (claim 5478), $43 million to capped calls (claim 5476), and $53 million to a prepaid forward (claim 5477)—financial engineering rather than organic growth deployment.

Dilution acceleration is severe. Applied Digital expanded through at least five preferred series in nine months, with the CIM facility eventually reaching $200 million total capacity (claim 5660) and outstanding balances of $125 million by October (claim 5549). Warrant packages exceeded $15 million (claims 6010, 6029, 6031). Stated uses—the $46 million Series E net proceeds and the $150 million "secured funding" in August (claim 3241)—lack clear asset deployment linkage. Garden City escrow release ($25 million, claim 5779) and project-level financing for 400 MW capacity (claim 3237) appear mentioned but not tracked to actual capex.

Sophisticated readers should monitor whether YA Fund exercises its $250 million commitment at unfavorable prices, the CIM accordion draw pace and remaining available tranches, and further delayed registration statements that would force equity acceleration. The October hedging trades suggest management concern about dilution impact on share price, signaling that current capital velocity may be unsustainable.

Applied Digital's 2024 capital-raising reveals a company cycling from scrappy to oversubscribed, with insufficient evidence that the instrument mix reflects coherent strategy rather than sequential opportunism. Between February and October 2024, management deployed preferred equity (Series E, E-1, F), convertible debt (YA II promissory notes, $300M senior converts), secured credit facilities (CIM $125-200M, per claim 5660), an equity commitment (YA SEPA $250M, per claim 5659), a PIPE ($160M, per claim 5782), and an ATM facility ($125M, per claim 5778)—a portfolio suggesting each new program addressed a financing bottleneck rather than an orchestrated capital plan.

Dilution trajectory compounds this reading. Early 2024 showed restraint: Series E at $50M nominal (claim 6119) yielding $46M net (claim 6120) with $9 preferred dividends (claim 6136). By May, management issued convertible promissory notes to Yorkville Advisors at $42.1M (claim 6106) with monthly conversion caps of $16M (claim 6111), then immediately sweetened CIM credit facility terms with warrants worth $6.3M (claim 6010) and consent shares of 100,000 each to Yorkville and YA II (claims 6032, 5988). These incremental equity concessions to debt and preferred holders require explanation. The second half accelerated: Series F at $50M preferred (claim 5657), a $250M SEPA with YA Fund (claim 5659), a $160M Goldman PIPE (claim 5782), and a $125M ATM (claim 5778) all announced by mid-September before the $300M convertible senior notes (claim 5481) followed in late October.

Use of proceeds tracking is opaque. Stated net proceeds from October convertibles totaled $362M (claim 5475), yet management allocated $43M to capped call transactions, $53M to prepaid forwards, and $31M to share repurchase (claims 5476-5478). This hedging and buyback deployment—rather than facility capex—suggests expected near-term stock pressure and management's choice to manage dilution through derivatives. Stated uses for Series F ($50M) and YA II convertible proceeds remain absent or generic ("financing discussions for HPC Ellendale," claims 5892, 5907). The Garden City escrow release ($25M) appears delayed.

Sophisticated readers should monitor: whether conversion cascades emerge as Series F converts (dividend rate $8, per claim 5835) and the $300M notes encounter capped upside via call spreads. CIM credit facility ($125-200M, per claim 5660) underpins the entire capital stack; any covenant stress forces immediate equity raises. Most critically, SEPA ($250M) and ATM ($125M) capacity totaled $375M yet actual draws remain sparse, suggesting capital stacking on standby for capex volatility the company misforecasted.

Applied Digital's capital sequence through October 2024 reveals a company that started with disciplined layering but pivoted toward accelerated, structurally hedged funding as growth capex requirements became apparent. The arc moves from preferred equity (Series E $46M net proceeds, per claim 6120 in May) through convertible debt (YA II prepaid advance $42.1M, per claim 6078) and secured lending (CIM $15M initial and $110M available, per claims 6026-6027 in June), finally climaxing in a $345M convertible senior notes offering in October (per claims 5481-5482, base plus greenshoe). The cadence accelerated sharply: May-June saw methodical instrument layering; August-September compressed multiple tranches into weeks; October then deployed $345M in a single capital event. However, the gap between announced and closed is material. Major facilities remain delayed: the S-1 registration effective date (claim 6108, May 23) never materialized in the data, Yorkville's promissory notes (claims 6079-6080) stay pending, the $25M Roth sales agreement (claim 6081) stalled, and the CIM term loan facility (claim 5753) appears half-drawn. This suggests not opportunism but constraint—the company announced more than it could close, likely pending regulatory or market clearances.

The mix shows neither coherent strategy nor pure opportunism, but rather forced adaptation. Preferred equity dominated early raises, but convertible debt and secured lending increasingly dominated capital structure by mid-year, indicating either equity dilution concerns or investor preference for downside protection. The October convertible offering itself deserves scrutiny: of $362M net proceeds (claim 5475), the company explicitly deployed $53M to a prepaid forward transaction, $43M to capped calls, and $31M to share repurchase (claims 5476-5478). This tri-part derivative structure caps the upside, hedges conversion risk, and offsets dilution—defensive mechanics, not growth mechanics. The $250M SEPA with Yorkville (repeatedly claimed, e.g., claim 5659) and $125M ATM facility (claim 5778) remain largely unstated as to utilization.

Use of proceeds tracking is opaque. The Ellendale HPC facility financing and 400 MW hyperscaler lease financing (claims 3237, 5892) are mentioned but deployment is not reconciled against the capital raised. The October convertibles' $235M residual (after derivatives/buyback) lacks stated destination.

Key watch items: S-1 effectiveness (gates future public market access), Garden City escrow release execution (lingering since August), SEPA drawdown activity, and whether cumulative convertible exercises trigger the prepaid forward/capped call mechanics. The delayed items outnumber the closed items by capital value, suggesting a second-act capital cycle may depend on regulatory or operational clearing events that have not yet surfaced in this data.

Applied Digital has executed an exceptionally rapid and complex capital raise from February through October 2024, layering multiple instruments that suggest both aggressive infrastructure ambitions and difficulty sustaining capital access through conventional channels. The company raised $50 million in Series E preferred equity (claim 6119), followed by $50 million in Series F preferred (claim 5781), then a $300 million convertible notes offering with $45 million greenshoe (claims 5481, 5482). Concurrently, the company accessed a $15 million initial CIM term loan with $110 million accordion capacity (claims 5974, 5975), a $250 million standby equity agreement with YA Fund (claim 5784), a $160 million PIPE placement (claim 5782), and a $125 million at-the-market facility (claim 5778). This is not coherent strategy; it is opportunistic stacking of every available capital market tool. The instrument mix—equity, convertible debt, secured lending, warrants, and contingent equity—reveals uncertainty about sustained single-source access. The repeated appearance of Yorkville and YA Fund entities, typically accessed by capital-constrained firms, signals that conventional venture markets were either closed at acceptable dilution terms or exhausted by September.

Deployment patterns partially explain the velocity. Stated use of proceeds for the October convertible offering ($362 million net, claim 5475) reveals partial capital diversion to shareholder protection rather than operations: $43 million for capped call transactions (claim 5476), $53 million for a prepaid forward (claim 5477), and $31 million for share repurchases (claim 5478). These hedging mechanisms, executed amid aggressive equity raises, signal management concern about dilution effects. Earlier disclosures reference a 400 megawatt hyperscaler lease (claim 3237) and HPC Ellendale facility financing (claims 5892, 5907), anchoring raises to infrastructure deployment. Yet the nine-month velocity suggests either unexpectedly high infrastructure capital intensity or a pre-public capital build executed ahead of market windows closing.

Sophisticated readers should track quarterly cash runway and burn rate, warrant conversion patterns from the substantial lender-issued tranches (CIM, AI Bridge), and whether HPC deployment remains on timeline. The multiple "delayed" status designations suggest regulatory friction. Most important: whether the company can stabilize funding cadence post-October, or whether warrant conversions and SEPA draws signal ongoing liquidity pressure beneath the capital raise headlines.

Applied Digital has assembled a financing stack that prioritizes capital velocity over capital discipline. From February through October 2024, the company executed at least seven distinct mechanisms concurrently: Series E and E-1 preferred offerings ($62.5M gross, per claim 5656), convertible promissory notes ($42.1M, claim 6106), secured lending from CIM ($125M stated, though delayed per claim 5753), a Standby Equity Purchase Agreement ($250M, claim 5784), private placements ($160M, claim 5783), an at-the-market offering ($125M, claim 5778), and ultimately a $300M convertible senior note (claim 5481). This layered approach suggests opportunistic capital-raising rather than a planned capital structure. The convertible promissory note included a monthly conversion cap of $16M (claim 6111), indicating initial caution on dilution, but that discipline eroded by October when the company issued $300M in convertibles with hedges—a prepaid forward transaction ($53M, claim 5477) and capped calls ($43M, claim 5476)—signaling management expects material stock appreciation.

Use of proceeds remains opaque. The company disclosed financing discussions for HPC Ellendale (claim 5892) and project-level facilities, but actual deployment pathways are not explicit. Of the $362M net proceeds from the October convertible offering (claim 5475), only approximately $127M is accounted for: $53M in prepaid forwards, $43M in capped calls, and $31M in share repurchases (claims 5477, 5476, 5478). The remainder—roughly $235M—is presumably reserved for capacity build-out, but this is not detailed. The $25M escrow release from Garden City (claim 5779) may supplement operational liquidity, though facility capital expenditures and hyperscaler lease deployment against the financing calendar remain unstated.

Dilution has accelerated. The 10M share equity compensation plan reservation (claim 5575) adds another vector. The YA Fund SEPA ($250M, claim 5780) and convertible stack position the company to access equity markets aggressively should debt covenants tighten. Watch whether proceeds flow into tangible capacity deployment versus accumulation. If facility utilization lags the financing calendar or stated HPC deployment is revised, the company's already-complex capital structure will show strain.

Applied Digital raised approximately $1 billion in gross capital commitments between February and October 2024, with marked acceleration from May onward. The cadence was relentless: Series E ($50M net, May), convertible promissory notes ($42.1M, May), YA II tranches (delayed through summer), CIM facility ($200M authorized), Series E-1 ($62.5M, September), $160M PIPE placement, $250M Standby Equity Purchase Agreement, and finally a $300M convertible senior notes offering (October, per claim 5481). Each window was seized independently rather than in sequence, suggesting capital-constrained opportunism rather than deliberate capital planning. The stated uses of convertible proceeds reveal the constraint: per claim 5477 and 5476, $53M was deployed to prepaid forward transactions and $43M to capped call hedges—essentially derivative positions to dampen dilution from conversion—while claim 5478 directed $31M to share repurchases. Stated capex needs (claim 3237: project-level financing for 400 MW hyperscaler lease; claim 5867: HPC Ellendale facility) were separately financed, leaving the main equity rounds orphaned of disclosed purpose. The capital mix was equally incoherent: venture preferred (Series E, E-1, F), unsecured convertible notes from Yorkville/YA II, secured CIM term lending with accordion tranches, warrant-sweetened facilities, and a SEPA providing $250M in backstop equity. No single instrument was sized or structured to meet stated needs; each appears to have been added as an available option. The dilution trajectory is steep and bifurcated: base equity dilution from preferred stock and the 10M equity compensation plan (claim 5575), plus conversion and warrant dilution embedded in the convertible and CIM facilities. The largest delayed items—Roth ATM, YA II notes, PIPE, SEPA, CIM accordion waiver (claim 5760)—suggest either disclosure lags or unmet conditions, requiring explanation. A reader should monitor: (1) whether the $300M convertible converts or remains debt, and at what stock price; (2) activation rates on the $250M SEPA and CIM accordion ($75M available, per claim 5620); (3) capex deployment and whether it justifies the capital raised; (4) whether the prepaid forwards and capped calls genuinely reduce dilution or merely defer recognition of it; (5) the composition and exercise pace of warrant packages issued to lenders, which remain largely unquantified in aggregate.

Applied Digital raised approximately $1 billion across nine months through cascading instruments—a cadence that accelerated sharply from May through October. The company began conservatively with a $50 million Series E offering in May (claim 6119), but by September had authorized a $250 million standby equity purchase agreement, $160 million Goldman-backed PIPE, and $125 million ATM facility (claims 5780, 5782, 5778). October brought a $62.5 million Series E-1 tranche (claim 5656) and a $300 million convertible senior notes offering with $45 million greenshoe (claims 5481, 5482). This sequence reflects opportunistic patching rather than coherent strategy: each instrument was layered to address a specific funding gap rather than as part of a planned capital structure.

The debt component shows visible strain. CIM provided a $200 million secured facility by September (claim 5660) with $15 million drawn initially and $110 million plus $75 million accordion capacity available (claims 6007–6009). Yorkville promissory notes totaling $50 million (claims 6079, 6080) required warrant issuances and common share grants as lender consent fees (claims 6029, 6031, 6032)—a signal that unsecured capital required equity sweeteners. B. Riley facilities were early-terminated and refinanced into CIM, indicating a rolling-short-to-longer-dated pattern. Against this, announced use of proceeds did not track actual deployment: asset sales in April ($160 million, claim 3240) and Garden City escrow releases ($25 million, claims 5779, 5742) addressed near-term liquidity, but project-specific financing for a 400 megawatt hyperscaler lease (claim 3237) and HPC Ellendale facility discussions (claims 5867, 5892) lacked corresponding capital deployment disclosures.

Registration delays created timing friction. The Form S-1 effectiveness requirement flagged as delayed in May (claim 6108) and repeated delays on the Roth ATM facility (claims 6081, 5986, 5754) suggest SEC or disclosure readiness issues. By late October, management deployed $53 million of convertible proceeds into a prepaid forward (claim 5477) and $43 million into capped calls (claim 5476), classic defensive hedges signaling concern about conversion dilution. A simultaneous share repurchase authorization (claim 5483) appears contradictory given maximum dilution from SEPA, PIPE, ATM, and convertibles.

Sophisticated readers should focus on whether debt conversions accelerate—$11.4 million was noted in October (claim 5593)—since operating cash flow may not cover capex over the next two to three quarters. The draw pace on the $200 million CIM facility and any accordion activation would signal runway constraints. Whether the prepaid forward and capped calls prove sufficient to cap dilution or force conversion economics at lower prices merits continuous monitoring.

APLD's 2024 capital evolution traces from methodical preferred-stock tranches to rapid instrument layering, suggesting management's confidence in single capital sources eroded through the year. The company opened with Series E preferred at 9 percent dividend, $50 million base issuance (claim 6119, claim 6136), and a $42.1 million convertible note restricted to $16 million monthly conversion (claim 6106, claim 6111)—structures suggesting deliberate dilution control. By late August, this discipline collapsed. Within weeks, APLD simultaneously committed to a $150 million facility (claim 3241), Series F preferred at 8 percent (claim 5833), and a $200 million CIM secured credit facility with accordion rights (claim 5660). September's announcements accelerated sharply: $62.5 million Series E-1 preferred (claim 5656), $250 million SEPA backstop with Yorkville Advisors (claim 5780), $125 million ATM authorization (claim 5778), and a $160 million PIPE (claim 5782), all within two weeks. The mix reflects opportunistic patching. The preferred tranches remain only partially deployed; conversion mechanics carry delayed effective dates (claims 6107, 6108); and the October $300 million convertible senior notes issuance (claim 5481) arrived before several earlier facilities had closed materially. This suggests a capital reset rather than incremental extension. Management then allocated $31 million of convertible proceeds to share repurchase and $43 million to capped-call hedges (claims 5478, 5476), moves inconsistent with organic valuation confidence. Announced use of proceeds—hyperscaler deployments and facility operations (claim 3237, 3241)—remains unverified against actual capex timing. Capital appears raised opportunistically against anticipated needs rather than in strict response to committed deployments. Forward signals center on draw patterns: monitor CIM accordion utilization, SEPA and ATM execution rates, and October convertible conversion timing to clarify whether operational scaling or capital structure management drove the year's raises.

Applied Digital's capital trajectory through 2024 reflects a progression from opportunistic patching to structured financial engineering. The first half exhibits a kitchen-sink approach: unsecured loans from AI Bridge ($20M, claim 6292), Series E preferred equity ($50M raised, per claim 6119), convertible promissory notes ($42.1M, claim 6106), and warrant-sweetened debt. Multiple instruments show delayed execution—the YA II promissory notes, Roth sales agreement, and Series E offerings all carry delayed status marks—suggesting execution friction in raising capital efficiently. By mid-year, coherence emerges around a CIM credit facility ($125M total commitments across initial $15M, subsequent $110M tranches, and a $75M accordion, per claims 5975, 6008, 6009) paired with a $250M Standby Equity Purchase Agreement with YA Fund (claim 5784), forming a structured backstop architecture. This phase couples asset-level financing (400 MW hyperscaler lease, claim 3237) with market-accessible credit facilities rather than relying on scattered preferred issuances. The October convertible offering marks an inflection. A $300M convertible note due 2030 (claim 5481) with $45M greenshoe option (claim 5482) yields $362M net proceeds (claim 5475). Critically, deployment diverges from typical infrastructure capex: $43M funds capped-call transactions, $53M a prepaid forward, and $31M a share repurchase (claims 5476-5478). This is financial structure optimization, not growth funding. Dilution has compounded substantially—Series E, E-1, and F preferred issuances crossed multiple tranches; CIM warrants and Yorkville consent shares added equity kickers; convertible debt will dilute upon likely conversion. Yet the October convertible's hedging deployment suggests management is engineering the dilution path deliberately rather than accepting mechanical conversion. A sophisticated reader should monitor (1) whether SEPA's $250M remains unused or draws accelerate (backstop facilities reveal confidence gaps), (2) accordion draw rates on the CIM facility—substantial capacity remains, signaling either conservative borrowing or headroom for stressed scenarios, (3) warrant exercise potential ($30M per claim 5725) and whether equity holders absorb or trigger it, and (4) Series F redemption timing (claim 5839 delayed), which will clarify whether preferred investors hold or force equity settlement. The capital structure has rotated from short-term bridging toward longer-dated convertible funding, but the scale of standby and accordion facilities suggests the company expects volatility in near-term cash generation.

Applied Digital pursued a capital structure strategy in 2024 that shifted materially from infrastructure investment to financial optionality. Through February-May, the company announced a $20 million AI Bridge unsecured loan (claim 6292, marked delayed as of the filing date), a $50 million Series E preferred offering (claim 6119, $9 dividend rate), and a $42.1 million convertible promissory note (claim 6106). These early financings tracked a coherent narrative around project-level funding for a 400 megawatt hyperscaler lease (claim 3237) and secured financing discussions for the HPC Ellendale facility (claim 5867, announced August 2024).

By June, however, the capital mix fragmented. APLD layered in a $200 million CIM credit facility (claim 5660, eventually consolidated into a $125 million term loan via claim 5753, marked delayed), a $250 million Standby Equity Purchase Agreement with Yorkville Advisors (claim 5780), a $125 million At-the-Market facility (claim 5778), and multiple warrant issuances tied to debt tranches. The company concurrently repaid earlier debt via common stock issuance to AI Bridge (claim 6137, $8.4 million of the original $20 million) and received consent-related equity grants to Yorkville (claim 6032, 100,000 common shares). By September, APLD added a $160 million Goldman-backed PIPE (claim 5782) and Series F Preferred ($50 million, 8% dividend rate per claim 5835).

The October-November period crystallized a pivot toward financial engineering. A $300 million convertible senior note offering (claim 5481, plus $45 million greenshoe per claim 5482) closed in late October. The disclosed use of proceeds reveals the shift: $53 million for a prepaid forward transaction (claim 5477), $43 million for capped call transactions (claim 5476), and $31 million for share repurchase (claim 5478). These hedge mechanisms consume $127 million of the $362 million net proceeds (claim 5475), leaving roughly $235 million for operations or debt service.

The original infrastructure narrative—Ellendale financing discussions, project-level deployment—appears no longer discussed in later filings. The company maintained multiple simultaneous funding commitments (SEPA still unfunded, ATM undrawn, CIM accordion features available) while converting Yorkville promissory notes and managing warrant dilution. A sophisticated reader should track: whether announced capex deployment materializes against stated data-center timelines, whether the prepaid-forward and capped-call hedges remain optimal if equity prices compress, and whether the layered capital structure reflects strategic optionality or constrained access to cheaper alternatives.

Applied Digital's 2024 capital-raising architecture progresses from exploratory financing to a deliberate pre-IPO structure. The company mobilized roughly $1.5 billion in gross commitments across thirteen distinct facilities between February and October, but the sequencing reveals strategic intent rather than opportunistic patching.

The dilution trajectory accelerates sharply. Early-stage rounds—Series E's $50 million net proceeds (claim 6120, May), AI Bridge unsecured $20 million (claim 6292, February), and a $42.1 million convertible promissory note (claim 6106, May)—represent measured venture-style deployment. By September, the pace doubles: Series F $50 million (claim 5781), Series E-1 $62.5 million (claim 5656), a $250 million Standby Equity Purchase Agreement with YA Fund (claim 5780), a Goldman-backed PIPE at $160 million (claim 5782), and a $125 million At-the-Market facility (claim 5778). October's $300 million convertible senior notes with a $45 million greenshoe option (claims 5481, 5482) caps the sequence.

The capital mix is coherent. Preferred equity and secured debt—CIM's $200 million accordion facility (claims 5660, 5974)—finance core operations and hyperscaler leases; project-level financing addresses specific facilities. The convertible notes include explicit hedging: $43 million allocated to capped call transactions and $53 million to a prepaid forward (claims 5476, 5477), mechanics typical of IPO preparation to offset shareholder dilution. This is not patching; it is scaffolding.

Announced use of proceeds tracks observable deployment. The CIM facility ($15 million initial, $110 million subsequent, $75 million accordion, claims 6026, 6027, 6028) correlates with advanced financing discussions for the HPC Ellendale facility (claim 5907, August). Garden City asset sale proceeds of $25 million (claim 5779) recirculate without pause. The equity compensation plan's $10 million reservation (claim 5575, October) signals imminent IPO mechanics.

A sophisticated reader should monitor the draw pace on the $250 million SEPA—its post-IPO utilization reveals whether Applied Digital trusts its near-term cash generation. The CIM accordion's deployment rate ($15 million drawn of $200 million available as of mid-October, claim 5549) warrants tracking; accelerating draws might signal faster capex execution or liquidity pressure. Preferred conversion dilution at IPO will be material to common holders. Finally, the AI Bridge and Yorkville facilities—with delayed prepaid advance maturity (claim 5838)—may require refinancing if convertible proceeds deploy to fixed assets rather than debt reduction.

Applied Digital's 2024 capital trajectory reveals a rapid escalation from traditional preferred equity into a layered, increasingly complex debt and hybrid stack. The company began with Series E preferred ($50M, claim 6119) and a $42.1M convertible (claim 6106) in May, then accelerated through a succession of facilities: CIM's secured credit arrangement with $15M initial borrowing and $110M subsequent availability (claims 5973, 5974), YA Fund's prepaid advances and promissory notes, and a $250M Standby Equity Purchase Agreement (claim 5780). By September, this had compounded into simultaneous exposure across CIM ($125M outstanding per claim 5549), YA convertible notes ($86-250M outstanding across various claims), and newly-layered facilities like a $125M ATM (claim 5778) and Goldman-backed PIPE ($160M, claim 5782). The October convertible offering—$300M senior notes due 2030 with a $45M greenshoe (claim 5481)—capped an eight-month period of opportunistic capital stacking rather than strategic sequencing. The company announced use of $362M net proceeds (claim 5475) as: share repurchase ($31M, claim 5478), capped call transactions ($43M, claim 5476), and prepaid forward transactions ($53M, claim 5477). This mix signals defensive mechanics overlaid on primary capital raising, not primary deployment. The original stated thesis—project-level financing for 400 MW hyperscaler infrastructure (claim 3237) and HPC Ellendale facility funding (claims 5892, 5907)—has become disconnected from the actual use of proceeds pattern, which emphasizes shareholder dilution mitigation rather than capex. Numerous claims carry delayed status, including CIM waivers (claim 5760), PIPE closings (claim 5548), and escrow releases, suggesting execution friction. A sophisticated reader should monitor: whether the nested equity layers (SEPA, ATM, convertible with capped calls) activate in sequence or redundantly, signaling capital market tightness; YA Note conversion velocity, which will reveal actual equity demand; and whether facility amendments and waivers accelerate, indicating strain on covenant capacity. The capped call structure deserves particular attention, as share price movements that trigger deliveries will materially affect ownership concentration.

Applied Digital executed a capital raise strategy markedly different in early and late 2024, suggesting a pivot from private-market stacking to public-market hedging. Through May-June, the company layered convertible debt—a $42.1 million promissory note (claim 6106) coupled with a $50 million Series F preferred and a $15 million CIM facility with $75 million accordion availability (claims 5757, 6007, 6028)—a structure typical of pre-IPO companies managing dilution while maintaining flexibility. This phase prioritized optionality: convertibles with conversion caps, warrant packages tied to facility tranches, and preferred stock with specified dividend rates ($9 for Series E per claim 6136, $8 for Series F per claim 5835) signaled a company managing multiple stakeholder classes and maintaining upside participation mechanisms.

By September-October, the mix shifted to large-scale public instruments: a $250 million Standby Equity Purchase Agreement (claim 5659), $125 million At-the-Market capacity (claim 5778), and a $160 million PIPE (claim 5729). Most notably, the company closed a $300 million convertible senior notes offering with a $45 million greenshoe (claims 5481, 5482), netting approximately $362 million (claim 5475). This late-stage financing diverges sharply from the spring instruments: convertibles provide lower immediate dilution but create maturity obligations, while the announced use of proceeds—$31 million for share repurchase, $53 million for prepaid forwards, and $43 million for capped calls (claims 5478, 5477, 5476)—prioritizes financial engineering over operational capital deployment. The repeated discussion of "financing discussions for HPC Ellendale Facility" (claims 5867, 5892, 5907) lacks concrete capital deployment, suggesting either negotiations remain incomplete or proceeds tracked elsewhere.

Delayed items warrant attention: the Garden City sale escrow $25 million repeatedly requires clarification (claims 3242, 5546, 5625), and multiple CIM tranches and Yorkville amendments (claims 5594, 5661) indicate ongoing covenant management. The trajectory reveals a company that front-loaded debt and preferred instruments, then moved to equity cushioning once public markets became available. Whether the shift from capex-oriented messaging to financial engineering reflects capital intensity proving lower than expected or strategic reorientation requires comparison to actual facility completions and power utilization rates.

APLD's financing sequence from February through October 2024 shows escalating capital intensity with a shift from improvised debt stacking toward institutional equity commitments. Through mid-year, the company assembled a patchwork of secured lending (CIM promissory note with $15M initial draw and $75M accordion feature, per claims 6007 and 6028) alongside Series E preferred equity at $50M net proceeds (claim 6120) and a $42.1M convertible note via Yorkville (claim 6106) with monthly conversion caps of $16M (claim 6111). This phase resembled opportunistic capital sourcing—each tranche paired with warrants or common share issuances for lender consent, suggesting constrained negotiating position. The cadence was rapid and uncoordinated: five distinct capital instruments layered within ninety days.

The pattern shifted decisively in September. APLD committed to a $250M Standby Equity Purchase Agreement with YA Fund (claim 5659), a $160M PIPE placement (claim 5729), and a $125M at-the-market facility (claim 5726)—collectively $535M in new capacity within two weeks. This acceleration suggests either imminent IPO pressures or material capex requirements insufficiently covered by prior facilities. A $25M asset sale from Garden City escrow (claim 5779) supplemented organic cash.

By October, the company pivoted to public-market instruments. It issued $300M in convertible notes due 2030 with a $45M greenshoe option (claims 5481 and 5482), raising an estimated $362M net (claim 5475). Critically, APLD allocated $31M to share repurchases (claim 5478), $53M to prepaid forward transactions (claim 5477), and $43M to capped call hedges (claim 5476). This financial engineering—approximately 34 percent of net proceeds—prioritized balance-sheet optimization and optionality over capex velocity, diverging from the infrastructure-focused deployment expected in early 2024 announcements.

The mix across instruments shows coherence in retrospect, not foresight. Early tranches were reactive, pairing secured lending with dilutive equity to secure lender comfort. Mid-year institutional commitments signaled a maturing capital access strategy. The Q4 convertible issuance, married to equity hedges and buybacks, suggests transition from raise-as-needed to disciplined capital structure management.

Watch whether APLD deploys the $500M-plus in undrawn facilities (CIM accordion, SEPA, at-market) and at what utilization rate. High burn would indicate original capex estimates were understated. Monitor convertible redemption or conversion timing; early conversion would signal equity appreciation but also renewed dilution.

Applied Digital's financing trajectory from February through October 2024 compressed a venture-to-public-market progression into nine months, revealing a deliberate shift from venture structures toward public-scale access. Early months show execution delays: the $20 million AI Bridge loan (claim 6292) and multiple Yorkville tranches remained unfunded alongside contingencies tied to registration effectiveness (claim 6108), signaling initial capital access friction. By May, the company pivoted to a preferred-equity model: Series E at $50 million with a 9% dividend (claims 6119–6120) coupled immediately to a $42.1 million convertible promissory note (claim 6106) with monthly conversion gates of $16 million. This structure front-loaded proceeds while deferring common-stock dilution; the prepaid advance to YA II PN (claim 6078) mechanically gatekeeled conversion rights.

Summer months layered in secured debt optionality. The CIM facility offered $15 million initial draw plus $110 million accordion availability (claims 6007, 5974) with warrant issuances totaling roughly $15.8 million across tranches. Series F preferred stock ($50 million, claim 5757) replicated the preferred-plus-warrant model at an 8% dividend. By September, the company accessed public-market scale: a $250 million Standby Equity Purchase Agreement (claim 5756), a $160 million PIPE (claim 5782), and a $125 million at-the-market facility (claim 5754). October's anchor was a $375 million convertible senior notes offering due 2030 with a $45 million greenshoe (claims 5481–5482), netting $362 million in proceeds.

Announced use of proceeds tracked away from capex: the $362 million funded share repurchases ($31 million), capped call transactions ($43 million), and prepaid forwards ($53 million) rather than the 400 MW hyperscaler or HPC Ellendale projects mentioned in mid-2024 claims (3237, 5907). This financial engineering—particularly capped calls layered atop convertibles—requires explanation: it signals either prior capex-funding sufficiency or active hedging against post-IPO dilution scenarios. The overall cadence appears intentional, migrating from venture instruments through growth-stage optionality to public access. What demands scrutiny is whether the $250 million SEPA and $110 million CIM accordion actually drew—many claims remain marked delayed—and whether capital deployment discipline holds if public equity pricing deteriorates.

Applied Digital's financing architecture from February to October 2024 reflects escalating capital intensity married to deferred regulatory clarity. The company began modestly with a Series E round ($50M net proceeds, claim 6120) and a single convertible note ($42.1M, claim 6106) in May, both backed by warrant dilution tied to loan tranches. By June, the capital stack had multiplied: a $15M initial CIM facility with a $110M accordion feature and warrants valued at $6.3M (claim 6010), prepaid advances from YA II totaling $82.1M (claims 6078–6080, though maturity dates marked delayed), and a Yorkville $25M ATM agreement (claim 5986, delayed). Each layer introduced conversion rights or warrant triggers, creating nested dilution vectors.

The mix resembles opportunistic layering rather than coherent strategy. The company drew from equity (Series E-1 $62.5M and Series F $50M), debt ($200M CIM facility, claim 5660), preferred convertibles (Series E-1, E, F with 8–9% dividends), unsecured convertible notes ($42.1M), and standby equity lines ($250M SEPA with YA Fund, claim 5784). Notably, the AI Bridge $20M unsecured loan from February never materialized (marked delayed); the company instead issued common stock to prepay $8.4M of it in May (claim 6137), suggesting either refinancing necessity or changed terms.

Announced use of proceeds shifted from asset deployment to financial engineering. Early claims on Series E and Series F offerings lack specific use detail. By October, the $300M convertible senior notes offering (claim 5481, plus $45M greenshoe) explicitly designated $31M for share repurchase (claim 5478), $53M for prepaid forward transactions (claim 5477), and $43M for capped call transactions (claim 5476). This mix—capital raise to buyback to derivative hedges—suggests managing stock dilution through hedge mechanics rather than reducing absolute share count. The $25M Garden City escrow release (claim 5779) funded operations or debt paydown, undisclosed.

A sophisticated reader should monitor registration statement effectiveness; multiple tranches contingent on S-1 efficacy remain delayed (claims 6108, 6036). Watch YA Notes conversion pace ($86M outstanding as of 10/17, claim 5550) and whether prepaid advances convert or mature. The convertible notes' capped call structure warrants scrutiny: if strike prices are below trading levels, they accelerate dilution rather than cap it. CIM accordion drawdown velocity and Series F redemption mechanics will determine whether the company can service debt while managing preferred dividend obligations. The announced share repurchase ($31M authorization, claim 5483) should be tracked against actual execution to assess net dilution.

Applied Digital's capital stack reveals a company that moved from opportunistic patching toward strategic sequencing, though key commitments remain unreconciled. The trajectory from February through October 2024 divides cleanly: early reliance on mixed instruments (the $20M unsecured AI Bridge loan marked delayed, per claim 6292; B. Riley term loan termination per claim 6291) shifted to layered preferred equity—Series E at $50M stated (claim 6119), Series E-1 at $62.5M stated (claim 5656), Series F at $50M (claim 5781)—paralleled by convertible debt buildout, beginning with a $42M promissory note (claim 6106) and escalating to $300M in convertible notes due 2030 (claim 5481). This suggests the company recognized dilutive equity alone would not fund infrastructure growth at required velocity.

The June-September period reads opportunistic rather than coordinated. Multiple facilities were announced then marked delayed: a $250M Standby Equity Purchase Agreement (delayed Sep 10, claim 5756; stated Oct 3, claim 5622), a $125M at-the-market facility (delayed Sep 10, claim 5754; stated Oct 3, claim 5623), and a $25M Roth Capital sales agreement (delayed Jun 6, claim 6081). This suggests phased drawdown—the CIM accordion feature available at $75M (claims 5975, 6028)—or contingent commitments executable only after prior raises closed. The Garden City escrow release timing is inconsistent: stated at $25M released Aug 28 and Sep 10 (claims 3242, 5779), then marked delayed Oct 3 (claim 5625). This requires reconciliation.

Announced use of proceeds deviates from capex narratives. The $362M net from the October convertible (claim 5475) breaks into financial hedges: $53M for a prepaid forward and $43M for capped calls (claims 5477, 5476), plus only $31M for share repurchase (claim 5478). These are dilution offsets, not capacity additions. The earlier "project-level financing for 400 MW hyperscaler lease" (claim 3237) and "$150M secured funding" (claim 3241) lack deployment detail.

Watch three items. First, CIM facility utilization: total availability suggests $200M across tranches (claim 5660), but drawn amounts require accounting. Second, Series F conversion authorization remained marked delayed as of Oct 22 (claim 5512). Third, the Yorkville relationship merits scrutiny: consent shares (claim 6032), registration covenants (claim 5836, delayed), and conditional prepayment triggers (claim 6036) indicate leverage held by a mezzanine creditor with optionality into the capital structure.

Applied Digital has employed an accelerating, layered financing strategy that shifted from equity concentration to convertible debt dominance, though the coherence of that progression warrants scrutiny. Through May 2024, the company raised primarily via preferred equity—Series E for $50M net proceeds (claim 6120)—while staging a convertible promissory note for $42.1M (claim 6106). The June 2024 period marked a structural break: three independent debt facilities launched simultaneously—a $50M+$10M promissory note pair with Yorkville (claims 6079–6080, both delayed), a $15M initial commitment from CIM with $110M subsequent availability and $75M accordion capacity (claims 6026–6028, stated), and a $25M sales agreement with Roth Capital Partners (claim 6081, delayed). This simultaneous staging of multiple lenders at similar terms suggests less a unified capital strategy than opportunistic commitments negotiated under time pressure. The presence of delayed registration statements (claims 6108, 6107) and monthly conversion caps ($16M, claim 6111) further indicates the Yorkville and convertible facilities carried compliance friction that remains unresolved. By August–September, Applied Digital raised $50M in Series F convertible preferred and committed a $250M Standby Equity Purchase Agreement with YA Fund (claims 5781, 5780)—a structure typically deployed under liquidity strain. However, announced use of proceeds diverged sharply from capex intent: the $362M net proceeds from October's $300M convertible offering (claim 5475) allocated to $31M share repurchases (claim 5478), $53M prepaid forward transactions (claim 5477), and $43M capped call hedges (claim 5476), with facility capex unspecified. These financial engineering moves contradict the August disclosures of "financing discussions for HPC Ellendale Facility" (claims 5892, 5867, 5907). A sophisticated reader should monitor: (1) drawdown pace on the $125M CIM facility outstanding balance (claim 5549) and $75M accordion capacity, since acceleration signals cash depletion; (2) triggering of the $250M SEPA commitment with YA Fund, which would indicate covenant pressure and distressed equity pricing; (3) deployment of remaining convertible proceeds into asset purchases versus additional financial hedges or debt service, as the gap between announced capex and actual spend has already materialized.

Applied Digital's capital structure in 2024 reflects deliberate infrastructure-stage financing rather than opportunistic patching. The company issued Series E preferred at a $9 dividend rate (claim 6136), then Series F at $8 (claim 5835) and Series E-1 at $62.5M (claim 5656)—compressed through successive rounds. Project-level debt is tied to specific customer contracts, notably a 400 MW hyperscaler lease (claim 3237), suggesting capital deployment tracks to revenue generation.

The layering of simultaneously available facilities—$125M at-the-market (claim 5778), $250M standby equity (claim 5659), and a $125M credit facility with $75M accordion (claims 5660, 5975)—indicates intentional design rather than ad-hoc patching. This architecture funds expansion at multiple price points without returning to the market. The $300M convertible with $43M capped calls (claims 5481, 5476) implies management expects material share price appreciation.

Deployment tracking shows gaps worth explaining. The AI Bridge unsecured $20M loan (announced February, claim 6292) remains marked delayed. Yorkville's first and second promissory notes ($40M and $10M, claims 6079, 6080) also carry delayed status despite later conversions appearing stated. The S-1 effectiveness required for preferred registration rights remains marked delayed (claim 6108). These delays do not automatically indicate problems—committed facilities often precede draws—but they require explanation if proceeds deployment is claimed.

A sophisticated reader should track: whether the $75M CIM accordion activates (signaling capex momentum versus working capital pressure), whether the $250M SEPA draws (indicating equity weakness relative to conversion terms), whether announced hyperscaler revenues materialize on announced timelines, and what explains the gaps between stated and delayed status on foundational debt facilities. The compressed preferred terms and aggressive hedging on the convertible suggest confidence in stock performance, which should be testable against actual execution.

Applied Digital raised approximately $6 billion in debt financing between March and June 2026, deployed through specific infrastructure facilities. The $2.15 billion senior secured notes in March (claim 2247) and subsequent $1.59 billion notes in June (claims 213, 222) were earmarked for Polaris Forge 1 and 2 construction (claims 866, 868). A $300 million bridge facility (claim 342) and $430 million in secured revolving credit (claim 152) completed the capital stack. Equity remained largely undeployed—a $2 billion Series G convertible commitment increase (claim 154) was arranged but not drawn, leaving the capital structure debt-dominant by construction.

The financing pattern exhibits project-specific coherence rather than opportunistic patchwork. Proceeds track to named facilities with explicit end-uses, suggesting disciplined capital allocation. However, risk hinges on execution. FY2026 revenue of $611 million (claim 321) and a net loss of $250 million (claim 322) with negative $3 million free cash flow (claim 319) indicate the company remains in capex-heavy growth mode, not yet profitable. Q4 2026 revenue of $259 million (claim 858) showed sequential acceleration, and adjusted EBITDA of $42.4 million (claim 376) signals emerging unit-level economics, yet company-wide cash burn persists. The June balance sheet—current ratio of 4, debt-to-equity of 3 (claims 318, 317)—became stale immediately after the June debt issuances, and actual leverage is now measurably higher.

Contracted lease revenue of $36 billion at base term (claim 491) and $86 billion including renewal options (claim 492) anchors the capital plan to customer commitments, a necessary counterweight to the debt load. But these are multiyear revenue streams, not near-term liquidity.

Insider selling presents ambiguity. Sales accelerated from $1.9 million in Q4 2025 to $46.7 million in Q3 2026 without offsetting buys, suggesting either vesting-driven liquidation tied to equity appreciation or reduced executive conviction. The mechanics require clarification.

Readers should monitor: whether contracted revenue executes on schedule; when free cash flow inflects positive; and the sustainability of capital spending given current cash burn. The debt narrative is transparent and facility-specific, but execution concentration is high—large leases, rapid deployment, and multiyear payoff timelines leave little room for revenue slippage or capex overruns.

Applied Digital's record splits into two unrelated eras, and only the second describes customers. The 2006 to 2008 items come from the predecessor shell, Flight Safety Technologies. They comprise one government development contract with the DOT Volpe Center (claim 21053) and three vendor engagements (claims 21068, 21067, 21069): Broadridge, Wolf & Company and Pacific Stock Transfer. These are suppliers to the company, not buyers from it, and should not count toward commercial traction. From August 2021 the disclosures shift to crypto hosting, and the counterparty picture moves from anonymous to named over about seven months. The first filing cited only 2 unnamed customers taking more than 50% of available energy (claim 14506). By November that became 4 customers covering the full first-facility allocation plus 85MW at a second site, still unnamed (claims 14412, 14374, 14363). Not until March 2022 were the four identified: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14200, 14210, 14071). GMR's roughly 50% share of the 100MW site is the one allocation consistently quantified as a signed contract (claims 14411, 14231, 14201, 14072). Concentration is the central diligence issue. At least half of first-site capacity, and probably more, sits with JointHash and Bitmain. Both are tied to GMR and Bitmain, which are also strategic equity investors and service providers (claims 14505, 14377, 14378). Bitmain additionally supplies the company's own 200PH equipment order (claims 14504, 14416) and refers hosting leads (claim 14417). Its affiliate Antpool is the company's partner in the 1.21 Gigawatts LLC joint venture, which carries a disclosed $8,000 figure (claims 14234, 14212). The customer base therefore overlaps heavily with the shareholder and supplier base, and these agreements should be read as related-party arrangements rather than arm's-length demand. Among the forward-looking items, only the Texas 200MW wind facility is explicitly non-binding: it is a letter of intent subject to definitive agreements (claims 14204, 14074). The Antpool vehicle is a signed LLC agreement, but its 'up to 1.5GW' scope is aspirational (claim 14203). The SparkPool advisory relationship is also conditional (claim 14418). On revenue arithmetic, the record contains no contract values, pricing, term lengths or revenue guidance attached to any hosting deal. The only dollar figures are a $22,000 proxy fee and an $8,000 JV item. Neither relates to customer revenue, so the announced megawatt allocations cannot be reconciled to any disclosed revenue base. The claim list supplied was also truncated at claim 14061, so later disclosures were not assessed.

Applied Digital's record separates into signed agreements, unnamed customer counts, and conditional or advisory arrangements, and the signed column is narrower than the headline language suggests. The 2006-2008 entries come from the predecessor shell, Flight Safety Technologies. Only the Volpe Center award (claim 21053) is a customer contract; Wolf & Company, Broadridge and Pacific Stock Transfer (claims 21067, 21068, 21069) are service vendors and should not count toward commercial traction.

In the hosting era, counterparties were identified slowly. In August 2021 the company cited agreements with "2 customers" covering more than half of available energy without naming them (claim 14506), while saying GMR and Bitmain had hosting contracts (claim 14507). By November 2021 the count was "4 customers", still unnamed, covering the full first facility plus 85MW at a second site (claims 14412, 14374, 14363). Not until March 2022 were all four named: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 14232, 14200, 14210, 14071).

Several arrangements are not customer revenue. The Texas wind facility is a letter of intent expressly subject to definitive agreements (claims 14204, 14074). SparkPool's role is advisory and engineering support (claims 14418, 14358). Coinmint is a co-hosting provider, so there Applied is the customer (claim 14419).

Concentration is high and circular. GMR alone holds roughly 50% of the 100MW first site (claims 14411, 14201, 14072). The remaining capacity is spread across the other three named customers. Bitmain is at once an equity investor, the equipment supplier on the 200PH order, a hosting customer, and, through its affiliate Antpool, a joint-venture partner in 1.21 Gigawatts LLC (claims 14378, 14416, 14232, 14234). Customer, investor and supplier relationships therefore overlap heavily, and the arm's-length quality of contract pricing cannot be judged from this record.

The revenue arithmetic cannot be reconciled. No contract values, per-MW rates, terms or take-or-pay provisions appear in any claim. The only dollar figures are a $22,000 proxy-solicitation fee (claim 21068) and an $8,000 figure tied to the Antpool LLC (claim 14234), and neither is customer revenue. Capacity, not dollars, is the only quantifiable commitment: 100MW under the Amended and Restated Energy Services Agreement (claims 14233, 14202) plus 85MW at an unbuilt second site. The supplied record breaks off mid-claim after 14059, so later deals fall outside this read.

Applied Digital's record splits into two unrelated eras, and only the second bears on commercial diligence. The 2006–2008 items belong to the predecessor shell, Flight Safety Technologies. They comprise a single government development contract with the DOT Volpe Center (claim 21053) and three service vendors: an auditor, a proxy solicitor retained for about $22,000, and a transfer agent (claims 21067, 21068, 21069). None of the vendors is a customer. The substantive history begins in August 2021, and its defining feature is that counterparties were named late. That month the company said it had agreements with "2 customers" covering more than half of available energy, without identifying them (claim 14506). Separately, it credited GMR and Bitmain with hosting contracts meant to pre-fill the first 100MW site (claim 14507). By November the count had become "4 customers" absorbing all first-facility energy plus 85MW at a second site, still unnamed (claims 14412, 14374, 14363). Only the March 2022 annual filing identified them as JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 14232, 14200, 14210, 14071). These read as executed agreements, as do the Antpool joint-venture LLC agreement (claim 14234) and the amended energy services agreement (claim 14233). The Texas 200MW wind project, by contrast, is a letter of intent expressly "subject to entry into definitive agreements" (claims 14204, 14074). It should not be counted as contracted capacity. Concentration is severe and circular. GMR alone holds roughly 50% of the first 100MW (claims 14411, 14231, 14201). GMR and Bitmain are also strategic equity investors (claims 14505, 14377, 14378). Bitmain additionally supplies the 200PH rig order (claim 14504), feeds hosting leads (claim 14417), and through its affiliate Antpool co-owns the 1.21 Gigawatts venture (claim 14234). Two of the four named customers are therefore shareholders, and the largest supplier doubles as a customer. That makes these closer to related-party arrangements than arm's-length demand. Coinmint is a host for the company's own mining, not a customer (claim 14419). SparkPool is an adviser whose status statement is cut off in the record (claims 14418, 14358). The revenue arithmetic cannot be reconciled from this record. No contract values, per-megawatt pricing, terms or revenue recognition appear anywhere. The only dollar figures are the $22,000 proxy fee and an $8,000 amount tied to the Antpool LLC. The capacity math is internally consistent: GMR at about 50MW leaves about 50MW across three other customers. The shift from two customers above 50% to four customers at 100% shows that allocations were still moving within one quarter. The record ends in April 2022, so later counterparties cannot be assessed here.

Applied Digital's record splits into two unrelated eras. The 2006-2008 entries come from the predecessor shell, Flight Safety Technologies. Only the Department of Transportation Volpe Center award (claim 21053) is a customer contract. The auditor, transfer agent and Broadridge proxy engagements (claims 21067, 21069, 21068) are vendor relationships and should not be counted as commercial traction. The relevant history starts in August 2021.

In August 2021, the company described agreements with two unnamed customers covering more than 50% of available energy (claim 14506). In the same filing it named GMR and Bitmain as having hosting contracts (claim 14507). By November 2021 the count had become four customers, still unnamed. These four were said to absorb all of the first facility's energy plus 85MW at a second site (claims 14412, 14374, 14363). The four were not identified until March 2022: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (claims 14232, 14200, 14210, 14071). Investors therefore relied on unnamed customer counts for about seven months before the counterparties were disclosed.

Concentration is the main issue. GMR alone was repeatedly said to hold roughly 50% of the 100MW first site (claims 14411, 14231, 14201). GMR, Bitmain and SparkPool are also strategic equity investors (claims 14505, 14376, 14377, 14378). Antpool, a Bitmain affiliate, is the joint-venture partner in 1.21 Gigawatts LLC (claims 14234, 14203). On the record, at least two of the four anchor customers are also shareholders or affiliates of the JV partner. That overlap weakens how independent the demand signal is.

Several items are not customer revenue at all. The Texas 200MW wind facility is a letter of intent explicitly subject to definitive agreements (claims 14204, 14074). The Antpool JV's 'up to 1.5GW' is a framework ambition, not contracted capacity (claim 14203). The Coinmint co-hosting arrangement (claims 14419, 14371) and the Bitmain 200PH rig order (claims 14504, 14416) make Applied Digital the buyer, not the seller. The SparkPool work (claims 14418, 14358) consists of advisory and services agreements.

The revenue arithmetic cannot be reconciled from this record. No contract values, per-MW pricing or recognized hosting revenue are disclosed. The only dollar figure tied to these deals is the $8,000 associated with the 1.21 Gigawatts formation (claim 14234), and it reads as a capital commitment rather than revenue. The claim list provided was truncated at claim 14059, so later disclosures were not reviewed.

Applied Digital's record splits into two unrelated eras. The 2006-2008 entries come from the predecessor shell, Flight Safety Technologies. Only the Department of Transportation Volpe Center award reads as a signed customer contract (claim 21053). Broadridge, Wolf & Company and Pacific Stock Transfer are vendors the company paid, not customers (claims 21068, 21067, 21069). They should carry no weight in a commercial read of the current business.

In the crypto-hosting era, disclosure ran ahead of identification. In August 2021 the company cited agreements with two customers covering more than half of available energy but did not name them (claim 14506). It separately said GMR and Bitmain held hosting contracts to pre-fill the first 100MW site (claim 14507), which suggests, without confirming, that those were the two. By November the unnamed count had grown to four, covering all of the first facility and 85MW of a second (claims 14412, 14374, 14363). Only in March 2022 were all four named: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (claims 14232, 14210, 14200).

Concentration is the dominant issue. GMR alone contracted for roughly 50% of the 100MW capacity (claims 14411, 14231, 14201). GMR and Bitmain are also strategic equity investors (claims 14505, 14377, 14378). Bitmain is additionally the equipment supplier and a source of customer leads (claims 14416, 14417), and its affiliate Antpool is the company's joint-venture partner in 1.21 Gigawatts LLC (claims 14234, 14203). The anchor demand is therefore heavily related-party and circular rather than arm's-length.

Several items should not be counted as customer revenue. The Texas 200MW wind facility is a letter of intent that is expressly subject to definitive agreements (claims 14204, 14074). The SparkPool arrangement is advisory (claim 14418). Coinmint is a host the company paid (claim 14419). 'Strategic partner' language around SparkPool, GMR and Bitmain blends investor, vendor and customer roles.

Revenue arithmetic cannot be reconciled from this record. No contract values, hosting rates, terms or revenue figures accompany any customer claim. The only dollar amounts are the $22,000 proxy fee and the $8,000 joint-venture capitalization. The megawatt allocations are internally consistent: 100MW is fully allocated, with roughly 50MW to GMR. However, megawatts are not dollars. The record supplied was truncated at claim 14059, so later disclosures may change this picture.

Applied Digital's record splits into two unrelated eras. The 2006–2008 entries come from the predecessor Flight Safety Technologies. Only one of them is a customer-side award: the Volpe Center emulation contract (21053). The Broadridge proxy retention at roughly $22,000 (21068), the Wolf & Company auditor appointment (21067) and Pacific Stock Transfer (21069) are vendor relationships. They should not be read as commercial traction. The relevant story begins in August 2021, and much of it is language that sounds like customer revenue but describes something else. SparkPool, GMR and Bitmain are introduced as strategic equity investors (14505, 14376–14378). The services they provide run toward the company: GMR's asset-management agreement (14502), SparkPool's mining-pool advisory (14418) and Bitmain's preferred equipment access (14417). The 200PH Bitmain order is a purchase commitment (14504, 14416, 14360). The Coinmint co-hosting agreement places Applied Digital as the hosted party (14419, 14371). The Texas 200MW wind facility is explicitly a letter of intent 'subject to entry into definitive agreements' (14204, 14074). The Antpool vehicle, 1.21 Gigawatts LLC, is a joint-venture framework rather than a hosting contract (14234, 14212). Disclosure of hosting customers progressed from anonymous to named over about seven months. In August 2021 the company cited two unidentified customers for more than 50% of capacity (14506). In November it cited four unnamed customers absorbing the full first-site energy plus 85MW at a second site (14412, 14374, 14363). Not until March 2022 were the four identified: JointHash (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (14232, 14210, 14200). That naming exposes heavy concentration. GMR alone holds roughly half of the 100MW site (14411, 14231, 14201). Bitmain is simultaneously equity holder, equipment vendor, customer and, through Antpool, JV partner. At least half of contracted capacity therefore sits with related-party investors rather than arm's-length buyers. On the revenue side, the record supplies no contract values, per-MW pricing, term lengths or recognized hosting revenue to test against the announcements. The only dollar figures are the $22,000 proxy fee and an $8,000 figure tied to the Antpool LLC formation (14234). Neither bears on customer revenue. The arithmetic therefore cannot be reconciled from this record. 'Fully contracted' capacity claims should be treated as allocation statements, not demonstrated revenue, until filings attach dollar amounts. Note also that the claim list provided ends mid-entry at 14059 (April 2022), so later developments are not assessed here.

Applied Digital's record splits into two unrelated eras, and only the second bears on commercial diligence. The 2006 Department of Transportation Volpe Center award (claim 21053) is a genuine government contract, but it belongs to the predecessor Flight Safety Technologies. The 2008 items naming Wolf & Company, Broadridge and Pacific Stock Transfer (claims 21067, 21068, 21069) are vendor engagements, not customers, and should carry no weight. The hosting business that matters begins in August 2021.

The disclosures moved from anonymous to named over about seven months. In August 2021 the company cited agreements with two unidentified customers covering more than half of available energy (claim 14506). It separately said GMR and Bitmain had entered hosting contracts to pre-fill the first 100MW site (claim 14507). By November the count had grown to four customers, still unnamed, covering the full first facility plus 85MW at a second (claims 14412, 14374, 14363). Only in March 2022 were all four identified: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (claims 14232, 14200, 14210, 14071). These read as signed agreements, not MOUs. The Texas 200MW wind project is expressly a letter of intent subject to definitive agreements and should not be counted as contracted capacity (claims 14204, 14074).

Concentration is the central issue. GMR alone holds roughly 50% of the first 100MW (claims 14411, 14231, 14201). GMR and Bitmain are also strategic equity investors (claims 14505, 14377, 14378), which gives the anchor book a related-party character. Bitmain is simultaneously a hosting customer, the equipment supplier on a 200PH order (claims 14504, 14416), a source of customer leads (claim 14417), and, through its affiliate Antpool, the company's partner in the 1.21 Gigawatts LLC joint venture (claims 14234, 14203). The SparkPool and Coinmint arrangements are advisory and co-hosting relationships in which Applied Digital is the buyer, not the seller (claims 14418, 14419). The Energy Services Agreement is a supply-side input (claim 14233).

The revenue arithmetic cannot be reconciled from this record. No contract values, per-megawatt pricing, tenors or take-or-pay terms are disclosed. Capacity allocations (100MW plus 85MW) are the only quantitative anchor. The $8,000 attached to the Antpool LLC formation (claim 14234) is unexplained in the quoted text. The supplied record also truncates after claim 14059, so later disclosures were not assessed.

Applied Digital's customer record before its AI-infrastructure pivot is mostly signed agreements, but the counterparties stayed unnamed for months, and the customers overlap heavily with the company's own investors. The 2006–2008 entries come from the Flight Safety Technologies predecessor. They are a Volpe Center government contract (21053) and service vendors, namely the auditor, transfer agent and Broadridge's roughly $22,000 proxy fee (21067, 21069, 21068). None of these bear on the hosting business.

In August 2021 the company disclosed signed hosting agreements with GMR and Bitmain (14507). It also cited agreements with "2 customers" covering more than half of available energy, without naming them (14506). By November the unnamed count had grown to "4 customers" absorbing all of the first facility and 85MW of a second (14412, 14374, 14363). Those four were identified only in March 2022 as JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (14232, 14200, 14210, 14071). The agreements are described as entered into, not as MOUs. However, no term, price or take-or-pay provisions appear in the record.

Several relationships are not customer revenue at all. SparkPool's role is advisory and supply-side (14418, 14358). The Coinmint arrangement has Applied Digital as the hosted party (14419). The 200PH Bitmain order is procurement (14416). The Texas 200MW wind facility is a letter of intent, expressly "subject to entry into definitive agreements" (14074, 14204), and should not be read as contracted capacity.

Concentration is high and related-party in character. GMR alone holds roughly 50% of the 100MW first site (14411, 14231, 14211). GMR and Bitmain are both strategic equity investors (14505, 14377, 14378). Bitmain's affiliate Antpool is also the company's joint-venture partner in 1.21 Gigawatts LLC (14234, 14212). So at least half the contracted load, and probably more, sits with shareholders or their affiliates.

The revenue arithmetic cannot be reconciled from this record. No contract values, per-MW rates or recognized hosting revenue are disclosed. The only figure tied to the hosting partners is an $8,000 item linked to the Antpool JV formation (14234), and the record gives it no units or context. Capacity statements (100MW plus 85MW fully allocated) are the only quantifiable claims, and they measure power commitments rather than dollars. The claim list is also truncated at 14061, so later disclosures may alter this picture.

Applied Digital's record splits into two unrelated eras. The 2006–2008 claims belong to the predecessor shell, Flight Safety Technologies. Only the Volpe Center contract (21053) was revenue-bearing; Broadridge, Wolf & Company and Pacific Stock Transfer (21068, 21067, 21069) are vendors and say nothing about customer demand. The substantive story begins in August 2021, and its defining feature is how far the early customers overlap with the company's own investors.

The naming of counterparties moved in stages. In August 2021 the company cited agreements with two unidentified customers covering more than half of available energy (14506). By November that had become four unnamed customers absorbing the full 100MW at the first facility plus 85MW at a second (14412, 14374, 14363). The four were named only in the March 2022 filings: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (14232, 14210, 14200). Those are signed hosting agreements, not MOUs.

The concentration is related-party in nature. GMR and Bitmain are strategic equity investors (14505, 14377, 14378). GMR alone holds roughly 50% of the 100MW (14411, 14231, 14201). Bitmain is also the equipment supplier (14504, 14416) and the source of hosting leads (14417). Its affiliate Antpool is the joint-venture partner in 1.21 Gigawatts LLC (14234, 14203). Before ground was broken, the first site was described as pre-filled by these partners (14507). That pre-fill shows insiders committing to the site, not independent third-party demand. F2Pool and Hashing LLC are the only named hosting customers without a disclosed equity tie.

Several items should be kept apart from signed revenue contracts. The Texas wind project is an explicit letter of intent "subject to entry into definitive agreements" (14204, 14074). The SparkPool arrangement is advisory (14418, 14358). The Coinmint agreement makes Applied Digital the customer, not the provider (14419). The Antpool JV is a capital vehicle whose "up to 1.5GW" framing is aspirational (14203). The $8,000 figure attached to 14234 is ambiguous and should not be read as deal value without the underlying filing.

Revenue arithmetic cannot be reconciled from this record. None of the claims disclose contract pricing, per-MW rates, term length or recognized revenue, so there is no basis to test the 185MW of contracted capacity against reported results. The claim set also ends partway through the April 2022 filings (14059 onward is cut off). Later developments, including any shift away from crypto hosting toward other compute customers, are therefore outside this analysis.

Applied Digital's record falls into two eras. The 2006-2008 entries belong to the predecessor shell, Flight Safety Technologies. The Volpe Center award (21053) is the only customer-type contract from that period. Broadridge, Wolf & Company and Pacific Stock Transfer (21068, 21067, 21069) are service vendors and should not be read as commercial traction. The substantive record begins in August 2021, and the excerpt provided ends mid-entry at claim 14059, so later disclosures are outside this read.

Customer disclosure moved from anonymous to named over roughly seven months. In August 2021 the company cited agreements with two unidentified customers covering more than half of available energy (14506). It also named GMR and Bitmain as hosting counterparties expected to fill the first planned 100MW site (14507). By November the count had risen to four customers covering the full first facility and 85MW of a second, still unnamed (14412, 14374, 14363). Only in March 2022 were all four identified: JointHash Holding (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (14232, 14210, 14200).

The naming exposes heavy concentration and a related-party tilt. GMR alone is described as holding roughly 50% of the 100MW first facility (14411, 14201). GMR and Bitmain are also disclosed strategic equity investors, alongside SparkPool (14505, 14377, 14378). So at least two of the four anchor customers are shareholders, and the earliest 'pre-filled' capacity came from investor-affiliated parties. F2Pool and Hashing LLC appear to be the only arm's-length names, and their share is not disclosed.

Several headline relationships are not customer revenue at all. The Bitmain 200PH order (14504, 14416) is an equipment purchase, and the Coinmint arrangement (14419) has Applied Digital as the hosted party. SparkPool's role is advisory (14418, 14358). The 1.21 Gigawatts LLC with Antpool, a Bitmain affiliate, is a joint-venture formation with only an $8,000 figure attached (14234, 14203). The Texas wind facility is explicitly a letter of intent subject to definitive agreements (14204, 14074).

The revenue arithmetic cannot be reconciled from this record. None of the customer claims discloses contract value, per-MW pricing, term, prepayments or take-or-pay provisions. The capacity figures (100MW, 85MW, GMR's roughly 50%) are internally consistent, but they describe energy allocation, not dollars. Any revenue projection built on these four agreements rests on terms the claims do not supply.

Applied Digital's record contains few true signed-customer disclosures, and the counterparties were named only after a delay. The 2006-2008 items come from the predecessor Flight Safety Technologies shell. These are a Volpe Center development contract (claim 21053) and vendor engagements with Broadridge, Wolf & Company and Pacific Stock Transfer (claims 21068, 21067, 21069), none of which bear on the hosting business. In the August 2021 filing, the company said it had agreements with "2 customers" covering more than half of the available energy, without naming them (claim 14506). By November 2021 this had grown to "4 customers" covering the full first facility and 85MW at a second site, still unnamed (claims 14412, 14374, 14363). Only in March 2022 were the four identified as JointHash Holding (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 14232, 14200, 14210, 14071). GMR's contract for roughly 50% of the 100MW capacity is the one hosting commitment named and sized from the start (claims 14411, 14231, 14201). Concentration is the central diligence issue. Half of first-site capacity sits with GMR. Two of the four named customers, GMR and Bitmain, are also strategic equity investors, alongside SparkPool (claims 14505, 14377, 14378). Bitmain's roles overlap heavily: it supplies equipment (claims 14504, 14416), refers hosting customers (claim 14417), is itself a customer, and its affiliate Antpool is the 1.21 Gigawatts joint-venture partner (claims 14234, 14203, 14212). The early customer base is therefore largely made up of insiders and related parties, not arm's-length demand. Several items should be kept apart from customer contracts. The Texas 200MW wind project is a letter of intent explicitly "subject to entry into definitive agreements" (claims 14204, 14074). The 1.21 Gigawatts entity is a framework joint venture for up to 1.5GW, and the $8,000 associated with it in claim 14234 reflects formation scale, not committed capacity. The SparkPool and GMR arrangements are advisory or services agreements (claims 14418, 14502, 14359). Coinmint is a vendor that hosts Applied Digital's own rigs, not a customer (claim 14419). The revenue arithmetic cannot be reconciled from this record. No contract values, per-MW pricing, terms or recognized revenue are disclosed against any named customer, so the 100MW-plus-85MW allocation has no dollar figure to test. The supplied record is also truncated at claim 14059, so later disclosures, including any shift toward HPC or AI tenants, are not reflected in this read.

Applied Digital's record splits cleanly into two eras, and only the second involves customers in any commercial sense. The 2006-2008 entries belong to predecessor Flight Safety Technologies. The Volpe Center award (claim 21053) was a genuine government development contract. The Broadridge, Wolf & Company and Pacific Stock Transfer items (claims 21068, 21067, 21069) are vendor engagements and should not be read as demand signals. The hosting-era disclosures from 2021 onward begin with heavy reliance on 'strategic partner' framing. SparkPool, GMR and Bitmain are described as equity investors who also supply advisory services, equipment access and hosting leads (claims 14505, 14376, 14417). In August 2021 the company reported agreements with '2 customers' covering more than half of available energy without naming either (claim 14506). By November that had become '4 customers' absorbing the full first-facility allocation plus 85MW at a second site, still unnamed (claims 14412, 14374, 14363). Names first appear in the March 2022 filings: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (claims 14232, 14200, 14210). That disclosure confirms how concentrated the book is. Two of the four signed hosting customers are affiliates of the company's own equity investors. GMR alone holds roughly 50% of the 100MW site (claims 14411, 14201). Bitmain is also equipment supplier, lead generator and, through Antpool, joint-venture partner in 1.21 Gigawatts LLC (claims 14234, 14212). The Coinmint arrangement (claim 14419) runs the other way, with Applied Digital as the hosted party rather than the host. Signed instruments therefore consist of the hosting contracts, the amended Energy Services Agreement (claim 14233) and the Antpool LLC agreement. The Texas 200MW wind project remains a letter of intent expressly 'subject to entry into definitive agreements' (claim 14074), and the 1.5GW Antpool ambition is a framework rather than a commitment. Revenue arithmetic cannot be reconciled from this record. No contract values, per-megawatt rates, terms or recognized revenue are disclosed for any hosting customer. The only dollar figures are a $22,000 proxy fee and an $8,000 LLC contribution, neither of which bears on hosting economics. Readers should also note that the supplied claim list is truncated after claim 14059, so later customer disclosures are not reflected here.

Applied Digital's record falls into two eras, and only the second contains customer evidence. The 2006-2008 items come from the predecessor shell, Flight Safety Technologies. A Volpe Center development contract (claim 21053) is the only revenue counterparty. Broadridge, Wolf & Company and Pacific Stock Transfer (claims 21068, 21067, 21069) are vendors the company paid, not customers, and should not be counted as commercial traction.

The 2021-2022 hosting narrative moved from anonymous to named counterparties over about seven months. In August 2021 the company cited agreements with two unidentified customers covering more than 50% of available energy (claim 14506), alongside hosting contracts with GMR and Bitmain (claim 14507). By November the unnamed count had grown to four, covering the full first facility plus 85MW at a second site (claims 14412, 14374, 14363). The four were not identified until March 2022: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 14232, 14210, 14200, 14071). Those are signed agreements, but the filings describe them as accounting for energy allocation, not as stating contract value or term.

Concentration is the central diligence issue. GMR alone holds roughly half of the 100MW first site (claims 14411, 14231, 14201). GMR and Bitmain, two of the four named customers, are also strategic equity investors (claims 14505, 14377, 14378). Bitmain's affiliate Antpool is the company's partner in the 1.21 Gigawatts joint venture (claims 14234, 14212). A substantial share of contracted demand therefore sits with shareholders and their affiliates, which weakens it as evidence of arm's-length market demand. SparkPool's role is advisory and supply-side (claims 14418, 14358), and Coinmint is a co-hosting provider (claim 14419), so neither is a customer.

Several items are not binding customer commitments. The Texas 200MW wind facility is explicitly a letter of intent subject to definitive agreements (claims 14204, 14074). The Bitmain 200PH order (claims 14504, 14416) is a purchase the company made, and Bitmain's provision of hosting leads (claim 14417) is not a contract.

On revenue reconciliation, the record contains no disclosed contract values, pricing, terms or recognized revenue tied to any named hosting customer. Megawatt allocations therefore cannot be checked against reported figures. The only dollar amount, an $8,000 figure attached to the Antpool LLC formation (claim 14234), does not state its units and relates to joint-venture capital, not customer revenue. The source record is also truncated after claim 14059, so later disclosures were not reviewed.

Applied Digital's record splits into two unrelated eras, and only the second bears on its current customer base. The 2006 Department of Transportation Volpe Center award (claim 21053) is a genuine signed government contract, but it belongs to the predecessor Flight Safety Technologies. The 2008 items naming Broadridge, Wolf & Company and Pacific Stock Transfer (claims 21068, 21067, 21069) are vendor engagements, not customers, and should not be counted as commercial traction.

The 2021–2022 hosting pivot shows a clear progression from anonymous to named counterparties. In August 2021 the company cited agreements with "2 customers" covering more than half its energy without identifying them (claim 14506). By November that became four unnamed customers absorbing the full first-facility allocation plus 85MW at a second site (claims 14412, 14374, 14363). Only in the March 2022 filings were they named: JointHash Holding (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 14232, 14200, 14210, 14071). These read as signed agreements, though the capacity they cover was still described as "under development" (claim 14411).

Concentration is the central diligence issue. GMR alone holds roughly 50% of the 100MW first facility (claims 14411, 14231, 14201). GMR and Bitmain are also strategic equity investors (claims 14505, 14377, 14378), so the two largest hosting customers are shareholders as well. Bitmain is simultaneously customer, equipment supplier for a 200PH order (claims 14504, 14416), lead generator (claim 14417), and, through affiliate Antpool, joint-venture partner in 1.21 Gigawatts LLC (claims 14234, 14212). Counting Bitmain and Antpool as one economic group and GMR and JointHash as another, two related-party groups likely account for most contracted capacity.

Several items are softer than their headlines suggest. The Texas 200MW wind facility is explicitly a letter of intent "subject to entry into definitive agreements" (claims 14204, 14074). The SparkPool pool-advisory deal is conditional on the company deciding to proceed (claim 14418), and SparkPool's own status was in transition (claim 14358). The Coinmint arrangement (claim 14419) and the Energy Services Agreement (claim 14233) are supply-side relationships in which Applied Digital is the buyer.

The revenue arithmetic cannot be reconciled from this record. No contract values, per-MW pricing or recognized revenue are disclosed for any hosting customer. The only dollar figures are a $22,000 proxy fee, which is a cost (claim 21068), and an $8,000 figure tied to the Antpool joint venture whose units and nature are unstated (claim 14234). Capacity claims are internally consistent in megawatts, but they cannot be translated into revenue. The record supplied is also truncated after claim 14059.

Applied Digital's record splits into two unrelated eras, and only the second involves customers. The 2006–2008 entries come from the predecessor Flight Safety Technologies: a Volpe Center research contract (claim 21053) and three service vendors, namely Broadridge, Wolf & Company and Pacific Stock Transfer (claims 21068, 21067, 21069). The vendors are suppliers, not customers, and none of these entries bear on the hosting business.

The 2021–2022 hosting disclosures show customers being named progressively. In August 2021 the company cited agreements with "2 customers" accounting for more than half of available energy, without naming them (claim 14506). By November it had moved to "4 customers" for the full first-facility allocation plus 85MW at a second site, still unnamed (claims 14412, 14374, 14363). Only in the March 2022 filings are the four identified: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (claims 14232, 14200, 14210, 14071). GMR's contract for roughly 50% of the 100MW site is the one deal tied to a named counterparty from the start (claims 14411, 14201). These read as signed agreements, though no term, price or take-or-pay language is quoted.

Other items should not be read as customer revenue. The Texas 200MW wind facility is explicitly a letter of intent "subject to entry into definitive agreements" (claims 14204, 14074). The 1.21 Gigawatts LLC with Antpool is a joint-venture framework rather than a customer contract (claims 14234, 14203). The Bitmain 200PH order is equipment supply (claims 14504, 14416), the Coinmint arrangement has Applied Digital as the hosted party (claim 14419), and SparkPool's role is advisory (claim 14418).

Concentration is high and related-party in character. SparkPool, GMR and Bitmain are described as strategic equity investors (claims 14505, 14376). Antpool is a Bitmain affiliate. Two of the four named hosting customers are therefore investor affiliates, and GMR alone holds about half of the first site.

The revenue arithmetic cannot be reconciled from this record. No contract values, hosting rates or revenue figures are disclosed for any deal. The only dollar amounts are a $22,000 proxy fee (claim 21068) and an $8,000 Antpool figure whose units are truncated (claim 14234). Megawatt allocations are the sole quantified measure, and the record itself ends mid-quote at claim 14059.

Applied Digital's record splits into two unrelated eras, and only the second involves customers. The 2006–2008 entries belong to the predecessor shell, Flight Safety Technologies. The Volpe Center award (claim 21053) is the only genuine customer contract of that period. Broadridge, Wolf & Company and Pacific Stock Transfer (claims 21068, 21067, 21069) are vendors the company pays, not counterparties that pay it. They should carry no weight in any customer read.

The 2021 hosting pivot shows a clear move from unnamed to named counterparties. In August 2021 the company cited agreements with "2 customers" for more than half the energy under its Energy Services Agreement, without naming them (claim 14506). It also leaned on SparkPool, GMR and Bitmain as "strategic equity investors" (claim 14505). By November the count had grown to four unnamed customers, said to take all of the first facility's energy plus 85MW at a second site (claims 14412, 14374, 14363). The four were not named until the March 2022 filings: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (claims 14232, 14200, 14210, 14071). Only GMR's commitment, roughly 50% of the 100MW, is ever sized individually (claims 14411, 14201).

Separating signed from soft language matters here. The hosting contracts, the Amended and Restated Energy Services Agreement (claim 14202) and the Antpool LLC agreement forming 1.21 Gigawatts (claims 14234, 14212) are described as executed. The Texas 200MW wind project, by contrast, is explicitly a letter of intent "subject to entry into definitive agreements" (claims 14204, 14074). The SparkPool mining-pool arrangement is advisory, not revenue-bearing (claim 14418). The "up to 1.5GW" framing attached to the Antpool venture (claim 14203) is a capacity ambition, not contracted load.

Concentration is the central diligence issue. The customer base is very narrow, and it is entangled with the company's own shareholders. GMR and Bitmain are equity investors, suppliers or advisers, and also hosting customers. Antpool, a Bitmain affiliate, is a joint-venture partner as well. Two of four customers, holding at least half the first site, are therefore related to the cap table. That raises questions about arm's-length pricing and about how durable the revenue would be if those relationships soured.

The revenue arithmetic cannot be reconciled from this record. No contract value, per-MW rate, term or recognized revenue is disclosed for any hosting agreement. The only figures are a $22,000 proxy fee (claim 21068) and an $8,000 entry tied to the Antpool LLC (claim 14234), whose units and meaning are unclear. Both are cost items, not revenue. Investors are being asked to accept full-capacity commitments on faith. The record set also truncates at claim 14059, so later disclosures may change this picture.

Applied Digital's record splits cleanly into a pre-2021 shell era and a 2021–2022 hosting buildout, and only the second contains anything resembling commercial customers. The 2006 Volpe Center award (claim 21053) was a genuine government contract for the predecessor Flight Safety Technologies, but the 2008 items naming Pacific Stock Transfer, Wolf & Company and Broadridge (claims 21069, 21067, 21068) are vendors the company pays, not customers, and should carry no weight as demand evidence. Several 2021 relationships also run the wrong direction. The Bitmain 200PH rig order (claims 14504, 14416, 14360) and the Coinmint co-hosting agreement (claims 14419, 14371, 14357) are purchases of equipment and capacity by Applied Digital. Bitmain's 'preferred access' and hosting 'leads' (claim 14417) describe a relationship, not a contract.

The identification of hosting customers improved over time. In August 2021 the company cited agreements with 2 unnamed customers for more than 50% of available energy (claim 14506), alongside hosting contracts attributed to GMR and Bitmain (claim 14507). By November that had become 4 unnamed customers covering the full first-facility allocation plus 85MW at a second site (claims 14412, 14374, 14363). Only in the March 2022 filings were the four named: JointHash Holding (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 14232, 14210, 14200, 14071). These appear to be signed agreements. The Texas 200MW wind facility, by contrast, remains a letter of intent expressly subject to definitive agreements (claims 14204, 14074).

Concentration is the central diligence issue. GMR alone accounts for roughly 50% of the 100MW first facility (claims 14411, 14231, 14201). GMR, Bitmain and SparkPool are also strategic equity investors (claims 14505, 14377, 14378, 14376). Bitmain is simultaneously a shareholder, equipment supplier, hosting customer and, through its affiliate Antpool, joint-venture partner in 1.21 Gigawatts LLC (claims 14234, 14212, 14203). At least half the contracted capacity therefore sits with related or affiliated counterparties, and arm's-length demand is limited to F2Pool and Hashing LLC.

The revenue arithmetic cannot be reconciled from this record. No contract values, pricing per MW, term lengths or recognized hosting revenue are disclosed against the named agreements. The only dollar figures are a $22,000 proxy fee (claim 21068) and an $8,000 entry tied to the Antpool LLC formation (claim 14234), neither of which bears on customer revenue. The record supplied is also truncated after claim 14061, so later disclosures that might supply this reconciliation were not reviewed.

Applied Digital's record separates into three layers that should not be conflated. The 2006-2008 items predate the current business and come from the Flight Safety Technologies shell. One is a government development contract with the DOT Volpe Center (21053). The others are service vendors: an auditor, a proxy solicitor at roughly $22,000, and a transfer agent (21067, 21068, 21069). None of them is evidence of commercial demand. The 2021-2022 record is the substantive one, and several of its 'partnership' items are also supply-side or advisory rather than revenue-bearing. These include the Bitmain 200PH equipment order (14504, 14416), the SparkPool mining-pool advisory agreement (14418), the GMR asset-management services (14502), and the utility Energy Services Agreement (14233). The Coinmint co-hosting arrangement makes Applied the customer, not the vendor (14419). The Texas 200MW wind facility is explicitly a letter of intent subject to definitive agreements (14204, 14074). The 1.21 Gigawatts LLC with Antpool is a joint-venture formation, not a customer contract (14234, 14203).

The signed-customer disclosure sharpened over time. In August 2021 the company cited agreements with '2 customers' covering more than 50% of available energy without naming them (14506). It also said GMR and Bitmain held hosting contracts (14507). By November 2021 this had become '4 customers' covering the full first facility plus 85MW at a second, still unnamed (14412, 14374, 14363). Only in the March 2022 filings were the four identified: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (14232, 14210, 14200).

Concentration is high and partly related-party. GMR alone was described as holding roughly 50% of the 100MW first site (14411, 14231, 14201). GMR and Bitmain are also equity investors (14505, 14377, 14378), so at least half of contracted capacity appears to sit with shareholders. Bitmain also feeds hosting leads (14417) and is affiliated with the JV partner Antpool (14234). Diligence should treat these as arm's-length questions rather than independent demand signals.

On revenue arithmetic, the record contains no hosting revenue figures, per-MW pricing or contract values. The capacity claims (100MW first site, 85MW second) therefore cannot be reconciled with reported revenue from this material alone. The $8,000 figure attached to 14234 appears to concern JV formation rather than customer economics. The record supplied is also truncated after claim 14061, so later disclosures may change this picture.

Applied Digital's record splits into two unrelated eras. The 2006-2008 entries belong to its predecessor shell, Flight Safety Technologies. They comprise one government development contract with the DOT Volpe Center (claim 21053) and three vendor engagements: a transfer agent, an auditor, and Broadridge at roughly $22,000 (claims 21069, 21067, 21068). None of these are commercial customers of the current business, and they should carry no weight in assessing it. The supplied record is also truncated after claim 14061, so later disclosures are not assessed here.

The 2021-2022 hosting buildout shows a clear progression from unnamed to named counterparties. In August 2021 the company cited agreements with '2 customers' covering more than 50% of available energy, without identifying them (claim 14506). By November it cited '4 customers' covering the full first facility plus 85MW at a second, still unnamed (claims 14412, 14374, 14363). Only in March 2022 did filings name JointHash Holding (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 14232, 14210, 14200, 14071). For roughly seven months, readers were asked to credit customer counts without identities.

Many relationships described as partnerships are not customer revenue. SparkPool's mining-pool advisory and GMR's asset-management services (claims 14418, 14502, 14359) run toward Applied Digital. The 200PH Bitmain rig order (claims 14504, 14416, 14360) is a purchase. The Coinmint co-hosting arrangement (claims 14419, 14371) makes Applied Digital the hosted party. The Energy Services Agreement (claims 14233, 14202) is power supply. The Texas 200MW wind project is a letter of intent explicitly 'subject to entry into definitive agreements' (claims 14204, 14074). The Antpool joint venture, 1.21 Gigawatts LLC, is a framework vehicle disclosed with only $8,000 attached (claims 14234, 14212).

Concentration is severe and entangled. GMR alone holds about 50% of the first 100MW (claims 14411, 14201, 14211), and that figure is repeated verbatim across five filings without update. GMR, Bitmain and SparkPool are all equity investors (claims 14505, 14378, 14377). Bitmain is simultaneously shareholder, equipment supplier, customer and, through its Antpool affiliate, JV partner. At least half of contracted capacity therefore sits with related-party investors.

The revenue arithmetic cannot be reconciled from this record. No contract values, per-MW pricing, tenors or take-or-pay terms are disclosed. That leaves 185MW of claimed fill without any dollar figure to test against reported revenue.

Applied Digital's record splits into two unrelated eras, and only the second involves customers. The 2006-2008 entries come from the predecessor shell, Flight Safety Technologies. They consist of one government development contract with the DOT Volpe Center (claim 21053) and three service-provider engagements: Pacific Stock Transfer, Wolf & Company, and Broadridge (claims 21069, 21067, 21068). None of these bears on the hosting business, and the providers are vendors, not customers. From August 2021 onward, the company's language moves from anonymous to named. The August 2021 filing cites agreements with two unidentified customers covering more than half of available energy (claim 14506). November 2021 filings expand this to four customers, still unnamed (claims 14412, 14374, 14363). Only in March 2022 are the four identified: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 14232, 14210, 14200). GMR is consistently described as signed for about 50% of the 100MW first site (claims 14411, 14231, 14201). Several items read as relationships but are not revenue contracts. Coinmint is a co-hosting provider to Applied Digital, not a customer (claim 14419). The 200PH Bitmain order is an equipment purchase (claims 14504, 14416). SparkPool's role is advisory and investor-side (claims 14418, 14376). The Texas 200MW wind facility is explicitly a letter of intent, subject to definitive agreements (claims 14204, 14074). Concentration is the central diligence issue. GMR, Bitmain and SparkPool are all equity investors (claims 14505, 14378). GMR alone holds roughly half the first site, and Bitmain is both a customer and the source of hosting leads (claim 14417). Bitmain's affiliate Antpool is also the joint-venture partner in 1.21 Gigawatts LLC (claims 14234, 14212). At least two of the four named customers are therefore related to shareholders, and one investor cluster plausibly accounts for a majority of contracted megawatts. The disclosed figures cannot be reconciled with the announced deals. No contract value, hosting rate, term or revenue figure appears in any claim. The only dollar amounts are the $22,000 proxy fee (claim 21068) and the $8,000 figure tied to the Antpool LLC formation (claim 14234), which appears to be a formation-stage contribution rather than revenue. Megawatt allocations are therefore the only measurable commitment. The record provided ends partway through claim 14061, so later disclosures were not assessed.

Applied Digital's record mixes genuine hosting contracts with vendor relationships, insider-investor arrangements and one non-binding letter, so the customer count is smaller and more concentrated than the volume of claims suggests. The 2006-2008 items predate the current business. They come from the predecessor shell, Flight Safety Technologies, and cover a government research contract (21053) and service providers such as a transfer agent, an auditor and a proxy solicitor (21069, 21067, 21068). None of these is a customer of today's operation.

The only signed customer agreements are the hosting contracts for the first 100MW facility. Several other items run the other way. The Coinmint co-hosting deal (14419) makes Applied the customer. The 200PH Bitmain order (14504, 14416) is equipment Applied is buying. The SparkPool and GMR advisory roles (14418, 14502, 14359) are services provided to Applied. The Texas 200MW wind project is only a letter of intent, explicitly "subject to entry into definitive agreements" (14204, 14074). The 1.21 Gigawatts LLC (14234, 14212) is a joint-venture vehicle with Antpool targeting up to 1.5GW (14203), not a hosting contract.

On naming, customers stayed anonymous for about seven months. In August 2021 the company cited unnamed "2 customers" taking more than 50% of energy (14506). By November that had become "4 customers" for all 100MW plus 85MW at a second site (14412, 14374, 14363). The four were identified only in March 2022: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (14232, 14210, 14071).

That disclosure confirms heavy concentration with related parties. GMR and Bitmain are both equity investors (14505, 14378, 14377), and an earlier filing said those two alone would fill the first site (14507). GMR by itself holds roughly 50% of the 100MW (14411, 14231, 14201). Antpool, the joint-venture partner, is a Bitmain affiliate, which ties Bitmain's exposure to the growth plan as well. In practice, a single related-party group anchors both the base load and the expansion path.

The revenue arithmetic cannot be reconciled from this record. No hosting rates, contract terms, take-or-pay provisions or contract values are disclosed. The only dollar figures are $22,000 in proxy fees (21068) and $8,000 tied to the joint-venture formation (14234), which looks like a capital item rather than revenue. MW allocations cannot be turned into revenue without those terms. Note that the claim list supplied here ends partway through claim 14061, so any later disclosures are not reflected in this read.

Applied Digital's record splits into two unrelated eras, and only the second describes customers. The 2006-2008 items come from the predecessor shell, Flight Safety Technologies. They cover a Volpe Center development contract (claim 21053) and routine vendor engagements with an auditor, a proxy solicitor and a transfer agent (claims 21067, 21068, 21069). None of these bear on the current hosting business. In the 2021-2022 hosting era, the counterparties started out unnamed and were only identified later. In August 2021 the company cited agreements with '2 customers' covering more than half of available energy, without naming them (claim 14506). By November 2021 this had become '4 customers' covering all of the first facility and 85MW at a second (claims 14412, 14374, 14363), still unnamed. Only in the March 2022 filings were the four identified as JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (claims 14232, 14210, 14200, 14071). Those named hosting agreements, the Coinmint co-hosting arrangement (claims 14419, 14371, 14357) and the Antpool LLC agreement for 1.21 Gigawatts (claims 14234, 14212, 14203, 14073) read as executed contracts. The Texas 200MW wind site is explicitly only a letter of intent 'subject to entry into definitive agreements' (claims 14204, 14074). SparkPool's role is advisory and services work rather than hosting revenue (claims 14418, 14376, 14358). Concentration is severe and circular. GMR alone holds roughly 50% of the 100MW first site (claims 14411, 14231, 14201). GMR, Bitmain and SparkPool are also strategic equity investors (claims 14505, 14377, 14378). Bitmain supplies the miners (claims 14504, 14416), generates hosting leads (claim 14417), and sits on both sides of the Antpool joint venture through its affiliate. As a result, a material share of contracted capacity traces back to the company's own shareholders and suppliers rather than to arm's-length demand. On the revenue arithmetic, the claims provide no contract values, per-MW rates or terms, so the announced deals cannot be reconciled to reported revenue from this record. The capacity figures are internally consistent: GMR at about 50MW of 100MW is compatible with four customers absorbing the full 100MW. The Antpool figure of $8,000 (claim 14234) is ambiguous and should be checked against the filing to see whether it is in thousands. The supplied record also cuts off mid-claim at 14061, so later customer disclosures, including any HPC or AI-datacenter leases, are not assessed here.

Applied Digital's record splits into two unrelated eras, and only the second describes customers. The 2006-2008 entries belong to the predecessor shell, Flight Safety Technologies. The Volpe Center award (claim 21053) was a genuine government contract, but it was a development-and-demonstration engagement. The auditor, transfer-agent and Broadridge proxy items (claims 21067, 21069, 21068) are vendor relationships, and the $22,000 figure is a cost, not revenue. None of them bears on the hosting business.

The 2021 hosting narrative started with unnamed counterparties and named them only later. In August 2021 the company said it had agreements with "2 customers" covering more than half of available energy (claim 14506). A separate passage in the same filing said GMR and Bitmain contracts would pre-fill the first 100MW site (claim 14507), and the two statements do not obviously describe the same scope. By November the unnamed count had grown to four customers covering the full first facility plus 85MW at a second, unidentified site (claims 14412, 14374, 14363). The four were named only in March 2022: JointHash (a GMR subsidiary), Bitmain, F2Pool and Hashing LLC (claims 14232, 14210, 14200). That disclosure converts earlier descriptive language into identifiable signed agreements, but four months passed before it did.

Concentration is high and intertwined with related parties. GMR alone is described as holding about 50% of the 100MW (claims 14411, 14231, 14201). GMR, Bitmain and SparkPool are all equity investors (claims 14505, 14376-14378), and Bitmain is simultaneously an equipment supplier (claims 14504, 14416), a source of customer leads (claim 14417), and, through its affiliate Antpool, a joint-venture partner in 1.21 Gigawatts LLC (claims 14234, 14203). Effectively two investor-linked groups anchor the first site. SparkPool's mining-pool advisory role (claim 14418) and the Coinmint co-hosting arrangement (claim 14419) are supplier relationships, not customer ones. The 200MW Texas wind project is expressly a letter of intent "subject to entry into definitive agreements" (claims 14204, 14074) and should not be counted as contracted.

The revenue arithmetic cannot be reconciled from this record. No contract values, per-megawatt pricing, terms or revenue recognition schedules are disclosed for any hosting agreement. The only dollar figure attached to the hosting-era counterparties is an unexplained $8,000 tied to the Antpool LLC formation (claim 14234), whose units and nature are not stated. Capacity claims (100MW plus 85MW) are therefore uncorroborated by any dollar commitment, and the supplied record ends in April 2022 (claim 14060), so any later conversion of these agreements is outside this read.

Applied Digital's customer record splits into two unrelated eras, and only the second describes the current business. The 2006-2008 entries come from the predecessor Flight Safety Technologies. The Department of Transportation Volpe Center award (claim 21053) is the only genuine third-party contract in that period. The auditor, transfer agent and Broadridge proxy items (claims 21067, 21069, 21068) are service vendors, not customers, and the $22,000 figure is an expense.

The hosting pivot shows identification arriving late. In August 2021 the company cited agreements with two customers covering more than half of available energy without naming either (claim 14506). By November that became four unnamed customers absorbing all of the first facility plus 85MW at a second site (claims 14412, 14374, 14363). Names appeared only in March 2022: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (claims 14232, 14210). Even then, the language stayed prospective: the agreements "will account for" energy at a facility still under development (claim 14411).

Concentration is the central diligence issue. GMR alone holds roughly 50% of the 100MW site (claims 14411, 14231, 14201), leaving the other three customers to share the remainder in undisclosed proportions. The customer base also overlaps heavily with the company's own investors. GMR, Bitmain and SparkPool are all strategic equity investors (claims 14505, 14376). Bitmain is at once a shareholder, equipment supplier on the 200PH order (claims 14504, 14416), customer, and source of hosting leads (claim 14417). Its affiliate Antpool is the company's joint-venture partner (claim 14234). Revenue concentrated in shareholder-affiliated counterparties deserves related-party scrutiny rather than being treated as arm's-length demand.

Several headline items are framework rather than contract. The Texas 200MW wind project is a letter of intent expressly subject to definitive agreements (claims 14204, 14074). The 1.21 Gigawatts LLC with Antpool targets up to 1.5GW, yet its only disclosed figure is $8,000 (claims 14234, 14203), which reads as a formation contribution, not committed capital. The SparkPool mining-pool arrangement is advisory and conditional on a future decision (claim 14418). Coinmint (claim 14419) is a facility the company pays for hosting, so it is a cost, not a revenue source.

The revenue arithmetic cannot be reconciled from this record. No contract values, per-megawatt pricing, terms or recognized revenue are disclosed for any hosting agreement. The capacity claims are therefore unverifiable rather than inconsistent. The record supplied here ends in April 2022, so later conversions or terminations are not reflected.

Applied Digital's disclosed counterparty record is short and inward-looking, and the supplied history ends in April 2022, partway through claim 14061. The 2006 to 2008 entries (21053, 21067, 21068, 21069) belong to the predecessor shell, Flight Safety Technologies. Apart from the Volpe Center research contract, they record vendors such as an auditor, a proxy solicitor and a transfer agent rather than customers, so they carry no weight for commercial diligence. The relevant history begins in August 2021, when the company said it had agreements with two customers covering more than 50% of its contracted energy (14506). The same filing names GMR and Bitmain as hosting counterparties (14507), so the identification is partial rather than absent. By November 2021 the count had grown to four customers covering the full 100MW first facility and 85MW at a second site (14412, 14374, 14363). None of the four was named until March 2022, when the filing identified JointHash Holding Limited, a GMR subsidiary, along with Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (14232, 14200, 14210). Four months of unnamed customers is a gap diligence readers should note, though the counterparties were eventually disclosed. The firmer language attaches to these hosting agreements and to the Amended and Restated Energy Services Agreement (14233, 14202). The softer items include the Texas 200MW wind facility, which was a letter of intent expressly subject to definitive agreements (14204, 14074). They also include the SparkPool advisory arrangement (14418) and the Coinmint co-hosting agreement, under which Applied Digital is the customer rather than the provider (14419). Concentration is the central issue. GMR alone holds roughly 50% of the first site's capacity (14411, 14231, 14201). GMR and Bitmain are also strategic equity investors (14505, 14377, 14378). Bitmain is additionally the equipment supplier (14504, 14416), a source of customer leads (14417), and, through its affiliate Antpool, the joint-venture partner in 1.21 Gigawatts LLC (14234, 14212). Customer, shareholder, supplier and JV partner therefore overlap heavily, which makes these agreements closer to related-party arrangements than arm's-length demand. The revenue arithmetic cannot be reconciled from this record. No contract values, hosting rates, terms or recognised revenue are disclosed alongside the capacity allocations. The only dollar figure attached to the Antpool JV (14234) is an ambiguous $8,000, so the announced megawatts cannot be tested against reported income.

The record splits into two unrelated eras, and only the second describes Applied Digital's current business. The 2006–2008 entries come from the predecessor shell, Flight Safety Technologies. The Volpe Center award (21053) was a genuine signed government contract. The other three are vendors the company paid, not customers who paid it: the transfer agent (21069), the auditor (21067) and Broadridge's roughly $22,000 proxy fee (21068). None of them bears on hosting-customer diligence.

The 2021–2022 hosting disclosures show a clear progression from unnamed to named. In August 2021 the company cited agreements with "2 customers" for more than 50% of available energy (14506). The same filing identified those two as GMR and Bitmain (14507). By November 2021 the count had grown to "4 customers" covering the entire first facility plus 85MW at a second, still unnamed (14412, 14374, 14363). The four were not identified until the March 2022 filings: JointHash (a GMR subsidiary), Bitmain, F2Pool and Hashing LLC (14232, 14210, 14200). For roughly four months, then, the full-capacity claim rested on undisclosed counterparties.

Concentration is high, and it overlaps heavily with related parties. GMR alone is said to hold about 50% of the 100MW site (14411, 14231, 14201). GMR and Bitmain are also equity investors (14505, 14377, 14378). Bitmain plays several roles at once: equipment supplier (14504, 14416), source of customer leads (14417), hosting customer, and parent of Antpool. Antpool is the company's partner in the 1.21 Gigawatts LLC joint venture (14234, 14203). As a result, at least two of the four named customers, and arguably half the contracted energy, sit with insiders rather than arm's-length buyers.

Several items are not customer agreements. The Texas 200MW wind project is expressly a letter of intent "subject to entry into definitive agreements" (14204, 14074). SparkPool is an advisor and investor, not a customer (14418, 14358). In the Coinmint arrangement, Applied Digital is the one being hosted (14419). The phrase "agreements… which will account for" the available energy describes planned capacity allocation, not contracts already producing revenue.

On revenue arithmetic, no revenue, contract value or per-MW pricing figure appears anywhere in this record, so the announced deals cannot be checked against reported results. The only dollar figures are the proxy fee and the joint-venture capitalization (14234). The claim list also cuts off at claim 14060, which means later disclosures, including any shift away from crypto hosting, are not covered by this read.

Applied Digital's customer record before 2021 belongs to its predecessor, Flight Safety Technologies, and has no bearing on the hosting business. The only revenue counterparty from that era is a Department of Transportation Volpe Center development contract (claim 21053). Broadridge, Wolf & Company and Pacific Stock Transfer are vendors, not customers (claims 21068, 21067, 21069).

The hosting story begins in August 2021, when the company said it had agreements with two customers covering more than 50% of available energy but did not name them (claim 14506). In the same filing it described GMR and Bitmain as having hosting contracts that would pre-fill the first 100MW site (claim 14507), so the two unnamed customers were almost certainly those strategic investors. By November 2021 the count had risen to four customers. They were said to cover all energy at the first facility and 85MW at a second, but they were again unidentified (claims 14412, 14374, 14363). Only in March 2022 were the four named: JointHash Holding (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 14232, 14200, 14210). Throughout, GMR is the only counterparty given a sized commitment, at roughly 50% of the 100MW (claims 14411, 14201).

Concentration is the central diligence issue. GMR, Bitmain and SparkPool are all equity investors (claims 14505, 14377, 14378). The January 2022 1.21 Gigawatts joint venture is with Antpool, a Bitmain affiliate (claims 14212, 14234). That means at least half the contracted capacity, and probably more, sits with shareholders or their affiliates. F2Pool and Hashing LLC are the only named arm's-length customers, and their share is undisclosed. SparkPool's arrangements are advisory and services agreements, not customer contracts (claims 14418, 14376). Coinmint is a co-hosting supplier to the company, not a buyer (claim 14419).

The softer language deserves its own category. The Texas 200MW wind facility is a letter of intent expressly subject to definitive agreements (claims 14204, 14074). The Antpool vehicle is a framework for up to 1.5GW and not a revenue contract (claim 14203). The 85MW second-facility allocation has no corresponding energy agreement in the record, unlike the 100MW Amended and Restated Energy Services Agreement (claims 14233, 14202).

The revenue arithmetic cannot be reconciled from this record. None of the claims disclose contract pricing, term, take-or-pay provisions or recognized hosting revenue, so the 'total available energy' assertions cannot be tested against reported results. The claim set as supplied is also truncated after claim 14061, so any later customer disclosures were not reviewed.

Applied Digital's record splits into two unrelated eras, and only the second is relevant. The 2006–2008 entries reflect the predecessor Flight Safety Technologies. One is a government development contract with the DOT Volpe Center (claim 21053). The rest are vendor appointments: Broadridge as proxy solicitor at roughly $22,000 (21068), Wolf & Company as auditor (21067) and Pacific Stock Transfer as transfer agent (21069). None of these is a customer relationship bearing on the hosting business. In the 2021–2022 hosting era, the language moves from anonymous to named, and the sequence matters. In August 2021 the company cited agreements with two unidentified customers covering more than half of available energy (14506). The same filing named GMR and Bitmain as hosting counterparties (14507), so the two are probably the same parties, though the filing never says so. By November 2021 the count became four customers absorbing all of the first 100MW facility and 85MW at a second site, still unnamed (14412, 14374, 14363). Only in March 2022 were the four identified: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (14232, 14200, 14210, 14071). These read as signed agreements. Other items are not customer contracts. The Texas 200MW wind facility is a letter of intent expressly subject to definitive agreements (14204, 14074). SparkPool's role is advisory and supply, not purchasing (14418, 14358). In the Coinmint co-hosting arrangement Applied Digital was itself the customer (14419). Concentration is the central diligence issue. GMR alone holds roughly 50% of the first 100MW (14411, 14201, 14231). GMR and Bitmain are also strategic equity investors (14505, 14377, 14378). Bitmain is additionally the equipment supplier on a 200PH order (14504, 14416), a source of hosting leads (14417), and, through its affiliate Antpool, joint-venture partner in 1.21 Gigawatts LLC (14234, 14212). That JV is described as targeting up to 1.5GW, yet the filing ties an $8,000 figure to it, which suggests nominal initial capitalization. Half or more of contracted capacity therefore sits with shareholders who also supply the company and partner with it, which weakens the independence of the demand signal. On revenue, the claims on record disclose no contract values, per-MW pricing, terms or recognized revenue for any hosting agreement. Without those figures the announced deals cannot be reconciled against reported revenue; full allocation of 100MW plus 85MW describes capacity, not dollars. The record supplied is also truncated after claim 14061, so later conversions or terminations of these agreements cannot be assessed here.

Applied Digital's record splits into two unrelated eras, and only the second bears on commercial diligence. The 2006-2008 entries come from the predecessor, Flight Safety Technologies. They comprise a Volpe Center development contract (21053) and ordinary vendor engagements: Broadridge for proxy solicitation at roughly $22,000 (21068), Wolf & Company as auditor (21067) and Pacific Stock Transfer as transfer agent (21069). None of the vendor engagements are customer relationships. The hosting business that begins in 2021 is the substance. Its disclosure pattern moved from anonymous to named over about seven months. In August 2021 the company cited agreements with two unidentified customers covering more than 50% of available energy (14506), while separately naming GMR and Bitmain as hosting contract holders (14507). By November 2021 the count had become four customers absorbing all energy at the first facility plus 85MW at a second, still unnamed (14412, 14374, 14363). Only in March 2022 were the four identified as JointHash Holding (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (14232, 14200, 14210, 14071). Those are the only signed customer agreements with named counterparties on the record. Concentration is acute and reflexive. GMR alone holds roughly 50% of the 100MW first site (14411, 14231, 14201, 14072). GMR, Bitmain and SparkPool are all described as strategic equity investors (14505, 14376, 14377, 14378). Bitmain is simultaneously a customer, the equipment supplier on a 200PH order (14504, 14416), a source of hosting leads (14417), and, through its affiliate Antpool, the JV partner in 1.21 Gigawatts LLC (14234, 14212, 14203). Customer, shareholder, supplier and co-venturer therefore overlap substantially. That overlap warrants scrutiny of arm's-length pricing. Several items should not be read as customer wins. The Texas 200MW wind project is a letter of intent expressly subject to definitive agreements (14204, 14074). The SparkPool arrangements are advisory and services relationships (14418, 14358). In the Coinmint co-hosting agreement, Applied Digital is the customer rather than the provider (14419). The Energy Services Agreement is a power-supply contract with an unidentified utility (14233, 14202). The record does not let the revenue arithmetic be reconciled. No contract values, per-MW pricing, tenor or recognized hosting revenue are disclosed, only shares of capacity. The figure attached to the Antpool JV (14234) appears without units or context. The claims list also ends mid-entry at 14061, so later disclosures may change this picture.

Applied Digital's record splits into two unrelated eras. The 2006–2008 claims belong to the predecessor Flight Safety Technologies, and only one of them is a customer contract: the Department of Transportation Volpe Center award (claim 21053). The others are service vendors, namely an auditor, a proxy solicitor and a transfer agent (claims 21067, 21068, 21069), and should not be counted as commercial traction. The relevant history begins in August 2021. At that point the company said it had agreements with two customers covering more than 50% of available energy, without naming them (claim 14506). The same filing identified GMR and Bitmain as hosting counterparties for the first 100MW site (claim 14507), and both were already strategic equity investors (claim 14505). By November 2021 the count had grown to four customers covering the full first-facility allocation plus 85MW at a second site, still unnamed (claims 14412, 14374, 14363). Names arrived only in March 2022: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 14232, 14200, 14210). That disclosure confirms the agreements are signed, but it also shows concentration is high and largely related-party. GMR alone holds roughly 50% of the 100MW capacity (claims 14411, 14231, 14201). Two of the four customers are equity holders. The January 2022 1.21 Gigawatts joint venture is with Antpool, a Bitmain affiliate (claims 14234, 14212), so the Bitmain group sits on the investor, supplier, customer and JV-partner sides at once (claims 14417, 14416, 14378). Several items are softer than their headlines suggest. The Texas 200MW wind facility is a letter of intent that remains subject to definitive agreements (claims 14204, 14074). The SparkPool arrangement is an advisory and services relationship, not a hosting contract (claims 14418, 14376), and the record cuts off mid-sentence on what SparkPool is 'currently in the process of' (claim 14358). Coinmint runs the other way: Applied Digital is the co-hosted customer there, not the provider (claim 14419). On revenue, the record discloses megawatt allocations and contract counts but no contract values, pricing, terms or recognized revenue. The only dollar figures are $22,000 in proxy fees and the $8,000 JV formation item. The announced deals therefore cannot be reconciled against revenue from this record. The claim list provided is also truncated at claim 14060, so later customer disclosures are not reflected in this read.

Applied Digital's record shows a small set of named hosting counterparties, and most of them are also its investors or suppliers. The pre-2021 entries come from the Flight Safety Technologies predecessor. They cover a Volpe Center contract (21053) and service vendors, namely an auditor, a proxy solicitor and a transfer agent (21067, 21068, 21069). None of these is a customer of the current business and none should be counted as commercial traction. In August 2021 the company described agreements with two unnamed customers covering more than 50% of available energy (14506). In the same filing it said GMR and Bitmain hosting contracts would pre-fill the first 100MW site (14507). That is a fuller claim than the 50% figure, and the two statements are hard to reconcile. By November 2021 the count had grown to four unidentified customers taking all of the first facility's energy plus 85MW at a second site (14412, 14374, 14363). The four were named only in March 2022: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (14232, 14200, 14071). GMR alone holds roughly 50% of the 100MW capacity (14411, 14231). Concentration is therefore high and largely related-party. GMR, Bitmain and SparkPool are each described as strategic equity investors (14505, 14378). The 1.21 Gigawatts LLC joint venture is with Antpool, a Bitmain affiliate (14234, 14203). F2Pool and Hashing LLC are the only named customers with no disclosed equity or supply tie. Several other items describe vendor or advisory relationships rather than revenue. These include the 200PH Bitmain equipment order (14504, 14416), the Coinmint co-hosting agreement, in which Applied Digital is the customer (14419), and the SparkPool and GMR advisory arrangements (14418, 14359). The Texas 200MW wind facility is only a letter of intent, subject to definitive agreements (14204, 14074), and should not be treated as contracted capacity. The SparkPool quote in 14358 is truncated mid-sentence, so its status cannot be assessed from this record. The revenue arithmetic cannot be reconciled. No contract values, per-MW pricing, term lengths or recognized hosting revenue are disclosed in these claims. The only dollar figures are a $22,000 proxy fee (21068) and an $8,000 figure tied to the Antpool LLC (14234), and neither is customer revenue. The record supplied here also ends partway through claim 14060, so any later disclosures are not reflected in this read.

The record splits into two unrelated eras, and only the second bears on Applied Digital's customer base. The 2006-2008 entries belong to the predecessor shell Flight Safety Technologies. The Volpe Center award (claim 21053) was a genuine government contract, but the other three items are vendors, not customers: an auditor, a proxy solicitor engaged at about $22,000, and a transfer agent (claims 21067, 21068, 21069). None of them says anything about the hosting business. From 2021 onward, the disclosure moved from anonymous to named counterparties, but slowly. In August 2021 the company cited agreements with two unidentified customers covering more than half the available energy (claim 14506). By November that had become four customers absorbing all of the first facility's energy plus 85MW at a second site, still unnamed (claims 14412, 14374, 14363). Only in March 2022 were the four identified: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (claims 14232, 14200, 14210, 14071). Those are signed hosting agreements, but the record discloses no tenor, pricing or take-or-pay terms. Several other relationships are not customer revenue at all. Bitmain's 200PH order is an equipment purchase by the company (claims 14504, 14416). Coinmint hosted the company's own rigs, so Applied Digital was the customer there (claims 14419, 14357). SparkPool is an adviser and equity investor (claims 14418, 14376). The Texas 200MW wind project is only a letter of intent, explicitly subject to definitive agreements (claims 14204, 14074). The Antpool joint venture is a framework vehicle for up to 1.5GW, not a contract for capacity (claims 14234, 14203). Concentration is the central diligence issue. GMR alone holds roughly 50% of the 100MW first facility (claims 14411, 14231, 14072). GMR and Bitmain are also strategic equity investors (claims 14505, 14378), and Antpool is a Bitmain affiliate (claim 14212). That means at least half the identified customer slate, and most of the contracted megawatts, sits with shareholders or their affiliates, which makes these effectively related-party relationships. On revenue, the record does not support any reconciliation. No contract values, per-megawatt rates or revenue figures accompany the announced deals. The only dollar amounts disclosed are the $22,000 proxy fee and an $8,000 figure attached to the 1.21 Gigawatts entity (claim 14234), and neither bears on hosting economics. The record supplied also ends mid-entry at claim 14060 (April 2022), so later contract developments are not assessed here.

Applied Digital's record mixes signed hosting agreements, vendor relationships and explicitly non-binding arrangements, and the sorting matters. The 2006-2008 entries belong to the predecessor shell, Flight Safety Technologies. Only the Volpe Center award (claim 21053) was a revenue contract. The auditor, proxy solicitor and transfer agent (claims 21067, 21068, 21069) are service providers, and none of these bears on the current business.

The hosting book went from anonymous to named in stages. In August 2021 the company described agreements with two unidentified customers covering more than 50% of available energy (claim 14506). In the same filing it said GMR and Bitmain contracts would fill the first 100MW site (claim 14507), an internal tension between a majority and the whole. By November 2021 the count was four customers taking all of the first facility plus 85MW at a second, still unnamed (claims 14412, 14374, 14363). Only in March 2022 were the four identified: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (claims 14232, 14200, 14071).

Concentration is severe and largely related-party. GMR alone holds roughly 50% of the 100MW (claims 14411, 14231, 14201). Two of the four customers are also strategic equity investors (claim 14505). Bitmain is at once investor, equipment supplier for a 200PH order (claims 14504, 14416), lead source (claim 14417), customer, and, through its affiliate Antpool, joint-venture partner (claims 14234, 14212). Revenue from such counterparties warrants arm's-length scrutiny. SparkPool's role is advisory, not customer (claims 14418, 14376). The Coinmint co-hosting deal makes Applied Digital the customer and is a cost line (claim 14419).

Several headline items fall short of commercial commitments. The Texas 200MW wind facility is a letter of intent expressly subject to definitive agreements (claims 14204, 14074). The 1.21 Gigawatts LLC is a signed formation agreement, but the scale is aspirational: the claim title cites $8,000 (claim 14234), against a stated ambition of up to 1.5GW (claim 14203). The Energy Services Agreement is supply-side power, not demand (claims 14233, 14202).

On revenue arithmetic, the record does not reconcile because nothing in it can be reconciled. No claim discloses contract value, pricing per megawatt, term or minimum take. "Account for the total available energy" is a capacity allocation, not a dollar commitment. Until per-customer economics are disclosed, the signed hosting book supports a statement about megawatts spoken for, not revenue secured.

Applied Digital's record spans two unrelated eras. The 2006–2008 claims belong to the predecessor shell, Flight Safety Technologies. That era has one genuine customer award, the Department of Transportation Volpe Center wake-vortex emulation contract (21053). The rest are service-provider engagements: auditor Wolf & Company (21067), transfer agent Pacific Stock Transfer (21069) and Broadridge proxy solicitation at roughly $22,000 (21068). These are vendor relationships, not customers, and carry no commercial weight.

The 2021–2022 hosting pivot is where diligence matters. In August 2021 the company said it had agreements with two unnamed customers covering more than 50% of available energy (14506), and it separately named GMR and Bitmain as hosting contractors (14507). By November the count had grown to four unidentified customers taking the full first-facility energy plus 85MW at a second site (14412, 14374, 14363). The four were named only in March 2022: JointHash Holding (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (14232, 14210, 14200, 14071). For roughly four months, "agreements" with four customers were asserted without identification.

Concentration is the more consequential finding. GMR alone is described as contracting for about 50% of the 100MW first facility (14411, 14231, 14201). GMR, Bitmain and SparkPool are also strategic equity investors (14505, 14377, 14378). Bitmain additionally supplies equipment (14416, 14504) and hosting leads (14417), and its affiliate Antpool is the joint-venture partner in 1.21 Gigawatts LLC (14234, 14203). Half or more of first-facility capacity therefore sits with shareholders who also sell to or advise the company, which looks more like related-party economics than arm's-length demand. SparkPool's role is advisory (14418, 14358), not that of a hosting customer.

Framework language should be discounted. The Texas 200MW wind facility is a letter of intent expressly "subject to entry into definitive agreements" (14204, 14074). The Antpool LLC is a joint-venture vehicle, not a revenue contract, and the Energy Services Agreement (14233, 14202) secures power supply rather than customer demand.

The revenue arithmetic cannot be reconciled from this record. No contract value, hosting rate or term length is disclosed for any hosting agreement. The only dollar figures, $22,000 to Broadridge and an $8,000 amount tied to the 1.21 Gigawatts formation (14234), are immaterial. Capacity allocations in megawatts are stated, but pricing is not. The claim set supplied ends mid-record at 14059, so later developments are not assessed here.

Applied Digital's record falls into two distinct eras, and only the second involves customers. The 2006-2008 entries, from the predecessor Flight Safety Technologies, include a single genuine revenue counterparty, the DOT Volpe Center development contract (claim 21053). The rest are service providers the company paid: an auditor, a transfer agent, and Broadridge at roughly $22,000 for proxy solicitation (claims 21067, 21069, 21068). None of those should be read as commercial traction.

The hosting pivot shows a clear progression from anonymous to named counterparties. In August 2021 the company said it had agreements with two customers covering more than half of available energy, without naming them (claim 14506), although the same filing identifies GMR and Bitmain as hosting contract holders (claim 14507). By November 2021 the count had risen to four customers covering the full first-facility allocation plus 85MW at a second site, again unnamed (claims 14412, 14374, 14363). Not until March 2022 were the four identified as JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 14232, 14210). These read as signed agreements, as does the Amended and Restated Energy Services Agreement for 100MW (claims 14233, 14202). By contrast, the 200MW Texas wind facility is explicitly a letter of intent "subject to entry into definitive agreements" (claims 14204, 14074). The SparkPool pool-advisory arrangement (claim 14418) and the Coinmint co-hosting deal (claim 14419), where Applied Digital is itself the customer, are not customer revenue.

Concentration is the central issue. GMR alone holds about 50% of the 100MW site (claims 14411, 14231). GMR and Bitmain are also strategic equity investors (claim 14505). Bitmain is simultaneously a customer, the equipment supplier on the 200PH order (claims 14504, 14416), a source of hosting leads (claim 14417), and, through its affiliate Antpool, the company's JV partner in 1.21 Gigawatts LLC (claim 14234). In practice, two investor-affiliated groups appear to account for most of the contracted first-site capacity. That makes the related-party character of the book at least as material as its size.

Revenue reconciliation cannot be performed from this record. No contract values, pricing per MW, terms or recognized revenue are disclosed against any named hosting deal. The only dollar figures are the Broadridge fee and the $8,000 associated with forming the Antpool vehicle (claim 14234), and neither is customer revenue. Readers should treat the capacity-allocation claims as contracted demand of unverified economic value. The claim list provided ends mid-entry at claim 14059, so later disclosures were not assessed.

The 2006 to 2008 entries belong to Flight Safety Technologies, the shell that preceded Applied Digital, and only one of them is a customer relationship. The Department of Transportation Volpe Center award (21053) is a genuine government contract for a defined development scope. The auditor (21067), transfer agent (21069) and Broadridge proxy engagement at roughly $22,000 (21068) are vendors the company pays, not counterparties that generate revenue. The commercially relevant record begins in August 2021, and it should be read as a progression from unnamed to named counterparties. That month the company said it had agreements with two customers covering more than 50% of available energy but did not identify them (14506), while separately naming GMR and Bitmain as hosting contract parties (14507). By November the count had risen to four customers covering the full first facility plus 85MW at a second, still unnamed (14412, 14374, 14363). Only in March 2022 were all four identified: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (14232, 14210, 14200, 14071). Several other items are relationships with suppliers, advisors or hosts rather than customer agreements. The Bitmain order for 200PH of rigs (14504, 14416) is a purchase. The Coinmint co-hosting arrangement (14419) makes Applied Digital the customer. The SparkPool arrangement is advisory (14418). The Texas wind-power facility is explicitly a letter of intent, 'subject to entry into definitive agreements' (14204, 14074), and should not be counted as contracted capacity. Counterparty concentration is acute and circular. GMR, through JointHash, holds a contract for approximately 50% of the 100MW site (14411, 14231, 14201). GMR and Bitmain are also strategic equity investors (14505, 14377, 14378), and Bitmain's affiliate Antpool is the company's partner in the 1.21 Gigawatts LLC (14234, 14203). In practice, two of the four named customers are shareholders or affiliates of shareholders, so the customer base carries related-party character. The disclosed revenue arithmetic cannot be reconciled because the record contains no hosting rates, contract terms, minimum commitments or dollar values. The only figure tied to these deals is the $8,000 attached to the Antpool LLC formation (14234), which reads as a nominal capital contribution rather than a revenue commitment. Megawatt allocations (100MW plus 85MW) are capacity claims, not revenue, and investors should treat them as unpriced. The record supplied ends mid-claim at 14059 (April 2022), so later developments are not reflected in this read.

Applied Digital's record splits into two unrelated eras, and only the second bears on commercial diligence. The 2006 to 2008 entries reflect the predecessor Flight Safety Technologies. The Volpe Center award (21053) was a genuine government contract. The transfer agent, proxy solicitor and auditor engagements (21069, 21068, 21067) are vendor relationships rather than customers and should carry no weight. The hosting pivot begins in August 2021, and its counterparty disclosure tightened over the following eight months. The first filings named GMR and Bitmain as hosting contract parties (14507) but described the core book only as "agreements with 2 customers" covering more than half of available energy (14506), with no identification. By November 2021 the count had grown to four unnamed customers absorbing all energy at the first facility plus 85MW at a second site (14412, 14374, 14363). Not until March 2022 were the four named: JointHash Holding Limited, a GMR subsidiary; Bitmain Technologies Limited; F2Pool Mining; and Hashing LLC (14232, 14200, 14210, 14071). Those named agreements are the strongest evidence of signed customer contracts. Even so, the filings describe them as agreements that "will account for" capacity at a facility still "under development" (14231), not as revenue-generating service. The concentration picture is more troubling than the customer count suggests. GMR alone holds roughly 50% of the 100MW first site (14411, 14201). GMR, Bitmain and SparkPool are each strategic equity investors (14505, 14377). Bitmain is at once an equity holder, an equipment supplier with a 200PH order (14504, 14416), a source of customer leads (14417), a named hosting customer, and, through its affiliate Antpool, a joint venture partner in 1.21 Gigawatts LLC (14234, 14203). At least two of the four anchor customers therefore sit inside the shareholder base, which weakens any claim of arm's-length demand. Several items are not customer agreements at all. The Texas 200MW wind facility is a letter of intent expressly "subject to entry into definitive agreements" (14204, 14074). The Coinmint co-hosting arrangement makes Applied Digital the customer, not the provider (14419). The SparkPool and GMR advisory arrangements are services the company receives (14418, 14502). On revenue, the record discloses no contract values, pricing per megawatt, term lengths or recognized hosting revenue. The only dollar figures are the $22,000 proxy fee and the $8,000 joint venture item, so the announced deals cannot be reconciled to revenue from this record. Note that the claim list supplied here is truncated at claim 14060, so later filings may change this assessment.

Applied Digital's record splits into two unrelated eras. The 2006-2008 claims belong to the predecessor shell, Flight Safety Technologies. Only the Department of Transportation Volpe Center award (claim 21053) is a genuine signed customer contract. The auditor, transfer-agent and Broadridge items (21067, 21069, 21068) are vendor engagements, and the $22,000 figure is a cost, not revenue. The substantive history begins in August 2021, and there the language matters more than the volume of repetition suggests.

In August 2021 the company described unnamed hosting customers, first as '2 customers' taking more than 50% of available energy (14506). By November that became '4 customers' covering the full first facility plus 85MW at a second site (14412, 14374, 14363). The four were not identified until March 2022: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (14232, 14210, 14200, 14071). These read as signed agreements. However, the 85MW allocation attaches to a facility not yet built, and the stated aim was to 'pre-fill' capacity 'before breaking' ground (14507). That makes them forward capacity commitments, not operating revenue.

Concentration is the central diligence issue. GMR alone holds roughly 50% of the 100MW first facility (14411, 14377, 14231, 14201). GMR and Bitmain, two of the four named customers, are also strategic equity investors (14505, 14378). Bitmain is additionally the equipment supplier (14504, 14416) and a source of hosting leads (14417). Its affiliate Antpool is the company's joint-venture partner in 1.21 Gigawatts LLC (14234, 14212, 14073). The customer base is therefore substantially related-party, and arm's-length demand is limited to F2Pool and Hashing LLC.

Several items are not customer agreements at all. The Texas 200MW wind project is a letter of intent expressly 'subject to entry into definitive agreements' (14204, 14074). The SparkPool and GMR arrangements are advisory services the company receives (14418, 14358, 14359, 14502). The Coinmint co-hosting deal makes Applied Digital the customer, not the vendor (14419, 14371, 14357).

Revenue arithmetic cannot be reconciled from this record. No contract value, per-megawatt rate, term or recognized revenue is disclosed. The only dollar figures are the $8,000 JV capitalization (14234) and the predecessor's proxy fee (14068). Capacity arithmetic is internally consistent: GMR at about 50%, and four customers at 100% of 100MW. The supplied record ends mid-claim at 14061 (April 2022), so later customer developments are outside this read.

Applied Digital's record mixes signed hosting agreements with vendor relationships, equity tie-ups and one explicit letter of intent, so the customer base looks broader than it is. The pre-2021 entries come from the Flight Safety Technologies predecessor. Most of them are service providers rather than customers: a transfer agent, an auditor, and a proxy solicitor paid about $22,000 (claims 21069, 21067, 21068). The 2006 Volpe Center award (claim 21053) is an unrelated legacy government contract. The hosting story begins in August 2021, when the company said it had agreements with two unidentified customers covering more than half of available energy (claim 14506). It separately named GMR and Bitmain as hosting counterparties (claim 14507). By November 2021 this had grown to four customers taking all first-facility energy and 85MW at a second site, still unnamed (claims 14412, 14374, 14363). The four were not identified until March 2022: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (claims 14232, 14200, 14210). Several other relationships read like partnerships but put Applied Digital on the buying side. These include the Bitmain 200PH equipment order, the Coinmint co-hosting arrangement and the SparkPool advisory agreement (claims 14504, 14419, 14418). The Texas 200MW wind project is only a letter of intent, subject to definitive agreements (claim 14074). Concentration is the bigger issue. GMR alone holds about 50% of the 100MW first facility (claims 14411, 14231). GMR, Bitmain and SparkPool are also strategic equity investors (claims 14505, 14378). The January 2022 1.21 Gigawatts joint venture is with Antpool, a Bitmain affiliate (claims 14234, 14203). Taken together, the GMR and Bitmain groups appear to control most contracted capacity while also holding equity and supplying equipment. Arm's-length demand from F2Pool and Hashing LLC looks minor by comparison, though the record does not split the MW allocation among the four. The revenue arithmetic cannot be reconciled. None of the claims on record give a contract value, price per MW, term length or revenue figure for any hosting agreement. Capacity is described only in megawatts and percentages, and the only dollar amounts relate to the Antpool LLC formation and proxy fees (claims 14234, 14068). The claim list supplied also cuts off at claim 14059, so later filings that might add pricing or revenue disclosure could not be reviewed.

Applied Digital's record splits into two unrelated eras. The 2006-2008 entries belong to the predecessor shell, Flight Safety Technologies: a Volpe Center contract (claim 21053) and routine vendor appointments such as a transfer agent, Broadridge's roughly $22,000 proxy fee, and an auditor (claims 21069, 21068, 21067). None of these bears on the current hosting business. The relevant history begins in August 2021.

The signed hosting base was first disclosed without names. In August 2021 the company cited agreements with "2 customers" covering more than 50% of available energy (claim 14506). By November that had become "4 customers" covering the full first facility and 85MW of a second (claims 14412, 14374, 14363), still unidentified. The four were named only in March 2022: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (claims 14232, 14210, 14071). The exception is GMR, which was identified early as holding roughly 50% of the 100MW site (claims 14411, 14231). Several other relationships are not customer revenue. Coinmint is a co-hosting provider, so Applied Digital is its customer (claim 14419). SparkPool is an advisor (claims 14418, 14358). The Bitmain 200PH order is an equipment purchase (claims 14504, 14416). The Texas 200MW wind project is a letter of intent that is explicitly subject to definitive agreements (claims 14204, 14074).

Concentration is the central diligence issue. About half of the first facility is contracted to a single counterparty, GMR. GMR and Bitmain, two of the four named customers, are also strategic equity investors (claims 14505, 14377, 14378). Bitmain's affiliate Antpool is in addition the joint-venture partner in 1.21 Gigawatts LLC (claims 14234, 14212). The customer base is therefore heavily intertwined with the shareholder base, which should be treated as functionally related-party revenue until arm's-length pricing is shown.

The revenue arithmetic cannot be reconciled from this record. No claim discloses contract value, pricing per MW, term or minimum-take provisions. The only dollar figure tied to the venture, $8,000 for the 1.21 Gigawatts formation (claim 14234), is a capitalization amount and not revenue. Allocation statements such as "total available energy" and "approximately 50%" describe capacity, not dollars. The claim set provided ends mid-entry at claim 14059 (April 2022), so later disclosures that may supply these figures were not reviewed here.

Applied Digital's disclosure record splits into two unrelated eras, and only the second involves customers. The 2006-2008 claims belong to its predecessor, Flight Safety Technologies. The Department of Transportation Volpe Center award (claim 21053) was a real government contract. The other entries are service vendors rather than customers: a transfer agent (21069), a proxy solicitor engaged for about $22,000 (21068), and an auditor (21067). None of them bears on the current business.

The hosting-era record shows customers being named only gradually. In August 2021 the company cited agreements with two unidentified customers covering more than half of available energy (14506), while separately naming GMR and Bitmain as hosting counterparties (14507). By November 2021 the count had grown to four customers, still unnamed, said to absorb all of the first facility's energy and 85MW at a second site (14412, 14374, 14363). The four were not identified until March 2022: JointHash (a GMR subsidiary), Bitmain, F2Pool and Hashing LLC (14232, 14200, 14210, 14071). Roughly seven months therefore passed between the first customer claim and full identification.

The concentration picture is weaker than four names suggest. GMR alone was said to hold about 50% of the 100MW first site (14411, 14231, 14201). GMR and Bitmain are also strategic equity investors (14505, 14377, 14378). Bitmain plays at least four roles: equipment supplier through a 200PH order (14504, 14416), source of hosting leads (14417), customer, and, through its affiliate Antpool, joint-venture partner in 1.21 Gigawatts LLC (14234, 14203). Much of the early book therefore depends on investors and their affiliates rather than arm's-length buyers. SparkPool (14418, 14376, 14358) was an adviser and investor, not a customer. The Coinmint arrangement (14419) runs the other way, with Applied Digital as the hosted party.

Several items are not binding commitments. The Texas 200MW wind facility is a letter of intent that is expressly subject to definitive agreements (14204, 14074). The 1.21 Gigawatts vehicle, described as targeting up to 1.5GW, is a formation-stage LLC agreement, and its $8,000 figure looks like a nominal contribution rather than committed capital or revenue.

On revenue, no reconciliation is possible from this record. None of the hosting agreements carries a disclosed price, term, minimum take or contract value, so the energy-allocation claims cannot be tied to any reported revenue line. The record as supplied also stops in April 2022, mid-claim 14060, so later customer developments are not assessed here.

Applied Digital's disclosure record shows a company whose signed customer base was described anonymously for seven months before it was named, and whose named customers overlap heavily with its own investors and suppliers. The pre-2021 entries are not customer evidence. The Volpe Center award (21053) belongs to the predecessor shell, Flight Safety Technologies. Pacific Stock Transfer (21069), Broadridge (21068) and Wolf & Company (21067) are vendors the company pays, not counterparties that pay it.

The hosting book went from unnamed to named in stages. In August 2021 the company cited agreements with '2 customers' covering more than 50% of available energy, without naming them (14506). It did name GMR and Bitmain as hosting contract holders in the same filing (14507). By November the count was '4 customers' absorbing the full first-facility allocation plus 85MW at a second site, still unidentified (14412, 14374, 14363). Only in March 2022 were the four named: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (14232, 14200, 14210, 14071). These read as executed agreements.

The framework language sits elsewhere. The Texas 200MW wind facility is a letter of intent expressly 'subject to entry into definitive agreements' (14074, 14204). SparkPool appears as an equity investor and adviser, not a paying customer (14418, 14376, 14358). Coinmint is a host to which Applied Digital is itself the customer (14419). Bitmain's 'leads for potential hosting customers' (14417) is pipeline, not contract.

Concentration is severe and runs back to related parties. GMR alone holds roughly half of the 100MW first facility (14411, 14231, 14201). The Bitmain group is at once a hosting customer, an equipment supplier (the 200PH order, 14504, 14416), a strategic equity holder (14505, 14378), and, through its affiliate Antpool, a joint-venture partner in 1.21 Gigawatts LLC (14234, 14212). Two investor-linked groups therefore appear to account for most of the contracted capacity. That makes the commercial terms and their durability harder to assess at arm's length.

The revenue arithmetic cannot be reconciled from this record. Every customer commitment is expressed in megawatts, not dollars. No pricing, term length, take-or-pay structure or revenue figure is disclosed. The only dollar amounts on file are the $22,000 proxy fee (21068) and the $8,000 Antpool JV entry (14234), and neither is customer revenue. Note also that the claim list provided was cut off at claim 14060, so later disclosures were not reviewed.

Applied Digital's record separates into a pre-2009 shell period and a 2021-2022 hosting build-out. The early entries are not customer evidence. The Volpe Center award (claim 21053) was a government development contract of the predecessor, Flight Safety Technologies. Broadridge, Wolf & Company and Pacific Stock Transfer (claims 21068, 21067, 21069) are vendors the company paid, not counterparties paying it. The commercially relevant history begins in August 2021, and it begins vaguely. The company cited agreements with '2 customers' for more than half of available energy without naming them (claim 14506). In the same filing it said GMR and Bitmain held hosting contracts that would pre-fill the first 100MW site (claim 14507). By November 2021 the unnamed count had grown to four customers covering all energy at the first facility plus 85MW at a second (claims 14412, 14374, 14363), and only GMR was named, at roughly 50% of the first 100MW (claim 14411). Identification arrived in March 2022, when the four were disclosed as JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 14232, 14210). Several items read as customers but are not. The Bitmain 200PH order (claims 14504, 14416) is equipment the company bought. The Coinmint co-hosting agreement (claim 14419) made Applied Digital the hosted party. The SparkPool and GMR arrangements (claims 14418, 14502) are advisory services delivered to the company. The Texas 200MW wind facility is explicitly a letter of intent 'subject to entry into definitive agreements' (claims 14204, 14074) and should carry no backlog weight. Concentration is the central diligence issue. GMR alone took about half the initial capacity, and GMR and Bitmain together are two of the four named customers. Both are also strategic equity investors alongside SparkPool (claims 14505, 14376, 14377, 14378). The 1.21 Gigawatts joint venture is with Antpool, a Bitmain affiliate (claims 14234, 14203, 14212). Much of the early demand therefore came from shareholders and their affiliates, which makes it closer to related-party demand than an arm's-length market test. The revenue arithmetic cannot be reconciled from this record. No contract values, hosting rates, terms or revenue figures accompany any customer claim. The only dollar amounts are a $22,000 proxy fee and an $8,000 figure attached to the Antpool venture, and neither is customer revenue. The megawatt allocations are internally consistent, at 100MW plus 85MW fully assigned, but they show contracted capacity, not recognized or committed dollars. The supplied record ends mid-entry at an April 2022 restatement of the Antpool venture (claim 14059).

Applied Digital's record splits into two unrelated eras, and only the second bears on commercial diligence. The 2006-2008 entries belong to its predecessor, Flight Safety Technologies. The Volpe Center award is a genuine government contract (claim 21053), but Broadridge, Wolf & Company and Pacific Stock Transfer are service providers the company pays, not customers (claims 21068, 21067, 21069).

From August 2021 the disclosures describe a crypto-hosting business, and customer identities surfaced gradually. The first filing cites agreements with "2 customers" covering more than half of available energy, without naming them (claim 14506). A parallel passage says GMR and Bitmain hold hosting contracts that "pre-fill" the first 100MW site (claim 14507). By November the count had become four unnamed customers absorbing all first-facility energy plus 85MW at a second site (claims 14412, 14374, 14363). Only in March 2022 were the four named: JointHash (a GMR subsidiary), Bitmain, F2Pool and Hashing LLC (claims 14232, 14210, 14071).

GMR alone is described as holding roughly 50% of the 100MW (claims 14411, 14231, 14201). The concentration is also deeper than four names suggest. GMR and Bitmain are equity investors as well as customers (claims 14505, 14377, 14378), so the anchor tenants are related parties. Bitmain is further the equipment supplier on a 200PH order and the source of hosting leads (claims 14504, 14417). Its affiliate Antpool is the company's partner in the 1.21 Gigawatts LLC (claims 14234, 14203). One counterparty group therefore sits on the customer, supplier, shareholder and joint-venture sides at once.

Several relationships are not customer revenue at all. SparkPool's arrangements are advisory and engineering services the company receives (claims 14418, 14358). The Coinmint co-hosting deal has Applied Digital as the customer (claims 14419, 14371). The Texas wind facility is explicitly a letter of intent "subject to entry into definitive agreements" (claims 14204, 14074). The Antpool vehicle is a joint-venture formation, not a services contract.

Revenue arithmetic cannot be reconciled from this record. No claim discloses contract values, term lengths, per-megawatt pricing or take-or-pay provisions. The only dollar figures are the $22,000 proxy fee and an $8,000 figure attached to the Antpool LLC, and neither is customer revenue. The "85MW at a second facility" is allocated before any energy agreement for that site appears on record. Readers should treat the hosting backlog as capacity commitments from related-party anchors rather than quantified contracted revenue. Note that the supplied record breaks off mid-quote at claim 14059, so later filings are not reflected here.

Applied Digital's named-customer record is narrower than its volume suggests. The pre-2021 entries belong to the Flight Safety Technologies shell. One is a government contract (claim 21053). The rest are vendor relationships, namely a transfer agent, an auditor and a proxy solicitor (claims 21069, 21067, 21068), and none of them is a customer. The usable history begins in August 2021. At that point the company said it had agreements with '2 customers' covering more than half of available energy, without naming them (claim 14506). It then described GMR and Bitmain as having entered hosting contracts for its first 100MW site (claim 14507). By November 2021 the count had grown to '4 customers' covering the full first facility plus 85MW at a second site, again unnamed (claims 14412, 14374, 14363). Only in March 2022 did filings identify all four: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (claims 14232, 14210, 14071). That disclosure is the first point at which the full customer roster can be tested. Concentration was severe from the start and has not visibly diluted. GMR alone holds roughly 50% of the 100MW capacity (claims 14411, 14231, 14201). GMR and Bitmain are also strategic equity investors (claims 14505, 14377, 14378). Bitmain is additionally the equipment supplier (claims 14504, 14416) and a source of hosting leads (claim 14417). Its affiliate Antpool is a joint-venture partner in 1.21 Gigawatts LLC (claims 14234, 14212). Investor, supplier, customer and JV partner therefore overlap substantially, so the hosting book carries related-party character and deserves scrutiny on pricing and enforceability. Several items should be kept out of the signed-customer column. The Texas 200MW wind facility is a letter of intent that is expressly 'subject to entry into definitive agreements' (claims 14204, 14074). The SparkPool arrangement is advisory (claims 14418, 14358). The Coinmint agreement runs the other way, with Applied Digital as the hosted party (claim 14419). On revenue, nothing in the record discloses contract values, tenor, per-MW pricing or recognized revenue tied to these agreements. The $8,000 attached to the Antpool LLC (claim 14234) looks like a formation or capital figure, not income. As a result, the announced deals cannot be reconciled to any revenue arithmetic from this record. Capacity allocation (100MW plus 85MW) is the only quantified commitment. The claim list supplied ends mid-entry at claim 14061, so later customer disclosures were not reviewed.

Applied Digital's record splits into two unrelated eras. The 2006-2008 entries come from the predecessor shell, Flight Safety Technologies. Only the Volpe Center contract (21053) is revenue-side. Broadridge, Wolf & Company and Pacific Stock Transfer (21068, 21067, 21069) are vendors the company paid, not customers, and should carry no weight in a customer read. The relevant history starts in August 2021.

At that point the customer book was anonymous. The company cited agreements with '2 customers' accounting for more than 50% of available energy (14506). In the same filing it said GMR and Bitmain had hosting contracts to pre-fill the first 100MW site (14507). By November the count had become '4 customers' covering the full first facility plus 85MW at a second site (14412, 14374, 14363), still unnamed. Names arrived only in March 2022: JointHash Holding (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (14232, 14210, 14071). So for roughly seven months, investors were asked to credit capacity commitments from counterparties who were described but not identified.

Concentration is the more material issue. GMR alone is said to hold about 50% of the 100MW (14411, 14231, 14201). GMR, Bitmain and SparkPool are all strategic equity investors (14505, 14377, 14378). Bitmain plays several roles at once. It is an equipment supplier through the 200PH order (14504, 14416), a source of customer leads (14417) and a hosting customer. Through its affiliate Antpool it is also the joint-venture partner in 1.21 Gigawatts LLC (14234, 14203). On this record, at least half of first-site capacity, and likely more, sits with shareholders who also supply hardware or services. That makes the counterparties less arm's-length than 'signed customers' implies.

Several items should not be read as customer agreements at all. The Texas 200MW wind facility is a letter of intent explicitly 'subject to entry into definitive agreements' (14204, 14074). 1.21 Gigawatts is an LLC formation with an $8,000 figure attached, which is a development vehicle, not contracted revenue (14234). Coinmint (14419) is a facility hosting Applied Digital's own miners, so Applied Digital is the customer there. SparkPool's Ether-pool arrangement (14418) is advisory.

The revenue arithmetic cannot be reconciled from this record. No contract values, hosting rates, terms or recognized revenue are disclosed alongside the four agreements. The excerpt also stops at claim 14059, so later filings that might allow a tie-out are not in view. The most that can be said is that the claims do not contradict each other on capacity: roughly 50MW to GMR, the balance of 100MW to the other three, and 85MW pre-sold at an unbuilt second site. Any dollar-denominated backlog claim should be treated as unverified until priced contracts appear.

Applied Digital's record splits into two unrelated eras, and only the second bears on its current customer base. The 2006-2008 entries describe the predecessor shell, Flight Safety Technologies. The Volpe Center award (claim 21053) was a genuine signed government contract. The transfer agent, auditor and Broadridge items (claims 21069, 21067, 21068) are vendor engagements, not customers. From August 2021 the language shifts to hosting, and the distinction between signed and aspirational matters more. The first filing cites agreements with two unnamed customers covering more than half the available energy (claim 14506). The same filing identifies those parties as GMR and Bitmain (claim 14507). Both are also strategic equity investors, alongside SparkPool (claim 14505). By November 2021 the count rose to four unnamed customers covering the full first facility and 85MW at a second (claims 14412, 14374, 14363). Names arrived only in March 2022: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (claims 14232, 14210, 14071). Concentration therefore never really diluted. GMR alone is repeatedly described as holding roughly half of the 100MW site (claims 14411, 14231, 14201), and at least two of the four customers are equity holders or their affiliates, so the base is related-party-heavy rather than arm's-length. Several items presented as partnerships are not customer revenue at all. SparkPool's arrangement is advisory and partly conditional on a decision not yet made (claim 14418). Coinmint and the Bitmain 200PH order are relationships in which Applied Digital is the buyer (claims 14419, 14416). The Antpool LLC is a joint-venture framework seeded with $8,000 that targets up to 1.5GW of datacenters (claims 14234, 14203), but it carries no committed offtake. The Texas 200MW wind facility is expressly a letter of intent subject to definitive agreements (claims 14204, 14074). Revenue arithmetic cannot be reconciled from this record. No contract values, per-megawatt pricing, term lengths or recognized revenue are disclosed, and the only dollar figures are costs or capital contributions. A reader can confirm that capacity was allocated to named, partly affiliated counterparties. A reader cannot confirm what that allocation is worth. The record supplied ends mid-entry at claim 14060, so later disclosures are not assessed here.

Applied Digital's customer record splits into two eras. The pre-2021 entries belong to its predecessor, Flight Safety Technologies. They describe one government contract, the Volpe Center wake-vortex award (claim 21053), along with service vendors such as a transfer agent, a proxy solicitor and an auditor (claims 21069, 21068, 21067). None of the vendor entries are customer relationships. The substantive record begins in August 2021. At that point the company disclosed agreements with "2 customers" covering more than half of available energy without naming them (claim 14506), while separately stating that GMR and Bitmain had hosting contracts to pre-fill the first 100MW site (claim 14507). By November 2021 the count had grown to four unnamed customers absorbing all capacity at the first facility and 85MW at a second (claims 14412, 14374, 14363). The four were not identified until the March 2022 filings: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 14232, 14210, 14200). These read as signed hosting agreements, not MOUs. Several other relationships are softer or point the other way. SparkPool is an equity investor and advisor, not a customer (claims 14376, 14418). Coinmint is a co-hosting provider to which Applied Digital was itself the customer (claim 14419). The Texas 200MW wind project is expressly a letter of intent "subject to entry into definitive agreements" (claims 14204, 14074). Concentration is the central diligence issue. GMR alone accounts for roughly 50% of the first 100MW (claims 14411, 14201, 14211). Bitmain is at once a strategic equity investor, the equipment supplier on a 200PH order, a source of hosting leads, a hosting customer, and, through its affiliate Antpool, a joint-venture partner in 1.21 Gigawatts LLC (claims 14505, 14416, 14417, 14234, 14203). In effect, two investor-affiliated counterparties anchor most contracted capacity, and the four-customer roster is related-party-heavy rather than an arm's-length book. The record provides no revenue, pricing, term or take-or-pay figures for any hosting agreement. Every commitment is expressed in megawatts. The one dollar figure attached to the Antpool venture ($8,000, claim 14234) is unexplained in the excerpt and cannot be tied to economics. Disclosed revenue therefore cannot be reconciled against the announced deals on this record. A reader should treat the capacity claims as signed but unpriced, and the Texas expansion as unsigned. The claims supplied end in April 2022 (claim 14060 is truncated), so later developments are outside this read.

Applied Digital's record splits into two unrelated eras. The 2006–2008 entries come from the predecessor shell, Flight Safety Technologies. The Volpe Center award (claim 21053) is the only true customer contract from that period. Broadridge, Pacific Stock Transfer and Wolf & Company (claims 21068, 21069, 21067) are vendors the company paid, not counterparties paying it, and should carry no weight in a customer read. The relevant history starts in August 2021. In that period, several relationships described as partnerships run in the wrong direction. SparkPool's services and Ether-pool advisory agreements (claims 14503, 14418), GMR's asset-management services (claim 14502) and the Coinmint co-hosting arrangement (claim 14419) all have Applied Digital as the buyer of services, not the seller. Disclosure of hosting customers moved from anonymous to named over roughly seven months. In August 2021 the company cited only two unidentified customers for more than 50% of available energy (claim 14506), while separately saying GMR and Bitmain had hosting contracts (claim 14507). By November that had become four unnamed customers covering the full first facility plus 85MW at a second (claims 14412, 14374, 14363). The four were not identified until March 2022: JointHash (a GMR subsidiary), Bitmain, F2Pool and Hashing LLC (claims 14232, 14210). Concentration is high and tangled with ownership. GMR alone is said to hold roughly 50% of the 100MW site (claims 14411, 14231). GMR, Bitmain and SparkPool are all described as strategic equity investors (claim 14505). The Bitmain complex is simultaneously a customer, the equipment supplier on a 200PH order (claims 14504, 14416), a source of customer leads (claim 14417), and, through its affiliate Antpool, a joint-venture partner in 1.21 Gigawatts LLC (claims 14234, 14212). At least half the contracted base therefore sits with equity holders, so the hosting contracts are better read as related-party arrangements than as arm's-length market validation. Separately, the Texas 200MW wind facility is explicitly a letter of intent (claims 14204, 14074) and should not count toward contracted capacity. The revenue arithmetic cannot be reconciled from this record. None of the claims discloses contract value, price per MW, term, or recognized hosting revenue. The only dollar figure attached to a 2021–2022 relationship is the $8,000 in the Antpool joint-venture label (claim 14234), and the quoted text does not state its units or purpose. The claim list also ends mid-entry at claim 14074, so any later named customers or revenue disclosures fall outside this read.

Applied Digital's record of named counterparties has two distinct phases. The pre-2009 entries relate to its predecessor, Flight Safety Technologies, and are not crypto-hosting customers. The Volpe Center award (claim 21053) is a genuine government contract. The auditor, proxy solicitor and transfer agent (claims 21067, 21068, 21069) are vendors, and none of them bear on commercial traction. The relevant story begins in August 2021, when the company said it had agreements with two customers covering more than half of available energy without naming them (claim 14506). In the same filing it identified GMR and Bitmain as parties to hosting contracts for the first 100MW site (claim 14507). By November 2021 the count had grown to four customers covering the full first facility and 85MW at a second, still unnamed (claims 14412, 14374, 14363). Only in the March 2022 filings were all four identified: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 14232, 14200, 14210, 14071). That is credible progression from anonymous to named signed agreements, but the concentration is severe. GMR alone holds a contract for roughly 50% of the 100MW site (claims 14411, 14231, 14201). Both GMR and Bitmain are also strategic equity investors (claims 14505, 14377, 14378). Bitmain is additionally the company's equipment supplier (claims 14504, 14416) and, through its affiliate Antpool, a joint-venture partner in 1.21 Gigawatts LLC (claims 14234, 14203). Much of the early revenue base therefore rests on shareholders with overlapping supplier and partner roles, which diligence should treat as quasi-related-party revenue. Several items describe services flowing to Applied Digital rather than customers paying it. SparkPool is an adviser and investor, not a hosting customer (claims 14503, 14418, 14358). Coinmint hosted the company's own rigs (claims 14419, 14371). The Texas 200MW wind facility is explicitly a letter of intent subject to definitive agreements (claims 14204, 14074) and should not be counted as contracted capacity. On revenue arithmetic, the record discloses no contract values, pricing per megawatt or term lengths for any hosting agreement; the only dollar figures are a $22,000 proxy fee (claim 21068) and an $8,000 item tied to the Antpool venture (claim 14234). Reconciliation of reported revenue to the announced deals is therefore not possible from these claims, and the record as supplied is truncated after claim 14072, so later customer disclosures were not reviewed.

Applied Digital's record splits into two unrelated eras. The 2006-2008 claims belong to the predecessor shell, Flight Safety Technologies. They comprise one government contract (21053) and three service vendors: a transfer agent, an auditor, and Broadridge at roughly $22,000 (21069, 21067, 21068). None of them bears on the current hosting business, and the vendors are not customers at all.

The hosting-era disclosures show identification arriving late. In August 2021 the company described agreements with '2 customers' covering more than 50% of available energy without naming them (14506). It separately said GMR and Bitmain had entered hosting contracts (14507). By November 2021 the count had become '4 customers' absorbing the full first-facility allocation plus 85MW at a second site, still unnamed (14412, 14374, 14363). Only in March 2022 were the four identified: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (14232, 14200, 14210, 14071). So for roughly seven months the contracted book rested partly on unidentified counterparties.

Several items are not customer revenue at all. SparkPool provides advisory and services, and its relationship was being wound down (14503, 14418, 14358). Coinmint hosts Applied Digital's own rigs, which makes Applied Digital the customer (14419). The Texas wind facility is a letter of intent only (14204, 14074). The 1.21 Gigawatts LLC is a joint-venture formation agreement with Antpool for prospective capacity, not a signed offtake (14234, 14203).

Concentration is the dominant issue. GMR alone holds about 50% of the 100MW first facility (14411, 14231, 14201). The Bitmain complex sits on every side of the business. It is a hosting customer, the supplier of the 200PH equipment order (14504, 14416) and a source of customer leads (14417). Through its Antpool affiliate it is also the JV partner. GMR and Bitmain are also strategic equity investors (14505, 14377, 14378). Two related-party groups therefore appear to account for most of the contracted capacity.

The revenue arithmetic cannot be reconciled from this record. Every capacity commitment is expressed in megawatts, with no contract values, rates, terms or recognized revenue attached. The only dollar figures are the $22,000 proxy fee and an $8,000 figure attached to the 1.21 Gigawatts LLC formation (14234), and neither speaks to customer economics. The record supplied also ends mid-claim at 14074, so any later disclosures are not reflected here.

The pre-2021 entries describe Applied Digital's predecessor shell, Flight Safety Technologies, and none of them is commercial customer evidence. The Volpe Center award is a government development contract (claim 21053). Broadridge, Wolf & Company and Pacific Stock Transfer are vendors the company paid, not customers that paid it (claims 21068, 21067, 21069). The relevant record therefore begins in August 2021, and it shows a clear progression from unnamed counterparties to named ones. At first the company disclosed agreements with "2 customers" for more than half of available energy without naming them (claim 14506). By November that had grown to "4 customers" covering the full first facility and 85MW of a second, still unnamed (claims 14412, 14374, 14363). Only in March 2022 did filings identify them as JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (claims 14232, 14210, 14200, 14071). Those four agreements, together with the Amended and Restated Energy Services Agreement for 100MW (claims 14233, 14202), are the strongest evidence of signed revenue contracts. The Texas wind facility, by contrast, is explicitly a letter of intent "subject to entry into definitive agreements" (claims 14204, 14074) and should not be counted. The SparkPool, GMR and Bitmain relationships are described as services, advisory or "preferred access" arrangements (claims 14503, 14502, 14417, 14418), not offtake contracts. Concentration is acute and circular. GMR alone is repeatedly said to hold roughly 50% of the 100MW first site (claims 14411, 14201, 14211). GMR and Bitmain are simultaneously hosting customers, equipment suppliers and strategic equity investors (claims 14505, 14507, 14377, 14378). The 1.21 Gigawatts joint venture is with Antpool, itself a Bitmain affiliate (claims 14234, 14203). In effect, two shareholder groups underpin most contracted capacity, and the company's own mining initially ran on Coinmint's co-hosting facility (claim 14419). On revenue reconciliation, the record supplies no revenue figures, contract values, pricing per megawatt or term lengths. The only dollar amounts are a $22,000 proxy fee and an $8,000 figure tied to the Antpool LLC formation (claims 21068, 14234). Capacity commitments therefore cannot be reconciled to disclosed revenue from this record. Diligence should obtain contract tenor, take-or-pay terms and related-party revenue share. The claim list also appears truncated after claim 14072, so later disclosures may alter this picture.

Applied Digital's record splits into two unrelated eras, and only the second contains customer evidence. The 2006 Volpe Center award (claim 21053) is a real government contract, but it belongs to the predecessor Flight Safety Technologies business. The 2008 entries for Broadridge, Wolf & Company and Pacific Stock Transfer (claims 21068, 21067, 21069) name vendors the company paid, not customers. None of them bears on the current hosting franchise. Several 2021 "partnerships" also run the wrong way for revenue diligence. The Bitmain 200PH rig order (claims 14504, 14416) and the Coinmint co-hosting arrangement (claim 14419) make Applied the buyer. The SparkPool services and Ether-pool advisory agreements (claims 14503, 14418) are inputs rather than demand, and claim 14358 hints, in truncated form, that SparkPool was winding something down. The customer disclosure sharpened over time. In August 2021 the company cited agreements with two unidentified customers covering more than half of available energy (claim 14506), alongside hosting contracts attributed to GMR and Bitmain (claim 14507). By November that had become four customers, still unnamed, filling the first facility and 85MW at a second (claims 14412, 14374, 14363). Only in March 2022 were all four identified: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (claims 14232, 14210, 14200, 14071). Concentration is the central issue. GMR alone accounts for roughly 50% of the 100MW first site (claims 14411, 14231, 14201). GMR and Bitmain are also strategic equity investors (claims 14505, 14376 through 14378), so at least two of the four anchor tenants are related shareholders rather than arm's-length buyers. The January 2022 1.21 Gigawatts JV (claims 14234, 14203) adds further dependence on the Bitmain group, since it is with Antpool, a Bitmain affiliate. The $8,000 figure tagged to that JV carries no stated unit or context. The Texas 200MW wind project is expressly a letter of intent subject to definitive agreements (claims 14204, 14074) and should not be counted as contracted capacity. On revenue, the arithmetic cannot be reconciled from this record. No contract values, hosting rates, terms or recognized revenue are disclosed against any of the named deals, so the announced capacity commitments cannot be tested against reported results. The supplied claim list also ends mid-entry at claim 14072, so any later disclosures fall outside this read.

Applied Digital's record separates cleanly into signed agreements and softer language, but the counterparties behind the signed agreements are less independent than the volume of announcements suggests. The four entries from 2006 to 2008 (claims 21053, 21067, 21068, 21069) belong to the predecessor shell, Flight Safety Technologies. Only the Volpe Center award is a customer contract. The rest are vendor engagements, including the $22,000 Broadridge proxy fee, and say nothing about the current hosting business.

The hosting customer base was disclosed in stages. In August 2021 the company said only that it had agreements with '2 customers' covering more than 50% of available energy (claim 14506), alongside a vaguer statement that GMR and Bitmain 'have entered into hosting contracts' (claim 14507). By November 2021 this had become four unnamed customers absorbing all 100MW at the first facility plus 85MW at a second (claims 14412, 14374, 14363). The four were not named until March 2022: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (claims 14232, 14210, 14071). These are signed agreements. However, the filings excerpted here give no contract values, terms or take-or-pay provisions.

Concentration is the central issue. GMR alone holds roughly half of the 100MW (claims 14411, 14231, 14201). GMR and Bitmain are also strategic equity investors (claims 14505, 14376). In addition, GMR is a paid service provider to the company (claims 14502, 14359), and Bitmain both supplies the 200PH rig order and refers hosting leads (claims 14504, 14416, 14417). Antpool, a Bitmain affiliate, is the joint-venture partner in 1.21 Gigawatts LLC (claims 14234, 14203). As a result, at least half of contracted capacity, and probably more, sits with insiders and their affiliates, not arm's-length buyers.

Several announcements are not customer revenue at all. The Coinmint co-hosting arrangement (claim 14419) and the SparkPool advisory and services agreements (claims 14503, 14418) run the other way, with Applied Digital as the buyer. The Texas 200MW wind project is only a letter of intent (claims 14204, 14074). The 1.21 Gigawatts entity is a framework joint venture with an $8,000 reference in claim 14234 and no committed capacity.

Revenue cannot be reconciled from this record. No claim gives a contract value, a per-MW rate or recognized hosting revenue, so the 185MW of commitments cannot be tied to reported financials. The second-facility 85MW allocation also appears without any disclosure of that site's status. The input was truncated after claim 14074, so any later disclosures that might close this gap were not reviewed.

Applied Digital's customer record splits into two unrelated eras, and only the second describes hosting customers. The 2006-2008 items belong to the predecessor shell, Flight Safety Technologies. The Volpe Center award (claim 21053) is a signed federal contract, but the auditor, transfer agent and Broadridge items (claims 21067, 21069, 21068) are vendor relationships, and the $22,000 proxy fee is an expense, not revenue. None should be read as commercial traction. From August 2021 the language moves from anonymous to named, but in stages. The first hosting disclosure cited "2 customers" taking more than half of available energy without naming them (claim 14506). A parallel passage said GMR and Bitmain had hosting contracts (claim 14507). By November 2021 the count was four unnamed customers covering the full first facility plus 85MW at a second (claims 14412, 14374, 14363). Identification arrived only in March 2022: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (claims 14232, 14210, 14200). Those four are signed agreements. The Texas 200MW wind project, by contrast, rests on a letter of intent (claims 14204, 14074), and the Coinmint arrangement makes Applied the customer rather than the host (claim 14419). Concentration is severe and circular. GMR alone is described as holding roughly 50% of the 100MW first site (claims 14411, 14231, 14201). GMR and Bitmain are also strategic equity investors (claims 14505, 14377, 14378), so at least half the named customer base overlaps with the shareholder register. Bitmain's role goes further still: it is the equipment supplier on a 200PH order (claims 14504, 14416), a source of hosting leads (claim 14417), a customer, and, through its affiliate Antpool, the partner in the 1.21 Gigawatts LLC (claims 14234, 14203). The $8,000 attached to that venture reads as nominal capitalization, not committed capital. The SparkPool relationships are advisory services (claims 14503, 14418). The disclosures also put customer allocations on an 85MW second facility, while the energy agreement on record covers only 100MW (claims 14233, 14202). The revenue arithmetic cannot be reconciled from this record. No contract values, per-MW pricing, terms or recognized hosting revenue appear in any cited claim, so the fully-contracted 185MW cannot be tested against reported figures. The record supplied also ends mid-entry at claim 14074, so any later counterparty changes are not reflected here.

Applied Digital's record splits into two unrelated eras, and only the second bears on the current business. The 2006-2008 claims belong to predecessor Flight Safety Technologies. Only the Department of Transportation Volpe Center award is a genuine customer contract (21053). Broadridge, Pacific Stock Transfer and Wolf & Company are vendors the company pays (21068, 21069, 21067), and they should not be counted as commercial relationships.

From August 2021 the disclosure follows a familiar pattern: unnamed customers first, names later. The August filing cited agreements with "2 customers" covering more than half of available energy (14506). The context identifies them as GMR and Bitmain, which are also described as equity-investing "strategic partners" (14505, 14507). By November the count had risen to "4 customers" absorbing all of the first facility and 85MW at a second (14412, 14374, 14363), still without names. Only in March 2022 were they identified as JointHash Holding (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (14232, 14210). The new names did little to dilute concentration. GMR alone holds roughly 50% of the 100MW site (14411, 14231, 14201). GMR and Bitmain are simultaneously shareholders, customers and service or equipment providers. Bitmain supplies 200PH of rigs and customer leads (14504, 14416, 14417), and its affiliate Antpool is the company's partner in the 1.21 Gigawatts LLC joint venture (14234, 14203). The hosting book therefore rests largely on related or interlocked counterparties rather than arm's-length demand.

Several items are not customer revenue at all. The Coinmint arrangement makes Applied Digital the hosted party (14419). The SparkPool agreements are advisory and engineering services flowing to the company (14503, 14418). The Texas 200MW wind project is expressly a letter of intent "subject to entry into definitive agreements" (14204, 14074). The 1.21 Gigawatts entity is a joint-venture framework, not a contract for capacity.

The revenue arithmetic cannot be reconciled from this record. The claims express commitments in megawatts and percentage allocations. None discloses pricing per megawatt, contract term, minimum take-or-pay obligations or expected dollar revenue. The only dollar figures are an $8,000 joint-venture formation amount (14234) and a proxy fee (21068). A reader can confirm that 100MW was nominally allocated across four named, partly affiliated customers. A reader cannot determine what those allocations are worth or how durable they are. The record supplied also ends mid-entry at claim 14072, so later disclosures may change this picture.

Applied Digital's record mixes very different kinds of agreements, and only a few are customer contracts. The 2006–2008 entries come from the predecessor shell, Flight Safety Technologies: a DOT Volpe Center development award (21053) and engagements of an auditor, a transfer agent and a proxy solicitor (21067, 21069, 21068). Apart from the Volpe award, these are vendor relationships, and none bears on the current hosting business. In the 2021 cycle, several named "partners" are also suppliers or counterparties to APLD. Coinmint hosts APLD's own rigs (14419). Bitmain sells APLD 200PH of equipment (14504, 14416). SparkPool and GMR provide advisory and asset-management services (14503, 14502, 14418). None of these is revenue-generating customer activity.

The real customer evidence is the hosting book, and it emerged slowly. In August 2021, APLD disclosed two unnamed customers covering more than 50% of the 100MW Energy Services Agreement (14506). By November it cited four unnamed customers covering the full first facility plus 85MW at a second (14412, 14374, 14363). Not until March 2022 did the filings name them: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (14232, 14210, 14200). For roughly seven months, investors relied on counts rather than identities.

Concentration is high and closely tied to related parties. GMR alone holds about 50% of the 100MW (14411, 14231, 14201). GMR and Bitmain are also strategic equity investors (14505, 14377, 14378), and the four-customer roster overlaps that investor group. The January 2022 1.21 Gigawatts LLC formed with Antpool, a Bitmain affiliate, is a joint-venture framework targeting up to 1.5GW, not offtake (14234, 14203). The 200MW Texas wind project rests only on a letter of intent (14204, 14074). Neither should be treated as contracted capacity.

Revenue reconciliation cannot be performed from this record. None of the hosting claims discloses pricing, term, take-or-pay provisions or dollar values. The only dollar figures are a $22,000 proxy fee (21068) and an $8,000 LLC contribution (14234). The claimed megawatts are therefore allocations of energy, not demonstrated revenue. The record also ends mid-claim at 14074, so any later conversions or terminations are not captured here.

Applied Digital's record splits into two unrelated eras, and only the second bears on customer diligence. The 2006-2008 entries come from its predecessor, Flight Safety Technologies. The Volpe Center award (claim 21053) is a genuine signed government contract. The transfer agent, proxy solicitor and auditor items (claims 21069, 21068, 21067) are vendor engagements, not customers, and the $22,000 Broadridge fee is an expense.

From 2021 the customer story firms up in stages. The August 2021 filing cites agreements with 2 customers covering more than half of available energy without naming them (claim 14506). It names GMR and Bitmain as hosting counterparties only in the same passage that identifies them as strategic equity investors (claims 14505, 14507). By November the unnamed count rose to 4 customers, covering the first facility plus 85MW at a second site (claims 14412, 14374, 14363). Only in March 2022 were all four named: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (claims 14232, 14200, 14210, 14071). These read as executed agreements, but terms, durations and pricing are not disclosed.

Several headline relationships are not customer contracts. The SparkPool and GMR arrangements are advisory and services agreements in which Applied Digital is the buyer (claims 14503, 14502, 14418, 14359). The Bitmain 200PH order is an equipment purchase (claims 14504, 14416). Coinmint hosted Applied Digital's own rigs, which makes the company Coinmint's customer, not the reverse (claims 14419, 14357). The Texas wind facility is explicitly a letter of intent (claims 14204, 14074). The 1.21 Gigawatts LLC with Antpool is a joint-venture formation, not offtake (claims 14234, 14203).

Concentration is the central issue. GMR alone holds roughly 50% of the 100MW first facility (claims 14411, 14201, 14211). Bitmain sits simultaneously as customer, equipment supplier, equity investor, lead source and, through Antpool, JV partner (claims 14378, 14417, 14234). At least half of contracted capacity therefore runs through shareholders, which makes the arm's-length quality of the economics hard to verify.

The revenue arithmetic cannot be reconciled from this record. No claim discloses per-MW rates, contract values or recognized hosting revenue. The only figures are the proxy fee and an unlabeled $8,000 tied to the Antpool LLC, which is likely in thousands but unconfirmed. The claim list supplied here also ends mid-record at claim 14074, so later disclosures may close part of this gap.

Applied Digital's record falls into two eras. The pre-2021 entries come from its Flight Safety Technologies predecessor. Of those, only the Department of Transportation Volpe Center award (claim 21053) is a genuine customer contract. Pacific Stock Transfer, Broadridge and Wolf & Company (claims 21069, 21068, 21067) are vendors the company pays, and the $22,000 Broadridge figure is a cost, not revenue. The relevant diligence story begins in August 2021, when the company said it had agreements with 2 unnamed customers covering more than half its contracted energy (claim 14506). By November 2021 this became 4 unnamed customers taking all the energy at the first facility plus 85MW at a second (claims 14412, 14374, 14363). Those counterparties were identified only in March 2022, as JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 14232, 14200, 14210, 14071). For roughly seven months, then, the customer base was disclosed only as counts and percentages. Concentration is the more important finding. GMR alone contracted about 50% of the 100MW first facility (claims 14411, 14231, 14211). GMR and Bitmain are also strategic equity investors (claims 14505, 14377, 14378). Bitmain is additionally the company's equipment supplier (claims 14504, 14416), and through its affiliate Antpool it is the company's joint-venture partner in 1.21 Gigawatts LLC (claims 14234, 14203). This means about half or more of the first facility's capacity was contracted with shareholders who sit on several sides of the relationship. SparkPool, also an investor, is a service provider rather than a hosting customer (claims 14503, 14418). The Coinmint arrangement (claim 14357) places Applied Digital as the hosted party, so it does not generate hosting revenue. The firmness of the language varies. The four named hosting agreements are presented as entered into. The Texas 200MW wind facility is only a letter of intent (claims 14204, 14074), and the 1.21 Gigawatts JV describes datacenter ambitions of up to 1.5GW rather than contracted load. The revenue arithmetic cannot be reconciled from the record. No contract values, tenors, per-megawatt pricing or minimum-take terms are disclosed. The only dollar figures are the Broadridge fee and an $8,000 entry tied to the JV, whose units and direction are unstated. Capacity allocations add up to 100MW plus 85MW, but they cannot be turned into a revenue expectation. Note that the claim set supplied is truncated after April 2022, so later named-customer developments are not assessed here.

Applied Digital's disclosed counterparties fall into three groups: signed hosting customers, suppliers and advisors described as partners, and non-binding development arrangements. The pre-2021 claims (21053, 21067, 21068, 21069) belong to the predecessor shell, Flight Safety Technologies. Apart from a single Volpe Center contract, they record service vendors such as an auditor, transfer agent and proxy solicitor, not commercial customers, and they say nothing about the current business. The substantive record starts in August 2021. At that point the company described agreements with '2 customers' accounting for more than 50% of available energy without naming them (14506), and in the same filing said GMR and Bitmain had hosting contracts covering its first planned 100MW site (14507). By November 2021 the unnamed count had grown to '4 customers' covering the full first facility and 85MW of a second (14412, 14374, 14363). The four were identified only in March 2022 as JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (14232, 14200, 14210, 14071). These are the only counterparties on record framed as signed hosting customers. Much of the 'partner' language describes the opposite relationship. SparkPool provides software, design and pool-advisory services (14503, 14418). Bitmain's role includes an equipment order and customer leads (14416, 14417). Coinmint hosts Applied Digital's own rigs (14419). The Amended and Restated Energy Services Agreement covers power supply, not demand (14233, 14202). The Texas wind facility rests on a letter of intent only (14204, 14074). The 1.21 Gigawatts LLC with Antpool, a Bitmain affiliate, is a joint-venture vehicle and not a revenue contract (14234, 14203, 14212, 14073). Concentration is high and turns partly on related parties. GMR alone is repeatedly credited with roughly 50% of the 100MW first facility (14411, 14231, 14211, 14201, 14072). GMR, Bitmain and SparkPool are all disclosed equity investors (14505, 14376, 14377, 14378). As a result, two of the four named customers are also shareholders, and Bitmain is at once customer, supplier, investor and JV affiliate. The underlying contracts may well be arm's-length, but that overlap deserves scrutiny. On revenue, none of the claims gives contract values, pricing per MW, term length or recognized hosting revenue. The only figures are megawatt allocations of 100MW plus 85MW and a JV capital figure, so the announced deals cannot be reconciled to any revenue line from this record. That reconciliation has to wait for the financial statements, and the claim list supplied here ends mid-record at 14074.

Applied Digital's record splits into two unrelated eras. The pre-2021 entries belong to its predecessor, Flight Safety Technologies. Only the Volpe Center award is a genuine customer contract (21053). Broadridge, Wolf & Company and Pacific Stock Transfer are vendors the company pays, not revenue counterparties (21068, 21067, 21069), and should be excluded from any customer tally.

The hosting-era disclosures show a clear progression from anonymous to named counterparties. In August 2021 the company cited agreements with "2 customers" covering more than 50% of available energy without naming them (14506), while separately saying GMR and Bitmain had signed hosting contracts (14507). By November the unnamed count had grown to "4 customers" covering all of the first facility and 85MW at a second site (14412, 14374, 14363). Only in March 2022 were the four identified: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (14232, 14200, 14210). These are described as entered agreements, not MOUs, so they qualify as signed. The Texas 200MW wind facility, by contrast, is a letter of intent expressly "subject to entry into definitive agreements" (14204, 14074) and should carry no contracted weight.

Several headline relationships are not customer relationships at all. SparkPool supplies software, design and advisory services (14503, 14418). The Coinmint arrangement makes Applied Digital the hosted party rather than the host (14419). The 200PH Bitmain order is an equipment purchase (14504, 14416). The Energy Services Agreement is supply-side power procurement (14233).

Concentration is high and related-party in character. GMR alone holds roughly 50% of the 100MW first facility (14411, 14231, 14201). GMR and Bitmain are also strategic equity investors (14505, 14377, 14378). Bitmain's affiliate Antpool is the company's partner in the 1.21 Gigawatts joint venture (14234, 14203). The anchor customers are therefore also shareholders and development partners, which weakens their value as independent demand validation.

The revenue arithmetic cannot be reconciled from this record. No contract values, hosting rates, terms or recognized revenue are disclosed against any named deal. The only dollar figure tied to the hosting era, the $8,000 attached to the Antpool LLC formation (14234), is a capitalization item, not revenue. The supplied record is also truncated at claim 14071, so later developments are not assessed here.

Applied Digital's record splits into two unrelated eras. The 2006-2008 claims belong to the predecessor shell, Flight Safety Technologies. Only the Volpe Center award (21053) is a revenue-bearing customer contract. The auditor, proxy-solicitor and transfer-agent entries (21067, 21068, 21069) are vendor relationships, and Broadridge's roughly $22,000 is a cost, not income. Nothing from that era bears on the current business.

The 2021-2022 hosting story moves from anonymity toward identification. In August 2021 the company cited agreements with two unnamed customers covering more than half of available energy (14506). It separately said GMR and Bitmain had hosting contracts (14507), so the two were probably the same pair, though the filing never says so. By November the count was four customers, still unnamed, absorbing all of the first facility plus 85MW at a second (14412, 14374, 14363). Only in March 2022 were they identified: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (14232, 14200, 14210, 14071).

Naming did not reduce concentration; it exposed it. GMR alone holds about 50% of the first 100MW (14411, 14201, 14072). GMR, Bitmain and SparkPool are also equity investors (14505, 14377). Bitmain is at once investor, equipment supplier (14416, 14504), lead source (14417), hosting customer, and parent of Antpool, the joint-venture partner in 1.21 Gigawatts LLC (14234, 14203). The anchor customers and the cap table substantially overlap, which weakens the evidentiary value of the contracts as arm's-length demand.

Several items are not customer agreements at all. The SparkPool arrangements are advisory and services contracts (14503, 14418). Under the Coinmint co-hosting deal Applied Digital is the client, not the provider (14419). The Texas wind facility rests only on a letter of intent (14204, 14074). The Antpool vehicle is a framework joint venture sized to as much as 1.5GW, with no committed offtake disclosed.

The revenue arithmetic cannot be reconciled because the record discloses no contract values, per-MW pricing, terms or recognized revenue. Capacity is the only checkable quantity, and it shows a gap. Customers are said to cover 185MW (100MW plus 85MW), but the only power agreement on record is the 100MW Amended and Restated Energy Services Agreement (14233, 14202). The second facility's 85MW has no power contract on record. The $8,000 attached to the Antpool formation (14234) is not explained in the quoted text and should not be read as deal value. The record supplied ends in April 2022, so later counterparties are outside this read.

Applied Digital's customer record divides into two unrelated eras. The 2006–2008 claims belong to the predecessor shell, Flight Safety Technologies. Only one of them is a customer agreement: the Volpe Center award (21053). The auditor, proxy-solicitor and transfer-agent entries (21067, 21068, 21069) are vendor engagements and should not be counted as commercial traction. The substantive record begins in August 2021.

The hosting pipeline started out largely anonymous. In August 2021 the company described agreements with two unidentified customers covering more than half of available energy (14506). It also said GMR and Bitmain had hosting contracts to pre-fill the first 100MW site (14507). By November 2021 this had grown to four unnamed customers absorbing all first-facility energy plus 85MW at a second site (14412, 14374, 14363). Only one allocation was quantified: GMR at roughly 50% of the 100MW (14411). The four counterparties were not named until March 2022: JointHash Holding (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (14232, 14200, 14210, 14071). That disclosure resolves the identification gap, but it also shows how concentrated the book is. About half of first-facility capacity sits with a single counterparty. GMR and Bitmain are also equity investors and service providers (14505, 14502, 14417). An Antpool affiliate of Bitmain is a joint-venture partner in 1.21 Gigawatts LLC (14234, 14203). The customer base is therefore heavily entangled with the company's own shareholders and suppliers, which weakens the independence of the demand signal.

Much of what reads as partnership is something other than customer revenue. The SparkPool arrangements are advisory and engineering services (14503, 14418). The Bitmain 200PH order is an equipment purchase by the company (14504, 14416). The Coinmint co-hosting deal makes Applied Digital the customer, not the provider (14419). The Texas 200MW wind project is only a letter of intent (14204, 14074). The 1.21 Gigawatts entity is a framework vehicle for up to 1.5GW with a nominal formation figure (14234), not contracted load. The Amended and Restated Energy Services Agreement (14233, 14202) is supply-side power, not demand.

Revenue arithmetic cannot be reconciled from this record. No claim discloses pricing, term, take-or-pay provisions or expected contract revenue. The only checkable figures are megawatt allocations, and the evolution from 'two customers, more than 50%' to 'four customers, 100%' reflects new signings rather than any inconsistency. The record as supplied ends at claim 14074, so later disclosures may change this picture.

Applied Digital's record through April 2022 has a few signed counterparties, but most of its announced capacity was described in aggregate before any customer was named. The pre-2021 claims belong to the predecessor shell, Flight Safety Technologies. A 2006 Department of Transportation Volpe Center contract (21053) is the only customer item there. The transfer agent, auditor and Broadridge engagements (21069, 21067, 21068) are vendor relationships and have no bearing on hosting demand. In the crypto-hosting era, the August 2021 disclosure cited 'agreements with 2 customers' covering more than 50% of available energy without naming either (14506). By November this had become four unnamed customers covering all of the first facility plus 85MW at a second site (14412, 14374, 14363). Only in March 2022 were they identified: JointHash Holding (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (14232, 14210, 14200, 14071). GMR's contract for roughly 50% of the 100MW site is the one allocation given a size (14411, 14231, 14201, 14072). The Coinmint co-hosting agreement (14419, 14371, 14357) is signed, but there Applied Digital is the customer, not the provider. Counterparty concentration is high and largely circular. SparkPool, GMR and Bitmain are described as 'strategic equity investors' (14505, 14376 to 14378). GMR is also the dominant hosting customer and the company's mining adviser (14359). Bitmain is simultaneously a shareholder, equipment supplier (14504, 14416, 14360), hosting customer, and source of customer 'leads' (14417). Through its affiliate Antpool, Bitmain is also a joint-venture partner in 1.21 Gigawatts LLC (14234, 14203, 14212, 14073). Two of four named customers, holding at least half the first site, are therefore related to insiders. The softer language deserves discounting. The Texas 200MW wind facility is a letter of intent expressly 'subject to entry into definitive agreements' (14204, 14074). The Antpool vehicle is a signed LLC agreement, but its disclosed figure is $8,000 at formation (14234) against an 'up to 1.5GW' ambition. SparkPool's pool advisory is conditional on a future decision (14418). The 85MW second-facility allocation predates any disclosed energy agreement for that site. On revenue arithmetic, the record contains no contract values, pricing, terms or recognized revenue, so the announced deals cannot be reconciled to financials. The only internal check is weak. GMR alone holding about 50% sits awkwardly with the earlier 'two customers, greater than 50%' framing (14506), which implies a very small second contract. The supplied record also ends mid-claim in April 2022, so later disclosures are not assessed here.

Applied Digital's customer record sorts into three tiers of firmness, and only the middle tier is plainly signed hosting revenue. The 2006–2008 entries come from the predecessor shell, Flight Safety Technologies. The Volpe Center award (claim 21053) was a genuine government contract, but the transfer agent, auditor and Broadridge engagements (claims 21069, 21067, 21068) are vendor relationships, not customers, and say nothing about the current hosting business.

The disclosure of counterparties came in stages. In August 2021 the company cited agreements with "2 customers" covering more than half of the energy under its Energy Services Agreement without naming them (claim 14506). By November that had become "4 customers," still unidentified (claims 14412, 14374, 14363). Only in March 2022 were the four named as JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (claims 14232, 14210, 14200). GMR alone holds roughly 50% of the first 100MW facility (claims 14411, 14231). Readers should treat the earlier unnamed counts as unverifiable until that March disclosure.

Concentration is tighter than the four-customer count suggests. GMR and Bitmain are also strategic equity investors (claims 14505, 14377, 14378). Bitmain is additionally the equipment supplier (claims 14504, 14416) and a source of customer leads (claim 14417). Its affiliate Antpool is the joint-venture partner in 1.21 Gigawatts LLC (claims 14234, 14212). At least two of the four hosting customers, likely holding the majority of capacity, are therefore shareholders with overlapping supplier or partner roles. That related-party profile warrants scrutiny of pricing and enforceability.

Several headline items are not customer contracts at all. SparkPool and GMR provide advisory and services to the company (claims 14503, 14418, 14359). Under the Coinmint co-hosting agreement, Applied Digital is the customer rather than the provider (claim 14419). The 200MW Texas wind facility rests on a letter of intent that is expressly "subject to entry into definitive agreements" (claim 14074). The Antpool vehicle is framed around up to 1.5GW of prospective datacenters, which is aspirational capacity rather than contracted demand (claim 14203).

The revenue arithmetic cannot be reconciled from this record. No contract values, per-megawatt rates, terms or recognized revenue are disclosed alongside the hosting agreements. The second-site 85MW allocation (claim 14412) is also not tied to an identified facility. Diligence should therefore anchor on the named March 2022 counterparties and discount the LOI and framework capacity until definitive terms appear.

Applied Digital's disclosed counterparty record is thinner than its volume suggests, because most claims repeat the same small set of 2021-2022 relationships across successive filings. The pre-2009 entries come from the predecessor shell, Flight Safety Technologies. Only the Volpe Center contract (claim 21053) is a revenue counterparty. The transfer agent, auditor and Broadridge (claims 21067, 21068, 21069) are vendors and carry no weight as customer evidence.

The hosting story begins with deliberate anonymity. The August 2021 filing cites agreements with '2 customers' covering more than 50% of available energy (claim 14506). By November that had become '4 customers' absorbing the full first-facility allocation plus 85MW at a second site (claims 14412, 14374, 14363), still unnamed. Identification arrived only in March 2022, when the four were disclosed as JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 14232, 14200, 14210). Those agreements appear to be signed, but the filings describe them only as 'agreements' that 'will account for' energy, which reads as capacity allocation rather than disclosed contract value or term.

Concentration is high and entangled. GMR alone contracted roughly 50% of the 100MW first facility (claims 14411, 14231, 14201). GMR and Bitmain are also strategic equity investors (claims 14505, 14377, 14378), and Bitmain is simultaneously equipment supplier, lead source and customer (claims 14416, 14417, 14360). Two of the four named customers are therefore insiders by economic relationship, which weakens these contracts as arm's-length demand validation.

Several items should not be read as customer wins. The Texas 200MW wind facility is explicitly a letter of intent 'subject to entry into definitive agreements' (claims 14204, 14074). The 1.21 Gigawatts LLC with Antpool is a joint-venture vehicle, not a customer contract (claims 14234, 14203, 14073). The SparkPool arrangements are advisory and service relationships flowing to the company (claims 14503, 14418), and Coinmint is a co-hosting provider for Applied Digital's own mining (claims 14419, 14371).

Revenue arithmetic cannot be reconciled from this record. No contract values, per-MW rates, durations or recognized revenue are attached to any hosting agreement. The only dollar figure in the period is the $8,000 tied to the Antpool LLC formation (claim 14234), which is not revenue. Diligence should require disclosed pricing and terms before treating the announced 100MW-plus-85MW pipeline as revenue backlog.

Applied Digital's customer record, as captured through April 2022, is dense with repetition but thin on economics, and the distinction between signed contracts and softer arrangements matters more than the volume of claims suggests. The pre-2021 entries belong to the predecessor shell, Flight Safety Technologies, and are mostly vendor relationships (transfer agent, auditor, a roughly $22,000 proxy solicitor) rather than customers (21069, 21067, 21068). The 2006 Volpe Center award is the only genuine customer contract from that era (21053).

The hosting story begins with unnamed counterparties. In August 2021 the company cited agreements with two unidentified customers covering more than half the available energy (14506), alongside hosting contracts with GMR and Bitmain (14507). By November the count had become four customers taking all of the first facility's energy plus 85MW at a second site, still unnamed (14412, 14374, 14363). Only in March 2022 were the four identified: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (14232, 14210). That disclosure arrived months after the capacity was described as fully contracted.

Concentration was high from the start and stays high. GMR alone is repeatedly said to hold approximately 50% of the 100MW site (14411, 14231, 14201). Two of the four named customers are also strategic equity investors (14505, 14377, 14378), so roughly half or more of contracted capacity sits with shareholders. Bitmain occupies nearly every role at once: equipment supplier (14416), source of hosting leads (14417), customer, investor, and, through its affiliate Antpool, joint-venture partner in 1.21 Gigawatts LLC (14234, 14203). SparkPool supplies services and advice and is not a customer (14503, 14418). Coinmint is a co-hosting provider to the company rather than a buyer from it (14419).

Several items fall short of signed commitments. The Texas 200MW wind facility is a letter of intent that is explicitly subject to definitive agreements (14204, 14074). The 1.21 Gigawatts entity is a framework for up to 1.5GW, and the only figure disclosed is a nominal $8,000 at formation (14234).

The revenue arithmetic cannot be reconciled from this record. None of the claims discloses contract values, tenor, per-megawatt pricing or minimum-take terms. "Fully contracted" therefore describes allocated capacity, not quantifiable revenue, and that capacity is heavily concentrated among related parties.

Applied Digital's record splits into two unrelated eras. The 2006-2008 entries belong to the predecessor shell, Flight Safety Technologies. The Department of Transportation Volpe Center award is the only signed customer contract from that period (claim 21053). Wolf & Company, Pacific Stock Transfer and Broadridge are vendors the company pays, not customers (claims 21067, 21069, 21068). Broadridge's roughly $22,000 fee is an expense, not revenue.

The 2021 hosting story starts with unidentified counterparties and only names them later. In August 2021 the company cited agreements with '2 customers' covering more than half of available energy, without naming either (claim 14506). By November that had become '4 customers' absorbing all of the first facility's energy plus 85MW at a second site, still unnamed (claims 14412, 14374, 14363). The four were not named until March 2022: JointHash Holding (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 14232, 14200, 14210, 14071).

Naming them clarifies the concentration. SparkPool, GMR and Bitmain were described as 'strategic partners' and were also equity investors (claims 14505, 14378, 14377). GMR alone held a contract for about 50% of the first 100MW (claims 14411, 14211, 14072). So at least two of the four anchor customers, likely holding the majority of contracted capacity, were shareholders. The Bitmain group also sits on several sides of the company. It is a hosting customer, an equipment supplier on the 200PH order (claims 14504, 14416), a source of customer leads (claim 14417), and, through its affiliate Antpool, a joint-venture partner in 1.21 Gigawatts LLC (claims 14234, 14212). SparkPool's role is advisory and services-based, not a hosting contract (claims 14503, 14418), and its 2021 description is cut off mid-sentence (claim 14358). The Coinmint arrangement makes Applied Digital the customer of Coinmint, not the other way around (claims 14419, 14371).

Several items should not be read as signed customer revenue. The Texas 200MW wind facility is explicitly a letter of intent (claims 14204, 14074). The Antpool vehicle is a joint-venture formation recorded at $8,000, not a customer contract.

The revenue arithmetic cannot be reconciled from this record. None of the hosting agreements disclose a price per megawatt, a term, take-or-pay provisions or a total contract value. The only dollar figures are the $8,000 joint-venture entry and the $22,000 proxy fee. A capacity commitment from related-party investors is not the same as contracted revenue. The record provided also stops partway through claim 14074, so any post-April 2022 data-center or AI-hosting customers fall outside this read.

Applied Digital's customer record before its 2021 pivot says little about commercial traction. The 2006–2008 entries come from the predecessor Flight Safety Technologies. Only the Volpe Center emulation contract (21053) is a revenue counterparty. Broadridge, Wolf & Company and Pacific Stock Transfer (21068, 21067, 21069) are vendors the company paid, not customers, and should not be read as commercial validation. The substantive story begins in August 2021, and the language there needs to be sorted carefully. The company said it had 'entered into agreements with 2 customers' covering more than half its contracted energy without naming them (14506). In the same filing it said GMR and Bitmain had signed hosting contracts to pre-fill the first 100MW site (14507). By November 2021 the count had grown to four customers, still unnamed, covering all first-facility energy plus 85MW at a second site (14412, 14374, 14363). The names did not appear until March 2022: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (14232, 14200, 14210). Those are signed agreements. The Texas 200MW wind facility is different. It is a letter of intent expressly 'subject to entry into definitive agreements' (14074, 14204) and should not be counted as contracted capacity. SparkPool's role is advisory and services-only (14503, 14418), and in the Coinmint co-hosting arrangement Applied Digital is the customer rather than the provider (14419). Concentration is the central diligence issue. GMR alone accounts for roughly 50% of the 100MW first facility (14411, 14201). GMR and Bitmain are also equity investors (14505, 14377, 14378). Bitmain additionally supplies equipment (14504, 14416), refers hosting leads (14417), and sits on the other side of the 1.21 Gigawatts joint venture through its affiliate Antpool (14234, 14203). The anchor demand therefore rests heavily on related parties who hold equity, sell hardware and co-invest. That overlap weakens how independent these signings are as evidence of market demand. On revenue arithmetic, the record supplied contains no contract values, per-MW pricing, terms or recognized hosting revenue, so it cannot be reconciled against these deals. The only dollar figures are a $22,000 proxy-solicitation fee (21068) and an $8,000 figure attached to the joint-venture formation (14234), and neither speaks to customer economics. The claim list provided also ends mid-entry at 14072, so later disclosures may change this assessment.

Applied Digital's record splits into two unrelated eras. The 2006–2008 entries belong to its predecessor shell, Flight Safety Technologies. Only the Volpe Center award (claim 21053) is a customer contract; the auditor, transfer agent and proxy solicitor (21067, 21069, 21068) are vendors and should not count as commercial traction. The relevant story begins in August 2021, and the record supplied here ends at an April 2022 filing (14071).

Naming came late. In August 2021 the company said it had agreements with two customers covering more than half of its first site's energy (14506). It credited GMR and Bitmain with hosting contracts (14507), but did not tie them explicitly to that count. By November 2021 the tally was four customers absorbing all of the first facility plus 85MW at a second (14412, 14374, 14363), still unidentified. Only in March 2022 were the four named: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (14232, 14200, 14210). Those are signed agreements in the company's own wording. The 85MW second-facility allocation, however, was never broken out by counterparty. The Texas 200MW wind project is expressly a letter of intent subject to definitive agreements (14204, 14074), and should be read as pipeline, not backlog. SparkPool's role is advisory and services work (14503, 14418, 14358), not customer revenue. Coinmint is a co-hosting arrangement in which Applied Digital was the customer, not the vendor (14419).

Concentration is high and becoming more circular. GMR alone was repeatedly described as contracted for about 50% of the 100MW site (14411, 14201, 14072). GMR, Bitmain and SparkPool were all disclosed as strategic equity investors (14505, 14377). Bitmain also supplied 200PH of equipment (14504, 14416), sourced hosting leads (14417), and through its affiliate Antpool co-formed the 1.21 Gigawatts LLC joint venture targeting up to 1.5GW (14234, 14203). As a result, a large share of contracted demand traces to shareholders who are simultaneously suppliers and joint-venture partners. Diligence should test whether that demand was negotiated at arm's length.

The revenue arithmetic cannot be reconciled from this record. None of the hosting claims discloses price per kilowatt-hour, term, take-or-pay provisions or contract value. The only dollar figures are the legacy $22,000 proxy fee (21068) and an $8,000 figure attached to the 1.21 Gigawatts formation (14234), whose units and meaning the claim does not specify. Capacity was described as fully allocated, but no stated revenue can be tested against it.

Applied Digital's record holds fewer signed revenue contracts than its partnership language suggests. The earliest entries come from the predecessor shell, Flight Safety Technologies. They include one government contract, the Volpe Center wake-vortex award (claim 21053). The rest are vendors, not customers: Broadridge at about $22,000, Pacific Stock Transfer and Wolf & Company (claims 21068, 21069, 21067). None of them bears on the hosting business.

Disclosure of hosting customers moved from anonymous to named over seven months. In August 2021 the company cited agreements with '2 customers' covering more than 50% of available energy without naming them (claim 14506). In the same filing it said GMR and Bitmain had hosting contracts (claim 14507). By November 2021 the count had become '4 customers', still unnamed, absorbing all of the first facility's energy and 85MW at a second site (claims 14412, 14374, 14363). Only in March 2022 were the four identified: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 14232, 14210, 14071). These read as signed agreements, but no term, pricing or take-or-pay structure is quoted.

Several relationships that look like customers are something else. SparkPool provides advisory and engineering services, so it is a vendor (claims 14503, 14418). Coinmint hosts Applied Digital's own machines, which makes Applied Digital the customer (claim 14419). The Bitmain 200PH order is a purchase by Applied Digital (claims 14504, 14416). The 1.21 Gigawatts LLC with Antpool is a joint-venture formation with $8,000 attached, not a revenue contract (claims 14234, 14203). The 200MW Texas wind facility is explicitly a letter of intent 'subject to entry into definitive agreements' (claims 14204, 14074).

Counterparty concentration is high and tangled with related parties. GMR alone holds about 50% of the 100MW site (claims 14411, 14231, 14201). GMR and Bitmain are also strategic equity investors (claims 14505, 14378, 14377), and Antpool is a Bitmain affiliate. That means at least two of the four named customers, and the JV partner, sit inside the investor base, so arm's-length pricing should be verified rather than assumed.

The record cannot reconcile revenue. No contract values, per-MW rates or revenue figures appear in any claim, so there is no way to test whether announced capacity translates into reported revenue. The source record also ends mid-quote at claim 14072, which leaves any later disclosures unreviewed.

Applied Digital's customer record resolves into a small, tightly connected group of counterparties, and several of the named 'partners' are not customers. The pre-2021 claims belong to the predecessor shell, Flight Safety Technologies. Only the Volpe Center award (21053) is revenue-bearing. Broadridge, Pacific Stock Transfer and Wolf & Company (21068, 21069, 21067) are vendors the company paid.

The hosting pivot began with anonymity. In August 2021 the company said it had agreements with '2 customers' covering more than 50% of available energy (14506), while also saying GMR and Bitmain had signed hosting contracts to pre-fill the first 100MW site (14507). By November the count had become '4 customers' covering all of the first facility plus 85MW at a second site, still unnamed (14412, 14374, 14363). Names appeared only in March 2022: JointHash Holding (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (14232, 14210, 14200, 14071).

Concentration is the central issue. GMR alone holds roughly 50% of the 100MW (14411, 14231, 14201). GMR and Bitmain are also strategic equity investors (14505, 14377, 14378). Bitmain is also the company's equipment supplier (14504, 14416) and a source of hosting leads (14417). Its affiliate Antpool is the company's joint-venture partner in 1.21 Gigawatts LLC (14234, 14212, 14073). At least half of first-site capacity, and plausibly more, therefore sits with parties that own equity, sell hardware to the company, or share a venture with it.

Several 'partner' claims describe services or purchases, not customer contracts. SparkPool provides advisory and engineering services (14503, 14418), and its truncated disclosure that it is 'currently in the process of' something is unresolved (14358). The Coinmint agreement makes Applied Digital the customer, not the provider (14419). The 200MW Texas wind facility is explicitly a letter of intent subject to definitive agreements (14204, 14074). The Antpool vehicle's $8,000 figure reads as nominal formation capital, not contracted demand (14234).

The revenue arithmetic cannot be reconciled from this record. No claim discloses contract value, per-MW pricing, term or revenue. The capacity figures also shift over time: more than 50% in August, 100% plus 85MW by November, with the second site's energy agreement not evidenced here. The claim list supplied ends mid-record at 14072, so later contracts are outside this read.

Applied Digital's record shows its hosting customers being named late, and it also mixes vendors, suppliers, investors and joint-venture partners into the same announcement stream. The pre-2021 items come from the predecessor shell and are not commercial customers. Broadridge, Pacific Stock Transfer and Wolf & Company are service providers the company pays (claims 21068, 21069, 21067). The Volpe Center award is the only government contract in the record (21053). Several of the crypto-era claims are also not customer revenue. SparkPool and GMR supply advisory, software and asset-management services (14502, 14503, 14418). The Bitmain 200PH order is an equipment purchase (14504, 14416). Coinmint is a co-hosting arrangement in which Applied is the customer (14419). The genuine hosting-customer thread starts in August 2021 with 'agreements with 2 customers' covering more than 50% of available energy. Those customers were unnamed (14506), although the same filing says GMR and Bitmain signed hosting contracts (14507). By November 2021 the count had become four unidentified customers taking all of the first facility's energy plus 85MW at a second site (14412, 14374, 14363). The four were named only in March 2022: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (14232, 14210, 14200). Concentration is the central diligence issue. GMR alone is repeatedly described as holding roughly 50% of the 100MW capacity (14411, 14231, 14201). GMR and Bitmain are also disclosed strategic equity investors (14505, 14377, 14378). This means at least two of the four customers, and probably most of the contracted megawatts, sit with shareholders. That raises related-party and pricing-independence questions that the claims do not resolve. The forward pipeline is softer than the headline language suggests. The Texas 200MW wind project is a letter of intent explicitly 'subject to entry into definitive agreements' (14204, 14074). The 1.21 Gigawatts LLC with Antpool, a Bitmain affiliate, is a framework joint venture capitalised at a nominal $8,000 that targets up to 1.5GW (14234, 14203, 14212). It is not a signed offtake. On revenue, the record as supplied contains no revenue, contract-value or per-megawatt pricing figures. The input also truncates at claim 14072. The announced megawatt allocations therefore cannot be reconciled against reported revenue from this record, and that gap should itself be flagged.

Applied Digital's pre-2021 record belongs to its predecessor, Flight Safety Technologies. It contains one genuine customer award, the Department of Transportation Volpe Center wake-vortex contract (claim 21053). The other entries are vendors: Broadridge at roughly $22,000 for proxy solicitation (21068), Pacific Stock Transfer (21069) and Wolf & Company (21067). None of them bears on the current hosting business. The operative record begins in August 2021, when the company said it had agreements with two unnamed customers covering more than half of the energy under its Energy Services Agreement (14506). In the same filing it named GMR and Bitmain as hosting counterparties (14507). By November the count had become four customers covering all of the first facility and 85MW at a second site, still without names (14412, 14374, 14363). Identification came only in March 2022, when the four were listed as JointHash Holding (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (14232, 14210, 14071). The disclosures describe these as agreements, not LOIs, which is to the company's credit. The Texas 200MW wind facility, by contrast, is explicitly a letter of intent 'subject to entry into definitive agreements' (14074) and should not be counted as contracted capacity. Concentration is the central issue. GMR alone is repeatedly described as holding roughly 50% of the first 100MW (14411, 14231, 14201). GMR, Bitmain and SparkPool are also strategic equity investors (14505, 14377, 14378), so the anchor customers are related parties rather than arm's-length buyers. Bitmain is simultaneously a customer, the equipment supplier behind the 200PH order (14504, 14416), a source of hosting leads (14417) and, through affiliate Antpool, a joint-venture partner in 1.21 Gigawatts LLC (14234). SparkPool's role is advisory and service-based rather than customer revenue (14503, 14418), and the November 2021 language hints at a change in its status (14358). The Coinmint agreement runs the other way, with Applied Digital as the hosted party (14419). On revenue, the record contains no disclosed hosting revenue, contract value, per-MW pricing or term length. The arithmetic therefore cannot be reconciled with the announced deals, and the $8,000 figure attached to the Antpool venture (14234) is too ambiguous in units to anchor anything. The claim list provided is truncated after early April 2022 (14072), so later customer shifts are outside this read.

Applied Digital's record divides into two unrelated eras. The 2006–2008 claims come from the predecessor shell, Flight Safety Technologies, and none of them shows commercial traction for the current business. The Volpe Center award (21053) was a government development contract. Broadridge, Pacific Stock Transfer and Wolf & Company (21068, 21069, 21067) are vendors the company paid, not customers. The substantive diligence question starts in August 2021.

The hosting counterparties were identified gradually, and that is the main pattern in the disclosures. The first filing said the company had agreements with "2 customers" covering more than 50% of available energy and did not name them in that sentence (14506), although it separately named GMR and Bitmain as hosting contract holders (14507). By November 2021 the count had risen to "4 customers" covering all of the first facility plus 85MW at an unidentified second site, still without names (14412, 14374, 14363). Only in March 2022 were the four named: JointHash (a GMR subsidiary), Bitmain, F2Pool and Hashing LLC (14232, 14210, 14200). The customer count doubled across two filings before any new name was disclosed.

Concentration was severe from the start and has not improved on this record. GMR alone is described as contracting for roughly half of the 100MW (14411, 14231, 14201). The customer base is also circular. GMR, Bitmain and SparkPool are each equity investors (14505, 14377, 14378). Bitmain is at once a shareholder, the equipment supplier on a 200PH order (14504, 14416), a hosting customer, a lead source (14417), and, through its affiliate Antpool, a joint-venture partner in 1.21 Gigawatts LLC (14234, 14212). SparkPool's agreements are services and advisory arrangements, not revenue contracts (14503, 14418). In the Coinmint arrangement Applied Digital is the hosted party, not the host (14419, 14357).

Not everything announced is signed. The Texas 200MW wind facility is a letter of intent that is expressly "subject to entry into definitive agreements" (14074, 14204). The Antpool vehicle is framed as targeting up to 1.5GW (14203), but its disclosed capitalization is $8,000 (14234). That describes a formation step, not committed capacity.

The revenue arithmetic cannot be reconciled. None of the claims discloses contract values, per-MW pricing, term length or recognized hosting revenue, so the announced megawatt allocations cannot be tied to any reported figure. The record also cuts off mid-claim at 14072. Until pricing and term disclosures appear, the "fully contracted" framing rests on allocation counts, not on dollars.

The record splits into two unrelated eras. The 2006–2008 claims belong to the predecessor shell, Flight Safety Technologies. Only the Volpe Center award (21053) is a customer contract. Broadridge (21068), Wolf & Company (21067) and Pacific Stock Transfer (21069) are vendors the company pays, and the $22,000 Broadridge figure is an expense, not revenue. The substantive history begins in August 2021, when the company said it had agreements with '2 customers' covering more than half of its energy (14506) without naming them. The same filing credited GMR and Bitmain with hosting contracts (14507). By November 2021 the count had risen to four unnamed customers covering the entire first facility plus 85MW at a second (14412, 14374, 14363). The four were identified only in the March 2022 filings: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (14232, 14200, 14210). For roughly seven months, then, the customer book was described by headcount rather than by name. Several relationships are not customer contracts at all. SparkPool is a services and advisory counterparty (14503, 14418), and 14358 hints that its status was changing. Coinmint is a host that the company itself paid (14419). The Texas wind facility is a letter of intent explicitly 'subject to entry into definitive agreements' (14204, 14074). The Antpool vehicle, 1.21 Gigawatts LLC, is a joint-venture shell rather than a revenue contract (14234, 14203). Concentration is heavy and circular. GMR alone holds about 50% of the 100MW first site (14411, 14231, 14201). GMR, Bitmain and SparkPool are also equity investors (14505, 14378). Bitmain appears at once as equipment supplier, source of customer leads, customer and, through its affiliate Antpool, joint-venture partner (14417, 14416). That means much of the contracted demand comes from shareholders and suppliers rather than arm's-length buyers. Revenue arithmetic cannot be reconciled from this record. No contract values, tenors, pricing or recognized revenue are disclosed for any hosting agreement, only megawatt allocations. Those allocations cite a 100MW Energy Services Agreement for the first facility (14233, 14202), but no equivalent power agreement is on record for the 85MW said to be committed at the second site. The record as supplied ends mid-claim at 14071, so later disclosures are not reflected here.

Applied Digital's record through April 2022 contains few signed customer contracts and a great deal of partnership and framework language. The 2006-2008 entries belong to the predecessor Flight Safety Technologies. Apart from the Department of Transportation Volpe Center award (21053), they describe vendors the company paid, not customers: Broadridge (21068), Wolf & Company (21067) and Pacific Stock Transfer (21069). Several 2021 relationships are also inbound services rather than revenue. SparkPool's role is advisory and engineering (14503, 14418). GMR provides asset-management and hedging services (14502). At Coinmint, Applied Digital was itself the hosted client (14419).

Customer identification arrived late. In August 2021 the filing cites agreements with "2 customers" covering more than half of available energy without naming them (14506), though an adjacent passage names GMR and Bitmain as hosting counterparties (14507). By November 2021 this had become "4 customers" absorbing all of the first facility's energy plus 85MW at a second site, again unnamed (14412, 14374, 14363). Only in March 2022 are the four identified: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (14232, 14200, 14210). The 85MW second-site allocation is never tied to a named facility or an executed energy agreement in the record.

Concentration is high and circular. GMR alone is said to hold roughly half of the 100MW first facility (14411, 14231, 14201). Bitmain is at once an equity investor (14505), the equipment supplier on a 200PH order (14504, 14416), a source of customer leads (14417), a hosting customer (14232), and, through its affiliate Antpool, the company's partner in 1.21 Gigawatts LLC (14234, 14203). The customer base therefore overlaps heavily with the strategic investor group. The 1.21 Gigawatts agreement is executed, but its "up to 1.5GW" scope is framework language, and the only figure disclosed is an $8,000 formation amount (14234). The 200MW Texas wind project is expressly a letter of intent "subject to entry into definitive agreements" (14204, 14074).

The revenue arithmetic cannot be reconciled because none of the hosting agreements disclose pricing, term, take-or-pay provisions or contract value. The only dollar amounts on record are a vendor fee (21068) and a nominal joint-venture capitalization (14234). The claimed capacity coverage is therefore unverifiable against revenue. Note also that the record provided is truncated at claim 14071.

Applied Digital's record splits into two unrelated eras, and only the second bears on the current business. The 2006-2008 entries come from the predecessor Flight Safety Technologies. The Volpe Center award (21053) was a genuine government contract, but Broadridge (21068), Pacific Stock Transfer (21069) and Wolf & Company (21067) are vendors the company paid, not customers.

The 2021 pivot to crypto hosting is where the distinction matters. In August 2021 the company said it had agreements with two customers covering more than half of available energy but named neither (14506). In the same filing it said GMR and Bitmain had hosting contracts that would pre-fill the first 100MW site (14507). By November the unnamed count had risen to four customers, covering the full first facility plus 85MW at a second (14412, 14374, 14363). Only in March 2022 were the four identified: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (14232, 14210). GMR alone was described as holding roughly 50% of the 100MW (14411, 14201).

Several named relationships are not customer revenue at all. SparkPool is an adviser and service provider (14503, 14418). Coinmint hosted Applied's own machines, so Applied was the customer (14419). The 200PH Bitmain order is a purchase (14504, 14416). The 1.21 Gigawatts LLC with Antpool, a Bitmain affiliate, is a framework joint venture for up to 1.5GW. The only disclosed figure attached to it is $8,000 (14234). The Texas 200MW wind project is expressly a letter of intent subject to definitive agreements (14204, 14074).

Concentration is the central diligence issue. GMR and Bitmain are simultaneously strategic equity investors (14505, 14377, 14378), hosting customers, and in Bitmain's case the equipment supplier and the source of hosting leads (14417). Through JointHash and Bitmain, the investor group likely controls well over half of contracted capacity. That makes the customer base both thin and related-party in character.

The revenue arithmetic cannot be reconciled from this record. No hosting agreement discloses price per kWh, term, minimum commitments or contract value. The disclosures are expressed in megawatts and share of available energy, not dollars. The only dollar amounts on record are a $22,000 proxy fee and an $8,000 JV formation figure, and neither is customer revenue. Capacity allocation should be treated as signed but unquantified until filings attach economic terms. The supplied record ends in April 2022 and truncates mid-claim (14072), so later counterparties fall outside this read.

Applied Digital's customer record divides cleanly into two eras, and only the second concerns customers. The 2006–2008 claims come from the Flight Safety Technologies predecessor. Only the Volpe Center award (21053) is a revenue contract. Broadridge (21068), Pacific Stock Transfer (21069) and Wolf & Company (21067) are vendors the company paid, not counterparties that paid it.

The hosting-era disclosures show a clear progression from unnamed to named. In August 2021 the company cited agreements with "2 customers" covering more than half of available energy, without naming them (14506). It separately said GMR and Bitmain held hosting contracts that let it "pre-fill" a site not yet built (14507). By November 2021 the count had grown to four unnamed customers covering the first facility plus 85MW at a second (14412, 14374, 14363). Only in March 2022 were all four named: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (14232, 14210). These are signed agreements, but they cover capacity at facilities still under development.

Much of the surrounding language does not describe customers at all. SparkPool's role is advisory and procurement (14503, 14418). The Bitmain 200PH order is a purchase by Applied Digital (14504, 14416). In the Coinmint co-hosting deal Applied Digital is the customer (14419). The Energy Services Agreement is supply-side (14233). The Texas wind project is a letter of intent expressly "subject to entry into definitive agreements" (14074, 14204). The 1.21 Gigawatts entity with Antpool is a joint venture formed for $8,000 (14234), not a revenue relationship.

Concentration is the central diligence issue. GMR alone contracted for roughly 50% of the 100MW first site (14411, 14231, 14201). GMR and Bitmain are also strategic equity investors (14505). Bitmain is at once investor, equipment supplier, hosting customer, lead source (14417) and, through Antpool, JV partner. Half or more of contracted capacity therefore sits with shareholders, which gives the book a related-party character.

Revenue arithmetic cannot be reconciled from this record. No claim discloses contract value, pricing per megawatt, term or recognized revenue. The only dollar figures, $22,000 (21068) and $8,000 (14234), are costs. The supplied record ends in April 2022 (claim 14072 is truncated), so later developments are not assessed here.

Applied Digital's disclosure record separates into a legacy layer and a hosting-era layer, and only the second bears on commercial diligence. The 2006 Volpe Center award (claim 21053) was a government development contract of Flight Safety Technologies, the predecessor entity. The 2008 items are service providers the company paid, not customers: its auditor (21067), transfer agent (21069) and a proxy solicitor at about $22,000 (21068).

The hosting pivot shows how customer language firmed up over time. In August 2021 the company cited agreements with two unnamed customers covering more than half of available energy (14506). It also said GMR and Bitmain had hosting contracts that would pre-fill the first 100MW site (14507). By November 2021 this had become four unnamed customers covering the whole first facility plus 85MW of a second (14412, 14374, 14363). Only in March 2022 were the four identified as JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (14232, 14210, 14200). These are described as entered agreements, not letters of intent. By contrast, the Texas 200MW wind facility is expressly a letter of intent subject to definitive agreements (14204, 14074). The SparkPool relationship is a services and advisory arrangement (14503, 14418), and the Coinmint agreement is one where Applied Digital was itself the hosted party (14419).

Concentration is the central issue. GMR alone holds roughly 50% of the 100MW capacity (14411, 14231, 14201). GMR, Bitmain and SparkPool are all equity investors (14505, 14377, 14378). Bitmain is at once a customer, the equipment supplier on the 200PH order (14504, 14416), a source of customer leads (14417), and, through its affiliate Antpool, a joint venture partner in 1.21 Gigawatts LLC (14234, 14212). At least half of contracted capacity therefore sits with shareholders, which raises related-party and arm's-length pricing questions that the record does not address. The 'strategic partner' label also blurs supplier, investor and customer roles.

Revenue arithmetic cannot be reconciled from this record. No contract values, per-megawatt rates, terms or recognized hosting revenue are disclosed. The only dollar figures are the legacy proxy fee and an $8,000 amount attached to the Antpool joint venture formation (14234). The claim set also ends mid-quote at 14071, so later disclosures are not assessed here.

Applied Digital's record separates into two unrelated eras. The 2006-2008 entries come from the Flight Safety Technologies predecessor. Only the Volpe Center award (claim 21053) is a customer contract, and it is a development-and-demonstration engagement. The Broadridge, Pacific Stock Transfer and Wolf & Company items (claims 21068, 21069, 21067) are vendors the company paid, not customers, and should carry no weight in any read of commercial traction.

The hosting-era disclosures show a clear progression from anonymous to named counterparties. In August 2021 the company described agreements with two customers accounting for more than half of available energy without naming them (claim 14506). In the same filing it credited GMR and Bitmain with hosting contracts for the first 100MW site (claim 14507). By November 2021 the count had become four unidentified customers covering the full first facility plus 85MW at a second (claims 14412, 14374, 14363). Only in March 2022 were the four identified: JointHash Holding (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 14232, 14210, 14071). This means that for roughly seven months investors were asked to credit 'customers' they could not verify.

Concentration is high and heavily related-party. GMR alone is repeatedly described as holding roughly 50% of the 100MW capacity (claims 14411, 14231, 14201). GMR and Bitmain are also equity investors in the company (claims 14505, 14377, 14378), and Bitmain is simultaneously a supplier (claims 14504, 14416) and a source of customer leads (claim 14417). Two of four named customers are therefore insiders, and they likely account for most of the contracted megawatts.

Several headline items are not customer agreements. SparkPool's role is advisory and services (claims 14503, 14418). The Coinmint arrangement makes Applied Digital the hosted party, not the host (claim 14419). The Texas 200MW wind project is expressly a letter of intent (claims 14204, 14074). The 1.21 Gigawatts LLC with Bitmain affiliate Antpool is a joint-venture formation agreement rather than a revenue contract (claims 14234, 14212), and the $8,000 figure attached to it is not explained by the quoted text.

On revenue arithmetic, none of the claims on record, which end in April 2022, disclose contract values, hosting rates, terms or recognized revenue. Reconciliation is therefore not possible from this record. The megawatt allocations are internally consistent (two customers exceeding 50%, later four at 100%), but they indicate capacity commitments rather than dollars, and whether the related-party contracts are priced at arm's length remains undisclosed.

Applied Digital's record divides into a pre-2021 shell history and a 2021-2022 hosting build-out. The early entries are not commercial customers. The Volpe Center award (claim 21053) was a government development contract held by predecessor Flight Safety Technologies. Broadridge, Wolf & Company and Pacific Stock Transfer (claims 21068, 21067, 21069) are vendors the company paid, not counterparties that paid it. None of them bears on the current business.

The hosting-era disclosures moved from anonymous to named, and the order matters. In August 2021 the company cited agreements with "2 customers" covering more than 50% of available energy, without naming them (claim 14506). A parallel passage credited GMR and Bitmain with hosting contracts that would pre-fill the first 100MW site (claim 14507). By November the unnamed count had grown to four customers taking all first-facility energy plus 85MW at a second site (claims 14412, 14374, 14363). Only the March 2022 filings named them: JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (claims 14232, 14200, 14210, 14071). The signed agreements therefore appear to be real, but for roughly seven months the company described them before identifying who had signed.

Concentration is the more material point. GMR alone is repeatedly credited with about 50% of the 100MW first facility (claims 14411, 14231, 14201, 14072). GMR and Bitmain are also disclosed strategic equity investors (claims 14505, 14377, 14378), so the two largest hosting counterparties are also shareholders. The January 2022 1.21 Gigawatts JV was formed with Antpool, a Bitmain affiliate (claims 14234, 14203). That deepens exposure to a single group across equity, customer, equipment supply (claims 14504, 14416) and development roles. SparkPool's role is advisory and services, not revenue (claims 14503, 14418). The Coinmint arrangement makes Applied Digital the customer, not the provider (claims 14419, 14371).

The Texas project should be treated as pipeline only. It rests on a November 2021 letter of intent that is expressly "subject to entry into definitive agreements" (claims 14204, 14074).

On revenue arithmetic, the record supplies no contract values, per-MW pricing, terms or recognized revenue for any hosting agreement. Megawatt allocations therefore cannot be reconciled to reported revenue from this data. The claim set is also truncated after claim 14073, so later developments are not assessed here.

Applied Digital's disclosed hosting book is thinner and more interlocked than its customer count suggests. In August 2021 the company described agreements with two unnamed customers covering more than 50% of available energy (claim 14506), alongside hosting contracts with GMR and Bitmain meant to pre-fill the first 100MW site (claim 14507). By November 2021 the count had become four customers, still unnamed, said to absorb all of the first facility's energy plus 85MW at a second (claims 14412, 14374, 14363). Only in the March 2022 filings are the four identified: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14210, 14200, 14071). That naming is the first point at which the signed-customer claim can be tested against identifiable counterparties. The move from two to four customers happened while the counterparties were still undisclosed.

Concentration is severe, and it overlaps with related parties. GMR, through JointHash, holds roughly 50% of the 100MW capacity (claims 14411, 14231, 14201). GMR and Bitmain are also strategic equity investors (claim 14505). Bitmain serves at the same time as equipment supplier through the 200PH order (claims 14504, 14416), as a lead source for hosting customers (claim 14417), and as a customer. Its affiliate Antpool is the joint-venture partner in 1.21 Gigawatts LLC (claims 14234, 14203). At least two of the four named customers are therefore shareholders or affiliates of shareholders. The two arm's-length names, F2Pool and Hashing LLC, have no disclosed capacity split. SparkPool is framed as a partner, but the record shows only advisory and services terms and no hosting contract (claims 14503, 14418, 14358). The Coinmint arrangement runs the other way, with Applied Digital as the hosted party (claim 14419).

The record contains some non-binding language. The Texas 200MW wind site is a letter of intent, explicitly subject to definitive agreements (claims 14204, 14074). The 1.21 Gigawatts entity is an LLC agreement that frames up to 1.5GW of development, not contracted load (claim 14203). No energy services agreement is disclosed for the 85MW second-facility allocation. Only the 100MW Amended and Restated ESA is cited (claims 14202, 14233).

Revenue arithmetic cannot be reconciled from this record. No contract values, rates, terms or recognized hosting revenue appear. The only dollar figure is an $8,000 reference attached to the Antpool LLC formation (claim 14234), which reads as a capitalization amount, not revenue. The claim set provided is also truncated after claim 14060, so later developments are not assessed here.

Applied Digital's early customer record is thinner than its volume of claims suggests, because most filings repeat the same small set of relationships. The first disclosure (claim 14506) cites agreements with two unnamed customers covering more than half of the available energy. By November 2021 this grew to four unnamed customers taking all energy at the first facility plus 85MW at a second site (claims 14412, 14374, 14363). The counterparties were only named in March 2022: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (claims 14232, 14200, 14210). For roughly seven months, then, the contracted-capacity narrative rested on anonymous customers.

What was actually signed differs from what was only proposed. The hosting agreements, the GMR contract for about 50% of the 100MW site (claims 14411, 14231), the Coinmint co-hosting arrangement (claim 14419) and the Amended and Restated Energy Services Agreement (claim 14233) are all described as entered into. Coinmint, however, is a facility where Applied is the customer, not a revenue source. The Texas wind project is only a letter of intent, expressly subject to definitive agreements (claims 14204, 14074). The SparkPool relationships are advisory and services arrangements, not customer contracts (claims 14503, 14418). The 1.21 Gigawatts LLC with Antpool (claims 14234, 14203) is a joint-venture agreement for future capacity, not a hosting commitment.

Concentration is high and circular. GMR alone is said to hold about half of the first site (claim 14201). GMR and Bitmain are each strategic equity investors (claim 14505). Bitmain is also the equipment vendor for the 200PH order (claim 14416), a source of hosting leads (claim 14417), a named hosting customer, and, through its affiliate Antpool, Applied's JV partner (claim 14212). Much of the early customer base overlaps with the shareholder register, so these contracts should be treated as related-party support rather than arm's-length demand.

The revenue side cannot be reconciled from this record. None of the claims discloses contract values, pricing per MW, term length or recognized revenue. The only number, the $8,000 attached to the Antpool LLC in claim 14234, is an unexplained figure, not deal economics. Capacity allocations are the only quantitative anchor, and they do not convert into revenue without pricing. The supplied record also cuts off mid-claim at 14060, so any later filings that name new customers or disclose revenue fall outside this read.

Applied Digital's early record is thinner on signed customers than its partnership language suggests. The firmest disclosures are hosting agreements. In August 2021 the company described agreements with two unnamed customers covering more than half the energy under its Energy Services Agreement (claim 14506). In the same filing it said GMR and Bitmain had signed hosting contracts intended to fill the first 100MW site (claim 14507). By November 2021 this had become four unnamed customers covering all energy at the first facility plus 85MW at a second (claims 14412, 14374, 14363). The four were only named in March 2022 as JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14200, 14071, 14060). That leaves roughly seven months in which 'customers' were counted but not identified. Several other relationships are not customer revenue at all. SparkPool's role is advisory and technical (claims 14503, 14418). GMR's services agreement runs toward Applied Digital, not from it (claim 14502). The Bitmain 200PH order is a purchase (claims 14504, 14416), and Coinmint is Applied Digital's own co-hosting provider (claim 14419). The Texas 200MW wind project is expressly a letter of intent subject to definitive agreements (claims 14204, 14074). The 1.21 Gigawatts LLC with Antpool is a joint-venture vehicle rather than a hosting contract (claims 14234, 14203). Concentration is severe and has barely diluted. GMR alone was repeatedly cited at about 50% of the 100MW capacity (claims 14411, 14231, 14201). GMR, Bitmain and SparkPool are also strategic equity investors (claim 14505), and Antpool is a Bitmain affiliate. That means at least half the named customer base, and likely most contracted megawatts, sits with shareholders or their affiliates. Growth from two customers to four came from counterparties already inside the investor circle, plus F2Pool and Hashing. On revenue, the record supplied contains no disclosed hosting revenue, contracted dollar values, pricing per megawatt or contract tenors. Reconciliation with the announced deals therefore cannot be performed from these claims. What can be said is that the contracted-capacity arithmetic is internally consistent: GMR's roughly 50MW plus three other customers filling 100MW. The disclosures also consistently describe the hosting operation as 'under development' (claim 14061), so any revenue attached to these agreements would have been prospective at the time. Readers should treat the 100MW as contracted but related-party-heavy. The Texas and Antpool capacity should be treated as unsigned or structural rather than customer demand.

Applied Digital's early customer record shows a small set of signed hosting contracts, but the disclosed counterparties are also the company's investors and suppliers. The first filings (August 2021) describe agreements with "2 customers" covering more than 50% of available energy without naming them (claim 14506), then name GMR and Bitmain as hosting counterparties for the first 100MW site (claim 14507). By November 2021 the count had become four unnamed customers covering all of the first facility plus 85MW at a second (claims 14412, 14374, 14363). Those four were only identified in the March 2022 filings as JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (claims 14232, 14200, 14210, 14071, 14060). So for roughly seven months, most of the contracted base was described but not identified. Once the names appeared, they did not show a diversified book.

Concentration is high and self-referential. GMR alone is repeatedly credited with approximately 50% of the 100MW capacity (claims 14411, 14231, 14201, 14061). GMR, Bitmain and SparkPool are all "strategic equity investors" (claims 14505, 14376). Bitmain is at once a shareholder, the equipment vendor on a 200PH order (claims 14504, 14416), a source of "leads for potential hosting customers" (claim 14417), and a hosting customer itself. It is also, through its affiliate Antpool, the counterparty in the 1.21 Gigawatts LLC joint venture (claims 14234, 14212, 14203). SparkPool's role is advisory, not revenue-bearing (claims 14503, 14418). The Coinmint arrangement makes Applied Digital the hosted party, not the host (claim 14419). Diligence should treat at least half of contracted capacity as related-party.

The softer language matters. The Texas 200MW wind facility is a letter of intent "subject to entry into definitive agreements" (claims 14204, 14074). The Antpool joint venture is a framework for up to 1.5GW of datacenters, not a customer commitment. Neither should be counted as contracted capacity.

On revenue, the record contains no disclosed pricing, contract terms, tenor or revenue figures tied to these agreements. The "$8,000" in the heading of claim 14234 appears to relate to the joint venture, not to hosting revenue. That means the arithmetic cannot be reconciled from these claims. Megawatt allocations (100MW plus 85MW) are the only quantitative anchor, and converting them to revenue requires terms the filings on record do not provide.

Applied Digital's early customer record is more concrete than most hosting-stage disclosures, but it took roughly seven months for the counterparties to be fully named, and nearly all of them are also investors or suppliers. In August 2021 the company described agreements with two unnamed customers that would take more than half the available energy (claim 14506). In the same filing it separately identified GMR and Bitmain as hosting-contract parties (claim 14507). By November 2021 the count had risen to four customers covering the full first facility plus 85MW at a second site, still unnamed (claims 14412, 14374, 14363). Only in March 2022 did the filings identify all four: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 14232, 14200, 14210). That March 2022 disclosure, repeated in claims 14071 and 14060, is the strongest signed-customer evidence on the record. Much of the remaining partnership language is services, supply or preliminary arrangements rather than hosting revenue. SparkPool's role is engineering and software support (claims 14503, 14358). Its Ether-pool arrangement is advisory and contingent on a future company decision (claim 14418). Bitmain's 'preferred access' and 'leads' are not contracted revenue (claim 14417). The Coinmint agreement makes Applied Digital the co-hosting customer, not the host (claim 14419). The Texas wind facility is explicitly a letter of intent, subject to definitive agreements (claims 14204, 14074). The 1.21 Gigawatts LLC with Antpool is a joint-venture formation, not a customer contract (claims 14234, 14212, 14203). Counterparty concentration is high and circular. GMR alone contracted for about 50% of the 100MW first site (claims 14411, 14231, 14201). SparkPool, GMR and Bitmain are all strategic equity investors (claims 14505, 14376–14378). Antpool is a Bitmain affiliate, so Bitmain appears as supplier, customer, investor and, through Antpool, joint-venture partner. Customer diversification therefore depends mainly on F2Pool and Hashing LLC, whose individual allocations are not disclosed. The revenue arithmetic cannot be reconciled from this record. None of the claims gives contract values, pricing per MW, term lengths or recognized hosting revenue. Only megawatt allocations are given, and the energy agreement cited covers only the first 100MW (claims 14233, 14202), leaving the 85MW second-site allocation without a matching disclosed power agreement. The claim list supplied also ends mid-record at claim 14061, so later filings may change this picture.

Applied Digital's early customer record is more concrete than most pre-revenue hosting stories, but the named counterparties are closely tied to the company as equity investors, and the filings give no revenue figures to test against. In the August 2021 materials the company described agreements with two customers expected to take more than half of the available energy (claim 14506) without naming them. It separately said GMR and Bitmain had hosting contracts that would pre-fill the first 100MW site (claim 14507). By November 2021 the count had grown to four customers covering all of the first facility's energy plus 85MW at a second site (claims 14412, 14374, 14363), still unidentified. Names first appear in the March 2022 filings: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14200, 14210, 14071, 14060). That disclosure is a real improvement. However, two of the four customers are affiliates of strategic equity investors (claims 14505, 14376, 14377, 14378), so the contracted base is partly related-party demand rather than arm's-length market validation.

Concentration has stayed high throughout. GMR is repeatedly described as holding roughly 50% of the first 100MW (claims 14411, 14231, 14201, 14211, 14072, 14061), and Bitmain's affiliate Antpool is the joint-venture partner in 1.21 Gigawatts LLC (claims 14234, 14212, 14203). A large share of both capacity and future expansion therefore rests on one investor-affiliated group. The SparkPool relationship is a services and advisory arrangement, not a customer contract (claims 14503, 14418, 14358). The truncated November 2021 language about SparkPool being 'in the process of' some change deserves scrutiny as a possible wind-down. The Coinmint co-hosting agreement (claims 14419, 14371, 14357) runs the other way: Applied Digital is the customer. The Texas 200MW wind facility is only a letter of intent, expressly subject to definitive agreements (claims 14204, 14074), and should not count as contracted capacity.

On arithmetic, the record provides capacity allocations but no contract values, pricing, terms or recognized revenue. The one dollar figure is an unexplained $8,000 attached to the Antpool LLC formation (claim 14234). The progression from two customers above 50% to four customers at 100% is internally consistent, with GMR at about half. Whether these megawatts turn into the implied hosting revenue cannot be reconciled from the claims on record.

Applied Digital's customer record through April 2022 splits into two groups: hosting agreements the company describes as signed, and a set of equity, services, and development arrangements that are easily mistaken for revenue relationships but are not.

The firmest disclosures are the hosting contracts. In August 2021 the company cited agreements with two unnamed customers covering more than half of available energy (claim 14506) and said GMR and Bitmain had signed hosting contracts to pre-fill the first 100MW site (claim 14507). By November the count had grown to four customers, still unnamed, covering the full first-facility allocation plus 85MW at a second site (claims 14412, 14374, 14363). Only in the March 2022 filings were all four identified: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies, F2Pool Mining, and Hashing LLC (claims 14232, 14210, 14200). The one quantified allocation, GMR at roughly 50% of the 100MW capacity, is repeated consistently (claims 14411, 14231, 14201, 14072, 14061).

Concentration is high and closely tied to insiders. Half of the first facility sits with a single counterparty, and two of the four named customers, GMR and Bitmain, are also strategic equity investors (claims 14505, 14378, 14377). Bitmain holds a further set of roles. It is the company's equipment supplier through a 200PH order (claims 14504, 14416), a source of customer leads (claim 14417), and, through its affiliate Antpool, a joint-venture partner (claim 14234). SparkPool, the third equity investor, appears only as a services and advisory provider, never as a hosting customer (claims 14503, 14418, 14358). Coinmint is a supplier relationship in which Applied Digital is the customer (claim 14419).

Several headline items are not customer contracts at all. The Texas 200MW wind project is a letter of intent expressly subject to definitive agreements (claims 14204, 14074). The 1.21 Gigawatts LLC with Antpool, framed as supporting up to 1.5GW of datacenters, is a joint-venture framework, not an offtake commitment (claims 14203, 14212). The Amended and Restated Energy Services Agreement is on the supply side (claims 14233, 14202).

The revenue arithmetic cannot be reconciled from this record. None of the claims disclose pricing, contract term, take-or-pay provisions, or revenue attributable to any customer. The 85MW second-facility allocation is also not matched by any disclosed energy agreement for that site. Readers should treat capacity-coverage statements as allocation claims, not as evidence of contracted revenue.

Applied Digital's early customer record is more concrete than most pre-revenue hosting stories, but the named counterparties were also its own strategic investors. In August 2021 the company said it had agreements with two unnamed customers covering more than half of the energy under its Energy Services Agreement (claim 14506). In the same filing it named GMR and Bitmain as hosting contract holders for the first 100MW site (claim 14507). By November 2021 the count had risen to four customers, still unnamed, said to absorb all first-facility energy plus 85MW at a second site (claims 14412, 14374, 14363). Only in March 2022 were the four identified: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14210). That gap of roughly seven months between the headcount and the names matters. GMR alone was described as holding about 50% of the 100MW (claims 14411, 14231, 14201). GMR, Bitmain and SparkPool were each strategic equity investors (claims 14505, 14377, 14378). So at least half of contracted capacity, and probably more once Bitmain's share is counted, sat with shareholder-affiliated counterparties. That is a related-party concentration rather than arm's-length demand. Several items should not be read as customer revenue. The Texas 200MW wind project was a letter of intent expressly subject to definitive agreements (claims 14204, 14074). The 1.21 Gigawatts LLC with Antpool, a Bitmain affiliate, is a joint-venture framework for future datacenters with a nominal initial capitalization (claims 14234, 14212). It deepens Bitmain exposure rather than diversifying it. The 200PH Bitmain rig order is a supplier purchase (claims 14504, 14416), and the Coinmint co-hosting agreement makes Applied the customer, not the vendor (claims 14419, 14371). SparkPool's agreements are advisory and engineering services (claims 14503, 14418). The Energy Services Agreement is power supply, not demand (claim 14233). On revenue arithmetic, the record discloses no contract pricing, term, take-or-pay provisions or revenue guidance tied to these agreements. The megawatt allocations therefore cannot be reconciled to any dollar figure. The 85MW second-site allocation (claim 14412) was never attributed to specific customers in the claims on record. Readers should treat the four signed hosting agreements as real but concentrated in investor-affiliated counterparties, and treat the LOI and JV as optionality.

Applied Digital's early customer record is built largely on counterparties that are also its equity investors and service providers, and the naming of those customers came well after the capacity claims. The first disclosure, in August 2021, said agreements with two unnamed customers would take more than half of the available energy under the Energy Services Agreement (claim 14506). A separate passage said GMR and Bitmain had hosting contracts that would pre-fill the first planned 100MW site (claim 14507). By November 2021 this had grown to four unnamed customers covering the full first facility and 85MW at a second (claims 14412, 14374, 14363). Only in March 2022 were the four identified: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14200, 14210). That disclosure was repeated verbatim through April 2022 (claims 14071, 14060).

Concentration is high and related-party. GMR alone is described as holding roughly 50% of the 100MW capacity (claims 14411, 14231, 14201). GMR, Bitmain and SparkPool are all strategic equity investors (claims 14505, 14377, 14378). Bitmain is also the equipment vendor for a 200PH order (claims 14504, 14416) and a source of hosting leads (claim 14417). Its affiliate Antpool is the joint-venture partner in 1.21 Gigawatts LLC (claims 14234, 14203, 14073). The SparkPool arrangements are advisory and services agreements rather than hosting revenue (claims 14503, 14418). The November 2021 description of SparkPool is truncated at "currently in the process of s" (claim 14358), so its status cannot be assessed from the record. The Coinmint co-hosting agreement runs the other way: Applied Digital is the hosted party (claims 14419, 14371).

The binding terms vary by agreement. The Texas 200MW wind facility is explicitly a letter of intent "subject to entry into definitive agreements" (claims 14204, 14074) and should not be counted as contracted capacity. The Antpool LLC agreement is a signed formation document, but it establishes a development vehicle, not a customer.

No contracted revenue, pricing, term length or megawatt-hour rates appear in any of these claims. The announced deals therefore cannot be reconciled to revenue. The only figure attached is an $8,000 amount in the title of claim 14234, and the truncated quote does not show what it refers to. Allocating 185MW across four named hosts is a capacity statement, not a revenue one.

Applied Digital's customer record through April 2022 is almost entirely hosting capacity pre-sold to its own strategic equity investors. The signed-agreement language is firmer than the disclosure behind it.

In August 2021 the company said it had agreements with "2 customers" covering more than 50% of available energy under its Energy Services Agreement, without naming them (claim 14506). In the same filing it described GMR and Bitmain as having entered hosting contracts to pre-fill the first 100MW site (claim 14507). By November the unnamed count had risen to four customers, covering the full first facility plus 85MW at a second (claims 14412, 14374, 14363). Those four were not identified until March 2022: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14200, 14210). For roughly seven months, then, the headline contracted-capacity claim rested on anonymous counterparties.

Concentration is the more material issue. GMR alone is described as holding a contract for approximately 50% of the 100MW first facility (claims 14411, 14231, 14201). GMR, Bitmain and SparkPool are all strategic equity investors (claims 14505, 14378). Bitmain also plays several other roles. It is an equipment supplier on a 200PH order (claims 14504, 14416), a source of customer "leads" (claim 14417), a named hosting customer, and, through its affiliate Antpool, a partner in the 1.21 Gigawatts LLC joint venture (claims 14234, 14212). At least half of first-site capacity, and likely more, therefore sits with parties that also own equity, supply hardware or co-invest. That makes the arm's-length quality of these contracts a legitimate diligence question.

Several items should not be read as customer revenue at all. The SparkPool and GMR arrangements are services and advisory agreements flowing to Applied Digital (claims 14503, 14418, 14359). Under the Coinmint co-hosting agreement Applied was the customer (claim 14419). The Texas 200MW wind facility is a letter of intent explicitly "subject to entry into definitive agreements" (claims 14204, 14074). The Antpool vehicle is a joint venture, not a hosting contract.

The revenue arithmetic cannot be reconciled from this record. None of the claims discloses contract values, per-MW pricing, tenor, take-or-pay terms or recognized revenue. The "$8,000" figure attached to the 1.21 Gigawatts claim appears only in a label, and the excerpt does not establish its units (claim 14234). The contracted megawatts are asserted, but the dollars they imply are not.

Applied Digital's early customer record is more concrete than many hosting-stage disclosures, but the counterparties were identified gradually, and several of them sit on both sides of the relationship. In August 2021 the company said it had agreements with two customers covering more than half of the energy under its Energy Services Agreement, without naming them (claim 14506). The same filing separately credited GMR and Bitmain with hosting contracts meant to pre-fill the first 100MW site (claim 14507). By November 2021 the anonymous count had grown to four customers, now said to absorb all of the first facility's energy plus 85MW at a second site (claims 14412, 14374, 14363). Not until March 2022 were all four named: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14200, 14210). Those are signed agreements. The November 2021 Texas 200MW wind project is different: it is a letter of intent the company itself describes as subject to definitive agreements (claims 14204, 14074). SparkPool's role is services and advisory, not purchasing. Its Ether-pool agreement is also conditional on the company choosing to proceed (claims 14503, 14418). Coinmint is a supplier to Applied Digital, not a customer of it (claim 14419).

Concentration is high and entangled. GMR alone is repeatedly described as holding roughly 50% of the 100MW capacity (claims 14411, 14201, 14072). GMR, Bitmain and SparkPool are all strategic equity investors (claims 14505, 14378). Bitmain is at once an equipment supplier through a 200PH order (claims 14504, 14416), a hosting customer, a source of customer leads (claim 14417), and, through its affiliate Antpool, a joint-venture partner in 1.21 Gigawatts LLC (claims 14234, 14212). Diversification from two unnamed customers to four named ones is real. However, two of the four are affiliates of shareholders, so the independent customer base is narrower than the headline count suggests.

The revenue arithmetic cannot be reconciled from this record. None of the claims discloses contract value, pricing per megawatt, term length or recognized revenue. The only figures that tie out are capacity figures: the four customers filling 100MW plus 85MW, with GMR at about 50MW. Diligence should require the underlying hosting agreements before crediting any revenue run-rate to these relationships.

Applied Digital's early record reads as a stack of signed hosting agreements, though the parties behind them took months to name and most are tied to the company as shareholders. The first filing described hosting contracts with GMR and Bitmain that would use the capacity of the first planned 100MW site (claim 14507). The same filing cited agreements with '2 customers' covering more than 50% of available energy without naming them (claim 14506). By November 2021 the count had grown to four unnamed customers. Together they were said to absorb all energy under the Amended and Restated Energy Services Agreement plus 85MW at a second facility (claims 14412, 14374, 14363). Only in March 2022 were the four identified: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14210, 14200). That is a gap of roughly four months in which the whole book was anonymous.

Several headline relationships are not customer revenue at all. SparkPool is a service provider and adviser, not a customer (claims 14503, 14418). Coinmint hosts Applied's own rigs, so Applied is the customer there (claim 14419). The Bitmain 200PH order is a purchase (claims 14504, 14416). The Energy Services Agreement is supply-side (claim 14233). The January 2022 1.21 Gigawatts LLC with Bitmain affiliate Antpool is a joint-venture formation aimed at up to 1.5GW, not a capacity contract (claims 14234, 14203). The November 2021 Texas 200MW wind arrangement is explicitly a letter of intent 'subject to entry into definitive agreements' (claims 14204, 14074).

Concentration is high and has not diversified. GMR alone holds roughly half of the 100MW (claims 14411, 14231, 14201), and Bitmain appears to account for much of the rest. SparkPool, GMR and Bitmain are all strategic equity investors (claims 14505, 14377, 14378). Bitmain also sits on the supplier and JV sides of the ledger, which makes the core customer base substantially related-party. F2Pool and Hashing LLC are the only named customers without a disclosed equity tie. Their share of capacity is not stated.

The revenue arithmetic cannot be reconciled from this record. Every commitment is expressed in megawatts or capacity share, and no contract values, pricing, terms or recognized revenue appear in the claims. Some of the MW figures also shift between filings, and several quotes are truncated (e.g., claims 14358, 14059). Diligence should obtain the underlying contracts, their duration and termination rights, and the related-party pricing before treating the 100MW-plus-85MW figure as contracted revenue.

Applied Digital's early customer record is more concrete than most development-stage hosting stories, but it is also tightly interlocked with its own investors. In August 2021 the company described hosting contracts with GMR and Bitmain that would pre-fill its first 100MW site (claim 14507), and separately cited agreements with two unnamed customers covering more than 50% of available energy (claim 14506). By November the count had grown to four unidentified customers covering the full first-facility allocation plus 85MW at a second site (claims 14412, 14374, 14363). Only in the March 2022 filings were all four named: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14200, 14210, 14071, 14060). These read as signed agreements rather than MOU language, and the Amended and Restated Energy Services Agreement for 100MW gives them a supply-side anchor (claims 14233, 14202).

The softer language sits elsewhere. SparkPool, GMR and Bitmain are presented as 'strategic partners' and equity investors (claims 14505, 14376, 14377, 14378), and several of these relationships are services, advisory or supply arrangements rather than hosting revenue (claims 14503, 14502, 14418, 14417). The Coinmint co-hosting agreement places Applied Digital as the customer, not the vendor (claims 14419, 14371, 14357). The Texas 200MW wind facility is explicitly a letter of intent 'subject to entry into definitive agreements' (claims 14204, 14074), and should not be counted as contracted capacity.

Concentration is the central diligence issue. GMR, through JointHash, holds roughly 50% of the first 100MW (claims 14411, 14201, 14061), and Bitmain appears simultaneously as equity investor, equipment supplier for a 200PH order (claims 14504, 14416), hosting customer, lead source, and, via its affiliate Antpool, joint-venture partner in 1.21 Gigawatts LLC (claims 14234, 14203, 14073). Two related-party groups therefore plausibly account for the majority of contracted load, and the customer base broadened only from two to four names while capacity commitments expanded.

On revenue arithmetic, none of the claims on record disclose contract pricing, term, or expected revenue per megawatt; the only dollar figure is an $8,000 reference attached to the Antpool JV formation (claim 14234). The announced deals therefore cannot be reconciled to any revenue guidance from this record, and capacity allocation percentages should not be read as a proxy for contracted revenue.

Applied Digital's early customer record is thinner than its volume of partnership language suggests, and most of what looks like breadth traces back to two counterparties that are also shareholders. The firmest items are the hosting agreements. In August 2021 the company described agreements with two unnamed customers covering more than 50% of the energy under its Energy Services Agreement (claim 14506), and separately said GMR and Bitmain had hosting contracts to pre-fill the first planned 100MW site (claim 14507). By November 2021 the unnamed count had grown to four customers taking all first-facility energy plus 85MW at a second site (claims 14412, 14363, 14374). Only in the March 2022 filings were those four identified: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14200, 14210, 14071, 14060). For roughly seven months, then, investors were asked to credit anchor customers without names.

The remaining counterparties should be read as something other than customers. SparkPool supplied software, procurement and site-design services and later an advisory arrangement for a mining pool (claims 14503, 14418, 14358). Coinmint was a co-hosting host for Applied Digital's own mining, not a buyer of its capacity (claims 14419, 14371). The Bitmain 200PH order is a supply commitment (claims 14504, 14416). The Texas 200MW wind facility is a letter of intent expressly subject to definitive agreements (claims 14204, 14074). The 1.21 Gigawatts LLC with Antpool, a Bitmain affiliate, is a joint-venture framework for up to 1.5GW of datacenters rather than contracted demand (claims 14234, 14203, 14212, 14073, 14059).

Concentration is the central diligence issue. GMR alone was said to hold about 50% of the 100MW first facility (claims 14411, 14231, 14201, 14072, 14061). GMR and Bitmain are also strategic equity investors (claims 14505, 14377, 14378). Bitmain additionally acts as equipment supplier, a source of hosting leads (claim 14417) and, through Antpool, a JV partner. At least half the contracted capacity therefore sits with related parties whose roles overlap. F2Pool and Hashing LLC are the only named customers without a disclosed equity or supply tie.

The revenue arithmetic cannot be reconciled from this record. None of the claims disclose contract pricing, term, or dollar value; the only figure is the $8,000 attached to the Antpool LLC formation (claim 14234), which is not a revenue number. The capacity arithmetic is internally consistent, moving from more than 50% with two customers to 100% plus 85MW with four. It remains stated in megawatts rather than dollars.

Applied Digital's early counterparty record mixes signed commercial contracts with strategic relationships, a letter of intent, and a joint-venture framework, and the three categories need to be read separately. The strongest disclosures are the hosting agreements for the first 100MW facility. These began as agreements with two unnamed customers covering more than half of available energy (claim 14506), grew to four unnamed customers covering the full first site plus 85MW at a second (claims 14412, 14363), and were finally named in March 2022 as JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14200). GMR's contract for roughly 50% of the 100MW capacity is described consistently as signed (claims 14411, 14231, 14211). The Coinmint arrangement (claim 14371) is a co-hosting agreement under which Applied Digital was the hosted party, so it is a supplier relationship rather than customer revenue. The SparkPool agreements (claims 14503, 14418) are advisory and services contracts. The Bitmain 200PH rig order (claims 14504, 14416) is procurement.

The softer items are the Texas facility and the Antpool joint venture. The 200MW wind-powered Texas site rests on a November 2021 letter of intent that the company itself says is subject to definitive agreements (claims 14204, 14074). The 1.21 Gigawatts LLC formed with Antpool (claims 14234, 14203) is an entity-formation agreement tied to an aspirational datacenter scale, not a contracted capacity commitment.

Concentration is the central diligence issue. GMR and Bitmain appear on several sides of the business at once: both are strategic equity investors (claim 14505), and both are hosting customers (claim 14507). Bitmain is also the equipment supplier and a source of customer leads (claim 14417), and its affiliate Antpool is the JV partner. At first the company said GMR and Bitmain alone would fill the first site (claim 14507). Adding F2Pool and Hashing modestly broadens the base. Still, about half of the initial capacity sits with one investor-affiliated counterparty, and a second investor holds another share.

On revenue, the claims on record contain no contract values, pricing, term lengths or recognized hosting revenue. The announced deals therefore cannot be reconciled to any revenue arithmetic from this record. The megawatt allocations are internally consistent across filings, but converting them into dollars would require pricing and ramp disclosures that are absent here.

Applied Digital's disclosure record through April 2022 describes capacity allocation and strategic relationships rather than revenue. The hosting agreements appear to be signed contracts, but none of the filings attaches a price, term or dollar value to them, so the arithmetic that can be reconciled is megawatts, not dollars. The August 2021 language was the thinnest. It cited agreements with two unnamed customers covering "greater than 50%" of available energy (claim 14506), alongside assertions that GMR and Bitmain had "entered into hosting contracts" to pre-fill the first 100MW site (claim 14507). By November 2021 the count had become four unnamed customers absorbing the entire first facility plus 85MW at a second (claims 14412, 14363). The four were named only in the March 2022 filings: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (claims 14232, 14200, 14210). The one quantified allocation is GMR's contract for roughly 50% of the 100MW site (claims 14411, 14231, 14201). That implies the other three customers share about 50MW, but no per-customer split was ever given. The 85MW second-facility allocation was never tied to a named counterparty or to a site under a definitive energy agreement; only the first site's Amended and Restated Energy Services Agreement is disclosed (claims 14233, 14202). Non-binding language sits alongside these contracts. The Texas 200MW wind facility is a letter of intent explicitly "subject to entry into definitive agreements" (claim 14074), and it should not be counted as capacity or backlog. The 1.21 Gigawatts LLC with Antpool is framed around up to 1.5GW of datacenters (claim 14203), but the record shows only a formation agreement and a nominal $8,000 figure (claim 14234), not committed capital or offtake. Concentration is the central diligence issue. The customer base is small, and it overlaps heavily with the company's own shareholders and suppliers. GMR and Bitmain are each simultaneously equity investors, service or equipment providers, and hosting customers (claims 14505, 14502, 14504, 14417). Bitmain's affiliate Antpool is also the joint-venture partner (claim 14212). SparkPool held the same investor-advisor role (claims 14503, 14418), though the later filing's truncated text suggests its status was changing (claim 14358). Coinmint appears as Applied Digital's own co-hosting provider rather than as a customer (claim 14419). On this record, at least half of first-site capacity, and plausibly more, rests with related strategic investors whose terms are undisclosed.

Applied Digital's customer record through April 2022 is thinner than its volume of repetition suggests, because a small set of statements is restated across successive filings rather than extended by new deals. The earliest disclosure described agreements with two unnamed customers covering more than half of available energy (claim 14506), while a parallel statement said GMR and Bitmain hosting contracts would fill the first planned 100MW site (claim 14507). By November 2021 the count rose to four unnamed customers covering the full first facility plus 85MW at a second (claims 14412, 14374, 14363), and only in March 2022 were they identified as JointHash (a GMR subsidiary), Bitmain Technologies, F2Pool and Hashing LLC (claims 14232, 14210, 14200). GMR's contract for roughly 50% of the 100MW capacity is the one specifically sized signed agreement, repeated at least seven times (claims 14411, 14231, 14201, 14072, 14061), and it covers an operation still described as under development.

Much of the rest sits outside signed hosting revenue. SparkPool, GMR and Bitmain are presented as strategic equity investors and service providers (claims 14505, 14503, 14502), meaning the two largest named customers also hold equity, which blurs arm's-length customer demand with investor support. Bitmain's role is framed as preferred equipment access and customer leads (claim 14417), and its 200PH order is an equipment purchase, not revenue (claims 14504, 14416). The SparkPool mining-pool arrangement is advisory and conditional (claim 14418). The Texas 200MW wind facility is an explicit letter of intent subject to definitive agreements (claims 14204, 14074), and the 1.21 Gigawatts LLC with Antpool, a Bitmain affiliate, is a joint-venture formation rather than a customer contract (claims 14234, 14203).

Concentration therefore went from two anonymous customers to four named ones, but with GMR near half the first site and Bitmain-linked entities on both the customer and JV sides, effective exposure runs to two related groups. On revenue arithmetic, the record contains no contract values, pricing, term lengths or recognized revenue, so the announced megawatt allocations cannot be reconciled to any disclosed figures; the only dollar amount cited, $8,000 tied to the Antpool entity formation (claim 14234), is immaterial to hosting economics. The capacity is contracted on paper, but its revenue remains unquantified and contingent on facilities still being built.

Applied Digital's disclosed customer base through early 2022 rests on hosting contracts, most of which were signed before the counterparties were identified. The August 2021 filings describe agreements with two unnamed customers covering more than 50% of available energy (claim 14506). By November 2021 that became four customers taking the full first-facility allocation plus 85MW at a second site (claims 14412, 14374, 14363). The March 2022 filings finally name them: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14200, 14210). For roughly seven months, then, the signed-customer count grew without a named counterparty, and readers had to infer identities from the 'strategic partner' language.

That language matters because most of the relationships on record are not customer revenue at all. SparkPool's role is advisory and technical (claims 14503, 14418, 14358). The Bitmain 200PH rig order is a purchase commitment running the other way (claims 14504, 14416). Coinmint is a co-hosting provider to the company, not a buyer (claims 14419, 14357). The 1.21 Gigawatts LLC with Antpool, a Bitmain affiliate, is a joint-venture framework for up to 1.5GW, with only a nominal $8,000 figure attached (claims 14234, 14203). The Texas 200MW wind project is explicitly a letter of intent 'subject to entry into definitive agreements' (claims 14204, 14074) and should carry no weight as contracted demand.

Concentration is high and largely related-party. GMR alone holds roughly 50% of the 100MW first facility (claims 14411, 14201). GMR and Bitmain are also strategic equity investors (claims 14505, 14377, 14378), so two of the four named customers are shareholders or their subsidiaries. Bitmain additionally supplies equipment and customer leads (claim 14417) and co-owns the JV through Antpool. F2Pool and Hashing LLC are the only arm's-length names, and their individual allocations are not disclosed.

The revenue arithmetic cannot be reconciled from this record. None of the claims states contract values, per-MW hosting rates, tenor, or recognized revenue. The capacity figures are also only partly consistent. GMR's ~50% share fits within the four-customer total for the 100MW facility, but the 85MW second-facility allocation has no disclosed split, site-level energy agreement, or energization date beyond the first site's Amended and Restated Energy Services Agreement (claims 14233, 14202). Contracted megawatts are therefore established; contracted dollars are not.

Applied Digital's early record is a small set of signed hosting contracts wrapped in a larger layer of partnership language, and the two should be read separately. The binding items are the hosting agreements. The August 2021 disclosure cites agreements with two unnamed customers covering more than half of available energy (14506), and says GMR and Bitmain signed hosting contracts to pre-fill the first 100MW site (14507). By November 2021 this had become four unnamed customers covering all of the first facility's energy plus 85MW at a second site (14412, 14374, 14363). Only in March 2022 were the four identified: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (14232, 14200, 14210, 14071, 14060). The Amended and Restated Energy Services Agreement for 100MW is the supply-side contract underneath them (14233, 14202). The Coinmint co-hosting agreement (14419, 14371, 14357) runs the other way, with Applied Digital as the hosted party, so it is not customer revenue. Everything else is softer. The SparkPool, GMR and Bitmain relationships are described as services, advisory, equipment access and customer leads (14503, 14417, 14418, 14358), not as purchase commitments. The Texas 200MW wind facility is a letter of intent explicitly 'subject to entry into definitive agreements' (14204, 14074). The 1.21 Gigawatts LLC with Antpool is an LLC agreement for up to 1.5GW of datacenters. The record attaches only an $8,000 figure to it (14234, 14203), which reads as formation capital rather than contracted demand. Concentration is the central diligence issue. GMR alone holds roughly half of the 100MW (14411, 14201). GMR and Bitmain are both hosting customers and strategic equity investors (14505, 14377, 14378), and Antpool is a Bitmain affiliate. That leaves two of four customers, and the only large joint venture, tied to shareholders. That is related-party revenue and deserves discount. On arithmetic, the record discloses no contract values, pricing per MW, term lengths or recognized revenue, only capacity shares. The claims are internally consistent: 'greater than 50%' from two customers grows to 100% from four as F2Pool and Hashing are added. But they cannot be reconciled to any dollar revenue figure from the claims on file. The input also truncates after claim 14060, so later developments are not assessed here.

Applied Digital's customer record divides cleanly along a naming axis rather than a signing axis. From the outset the company used the language of executed agreements — "entered into agreements with 2 customers which will account for greater than 50% of the available energy" (claim 14506, 2021-08-12), expanded to four customers covering the total available energy at the first facility plus 85MW at the second (claims 14412, 14374, 14363) — but withheld identities for roughly seven months. Identification arrives only in the March 2022 filings, where the four are named as JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14210, 14200, 14071, 14060). The earlier "strategic partner" cohort of SparkPool, GMR and Bitmain (claim 14505) was always named, but as equity investors and service providers rather than as revenue counterparties, and the two roles were presented in adjacent sentences in a way that blurs them.

Concentration evolved from acute to merely severe, and less than the customer count suggests. GMR is repeatedly described as having signed for approximately 50% of the 100MW first site (claims 14411, 14231, 14201, 14211, 14061, 14072), so one relationship carries half the initial book. Bitmain occupies four roles simultaneously: equity investor (14378), equipment supplier for the 200PH order (14504, 14416), lead-generator for hosting customers (14417), joint-venture partner through its Antpool affiliate (14234, 14203, 14212), and named hosting customer (14232). The contracted base is therefore substantially related-party rather than arm's-length. Coinmint (14419, 14371, 14357) runs the other direction entirely — Applied Digital is the customer there, not the host.

The softer items are labeled honestly. The Texas 200MW wind facility is a letter of intent "subject to entry into definitive agreements" (claims 14204, 14074), and the 1.21 Gigawatts LLC vehicle framed around up to 1.5GW of datacenters (14203) was capitalized at $8,000 (14234) — a structure, not a commitment.

The revenue arithmetic does not reconcile because it is never stated. Every commitment is denominated in megawatts, not dollars; no $/MW-month rate, contract term, or prepayment is disclosed in these claims, and the capacity itself was unbuilt when pre-filled (14507). Contracted MW is not revenue until price and tenor are shown.

Applied Digital's counterparty record separates cleanly into three tiers, and the distinctions matter more than the volume of announcements suggests. The genuinely signed tier is the hosting book: agreements with two customers covering more than half the available energy under the Energy Services Agreement as of August 2021 (claim 14506), expanded to four customers covering the total available energy at the first facility plus 85MW at a second by November (claims 14412, 14374, 14363), and a GMR contract for roughly 50% of the 100MW first site (claims 14411, 14231, 14201). The second tier is supplier and advisory relationships repeatedly described in customer-adjacent language: SparkPool provides software, hardware selection and site engineering (claims 14503, 14358), Bitmain provides preferred equipment access and hosting leads (claims 14417, 14378), and the 200PH Bitmain order (claims 14504, 14416) is a purchase obligation by Applied Digital, not revenue. The Coinmint co-hosting agreement (claims 14419, 14371, 14357) runs the same direction — Applied Digital is the customer there. The third tier is explicitly pre-contractual: the Texas 200MW wind facility is a letter of intent 'subject to entry into definitive agreements' (claims 14204, 14074), and the Antpool joint venture 1.21 Gigawatts LLC, framed around up to 1.5GW of datacenters (claim 14203), was capitalized at $8,000 (claim 14234).

On identification, the hosting customers went unnamed for roughly seven months. The August and November 2021 disclosures describe two, then four, counterparties purely by count (claims 14506, 14412, 14363), while naming GMR, SparkPool and Bitmain separately as strategic equity investors (claims 14505, 14376). Names arrived only in March 2022: JointHash Holding Limited, Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claim 14232).

That reveal is where concentration becomes legible rather than improving. JointHash is disclosed as a GMR subsidiary and Antpool as a Bitmain affiliate (claims 14232, 14234), so three of the four named customers, the joint-venture partner, the equipment supplier and two of three equity investors resolve to the same two relationships. Growth from two to four customers is nominal diversification against a fully subscribed 100MW.

The arithmetic does not reconcile to revenue because no revenue is attached. GMR is described as contracting for about 50MW (claim 14211) while JointHash is one of four customers absorbing the same 100MW (claim 14210), leaving the overlap unquantified, and no disclosed price per megawatt permits translating 185MW of contracted capacity into dollars.

Applied Digital's counterparty disclosure moves in a clear direction: from unnamed counts toward actual legal names, but only after roughly seven months. The August 2021 record refers to "agreements with 2 customers" covering more than 50% of available energy under the Energy Services Agreement without identifying either (claim 14506), while separately stating that GMR and Bitmain had "entered into hosting contracts" (claim 14507). By November the count had risen to four customers absorbing the entire first facility plus 85MW at a second site, still unidentified (claims 14412, 14374, 14363). Identification arrives only in the March 2022 filings, which name JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14200, 14210, 14071, 14060). The language throughout is signed-contract language rather than MOU or discussion language, with one clean exception: the Texas 200MW wind facility is explicitly a letter of intent "subject to entry into definitive agreements" (claims 14204, 14074), and should not be counted as contracted.

Concentration is the more important story, and it tightens rather than diversifies. GMR alone is repeatedly described as having signed for approximately 50% of the 100MW first facility (claims 14411, 14377, 14231, 14201, 14211, 14072, 14061), and the named customer list resolves GMR into JointHash, so one economic group holds roughly half the initial book. The remaining names are not arm's-length either: SparkPool, GMR and Bitmain are disclosed as strategic equity investors (claims 14505, 14376, 14377, 14378), and the Antpool joint venture partner in 1.21 Gigawatts LLC is an affiliate of Bitmain (claims 14234, 14212, 14203). Four customers therefore reduce to approximately two related-party clusters. Note also that SparkPool and Coinmint are not customers at all despite appearing in partnership framing: SparkPool is a software and advisory vendor (claims 14503, 14418, 14358) and Coinmint is Applied Digital's own co-hosting host (claims 14419, 14371, 14357), an inbound supplier relationship.

On revenue arithmetic, the record as presented supports no reconciliation, because it contains capacity commitments and no contracted dollar values. What can be tested is the capacity denominator, and it shifts: two customers covering "greater than 50%" (claim 14506) becomes four covering 100% plus 85MW (claim 14363) with only two incremental names ever disclosed. The 1.21 Gigawatts vehicle, capitalized at $8,000 (claim 14234), carries a "up to 1.5GW" ambition that is aspiration, not backlog.

The naming discipline in this record improves sharply at a single point, and the date matters. Through the second half of 2021 Applied Digital described its demand book by count rather than identity: two customers accounting for more than half the available energy under the Energy Services Agreement (claim 14506), then four customers covering the whole of the first facility plus 85MW at a second (claims 14412, 14374, 14363). The counterparties behind those counts were not disclosed until 2022-03-20, when the identical sentence finally names JointHash Holding Limited, Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claim 14232, repeated verbatim at 14200, 14210, 14071, 14060). The contractual language never changed across those eight months; only the identification did, which is the cleanest available evidence that the earlier unnamed counts were real signings rather than placeholders.

The concentration story is less reassuring than the headline progression from two customers to four suggests. JointHash is disclosed as a GMR subsidiary, and GMR is separately described as having signed for roughly 50% of the 100MW first site (claims 14411, 14231, 14201, 14072, 14061). Bitmain is the second named customer. Both GMR and Bitmain, along with SparkPool, are simultaneously disclosed as strategic equity investors (claims 14505, 14377, 14378, 14376) and as suppliers or advisors (claims 14504, 14503, 14418, 14417). So the customer count doubled while roughly half of first-facility capacity stayed with one related party, and at least two of four named customers sit on the cap table. The Coinmint arrangement runs the other direction entirely: there Applied Digital is the co-hosting customer, not the vendor (claims 14419, 14371, 14357).

Signature quality separates clearly at two items. The Texas 200MW wind facility is an explicit letter of intent, and the filing says so, adding that it is subject to entry into definitive agreements (claims 14204, 14074); no definitive agreement appears later in this record. The Antpool joint venture is a signed LLC agreement but not a revenue contract, and the arithmetic is the tell — $8,000 of capitalization against an up-to-1.5GW ambition (claims 14234, 14203, 14212, 14073, 14059).

Revenue reconciliation is not possible here, and that is itself the finding. Every disclosure is megawatt-denominated; no contract value, term, or price per kilowatt appears in any claim. Worth noting also that contracted capacity totals 185MW while the Amended and Restated Energy Services Agreement is described as covering 100MW (claims 14233, 14202), leaving the second site's 85MW outside the quoted energy arrangement. The Bitmain 200PH equipment order, due late in calendar 2021, is restated unchanged on 2021-11-01 and 2021-11-30 (claims 14416, 14360) with no delivery confirmation in the record.

Applied Digital's counterparty record separates cleanly into three tiers, and only one of them is a signed customer agreement. The genuinely contractual tier is the hosting book: two customers accounting for more than 50% of available energy as of August 2021 (14506), expanding to four customers covering the full first facility plus 85MW at a second site by November (14412, 14374, 14363). The aspirational tier is thinner than the announcement cadence suggests. The Texas facility is a letter of intent dated November 24, 2021, and the company says so plainly — the arrangement is 'subject to entry into definitive agreements by the parties' (14074, 14204). The 1.21 Gigawatts LLC joint venture with Antpool is framed around 'up to 1.5GW' of datacenters (14203, 14212, 14059, 14073) but is capitalized at $8,000 (14234); that is a formation document, not demand.

On identification, the gap is roughly seven months. From August through November 2021 the hosting customers are counted but not named (14506, 14412, 14374, 14363), while the counterparties that are named in the same breath — SparkPool, GMR, Bitmain — are described as 'strategic equity investors' (14505, 14378, 14376, 14377) and service providers (14503, 14502, 14418, 14358, 14359), which is a different relationship than paying for capacity. Names arrive only on March 20, 2022: JointHash Holding Limited, Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (14232, 14200, 14210, 14071, 14060).

That disclosure is where concentration becomes legible rather than improving. JointHash is identified as a GMR subsidiary, GMR holds roughly 50% of the 100MW first facility (14411, 14231, 14201, 14211, 14061, 14072), and Bitmain is simultaneously equity holder (14505), equipment supplier of the 200PH order (14504, 14416, 14360), lead-generator (14417), JV partner through Antpool (14234) and named customer. Two of four names are related parties; the Coinmint arrangement (14419, 14371, 14357) runs the other direction, with Applied Digital as the customer.

The revenue arithmetic cannot be reconciled because it is never stated. No claim discloses contract value, term, price per MW or energization date — only megawatt allocations. Note also that the November 21, November 30, March 27, April 7 and April 11 versions restate identical language, so the apparent volume of announcements reflects successive filings rather than distinct commercial events.

Applied Digital's customer disclosures move from anonymity to identification over roughly seven months, and the shift is the most informative thing in the record. In August 2021 the company reported agreements with two customers covering more than 50% of available energy under the Energy Services Agreement without naming either (claim 14506), even while naming GMR and Bitmain as holders of hosting contracts in the same filing (14507). By November the count was four customers, still unnamed, said to absorb the total available energy at the first facility plus 85MW at a second (14412, 14374, 14363). Only on 2022-03-20 do the counterparties appear: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (14232, repeated at 14200, 14210, 14071, 14060). That identification matters because it collapses the apparent diversification. GMR is separately described as having signed for approximately 50% of the 100MW capacity (14411, 14231, 14201, 14061), and JointHash is its subsidiary, so one relationship spans roughly half the book. Bitmain is simultaneously customer, equipment supplier under the 200PH order (14504, 14416), strategic equity investor (14505, 14378), lead-generator for other hosting customers (14417), and joint-venture partner through Antpool in 1.21 Gigawatts LLC (14234, 14212, 14203). Concentration did not fall from two names to four; it re-expressed itself across affiliated entities.

The softer items are explicitly flagged as such and should not be scored as contracts. The Texas 200MW wind facility is a letter of intent, stated as subject to entry into definitive agreements (14204, 14074). The SparkPool Ether pool arrangement is conditional on the company deciding to proceed (14418). The 1.21 Gigawatts vehicle is framed around up to 1.5GW of datacenters (14203) while the LLC itself is capitalized at $8,000 (14234) — an ambition figure, not a commitment. Signed and specific are the Coinmint co-hosting agreement (14419, 14371, 14357) and the four named hosting agreements.

Revenue arithmetic cannot be reconciled from this record, because none is disclosed. Every claim quantifies megawatts or petahashes; no contract value, hosting rate, or term appears. Note also that the Amended and Restated Energy Services Agreement covers 100MW (14233, 14202), while customer commitments total roughly 185MW across two sites — the second facility's energy supply is asserted but not evidenced here. The record ends 2022-04-11.

Applied Digital's customer record divides cleanly into two periods, and the dividing line is disclosure of names rather than any change in the underlying contracts. From August through November 2021 the company repeatedly asserted signed hosting agreements without identifying the counterparties: first "agreements with 2 customers" covering more than half the energy under the Energy Services Agreement (claim 14506), then "agreements with 4 customers" covering the total available energy at the first facility and 85MW at the second (claims 14412, 14374, 14363). The four were not named until March 2022, when they emerged as JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14210, 14200, 14071, 14060). GMR itself was named earlier and consistently as holding a contract for roughly 50% of the first 100MW (claims 14411, 14377, 14231, 14201, 14072, 14061), as was Coinmint LLC on the co-hosting side (claims 14419, 14371, 14357).

The concentration story is that it never loosened, it was only re-labelled. Two of the four eventual named customers are the same entities described as "strategic equity investors" and partners — GMR and Bitmain (claims 14505, 14377, 14378, 14376) — with Bitmain simultaneously a supplier under the 200PH equipment order (claims 14504, 14416, 14360), a lead-generator for hosting customers (claim 14417), and, through Antpool, a joint-venture counterparty (claims 14234, 14203). SparkPool, present throughout 2021 as a services and advisory partner (claims 14503, 14418, 14358), never appears among the named four; F2Pool occupies that slot instead.

On softer language, the Texas 200MW wind facility is explicitly a letter of intent "subject to entry into definitive agreements" (claims 14204, 14074) and should not be counted as contracted capacity. The 1.21 Gigawatts LLC vehicle with Antpool is a signed LLC agreement, but the only dollar figure anywhere in the record is its $8,000 capitalization against a stated ambition of up to 1.5GW (claim 14234).

Revenue arithmetic cannot be reconciled from this record at all. No claim discloses contract value, hosting rate, term, or per-customer megawatt allocation; the disclosures are denominated in percentages of capacity, so "total available energy" (claim 14200) is unverifiable against any revenue line.

The record contains far fewer distinct transactions than it contains announcements. The same four-customer sentence recurs essentially verbatim across claims 14232, 14210, 14200, 14071 and 14060 between March 20 and April 11, 2022, and the GMR hosting line recurs across 14411, 14377, 14231, 14211, 14201, 14072 and 14061. Counting filings rather than counterparties would roughly triple the apparent commercial base.

On naming, the August 2021 disclosure of agreements with two customers accounting for "greater than 50% of the available energy under the Energy Services Agreement" (14506) identifies no one, though the adjacent 14507 attributes hosting contracts to GMR and Bitmain, making the unnamed pair inferable rather than disclosed. Identification arrives only in March 2022, when the four are given as JointHash Holding Limited (a subsidiary of GMR), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (14232). Note the substitution: the party repeatedly credited with signing for "approximately 50% of our 100MW capacity" is GMR (14411), but the named signatory is its subsidiary. SparkPool is never a customer at all — 14503, 14418, 14376 and 14358 describe software, advisory and site-engineering services flowing toward Applied Digital, and 14358 breaks off mid-clause at "is currently in the process of s", which the record leaves unresolved.

Concentration nominally improves, from two customers above half the energy (14506) to four covering the entirety of first-facility energy plus 85MW at the second (14412, 14374, 14363). Substantively it does not. Two of the four named are Bitmain and a GMR affiliate, and 14505 discloses both, with SparkPool, as strategic equity investors — capacity contracted to shareholders. Bitmain simultaneously supplies the 200PH equipment order (14504, 14416), supplies hosting leads (14417), and through its Antpool affiliate co-owns the 1.21 Gigawatts LLC joint venture (14234).

On arithmetic, there is little to reconcile. Every claim expresses commitment in megawatts or percentages; none discloses contract value, term or hosting rate, so announced capacity cannot be tied to revenue. The sharpest disproportion is 14234 against 14203: a joint venture capitalized at $8,000 presented as the vehicle for up to 1.5GW of datacenters. Coinmint (14419, 14371, 14357) is a cost, not a customer — Applied Digital is the hosted party. The Texas 200MW arrangement remains a letter of intent, "subject to entry into definitive agreements by the parties" (14074).

The record divides cleanly along a naming line that Applied Digital did not cross until March 2022. From the August and November 2021 disclosures, the company consistently identified its equity-holding partners — SparkPool, GMR and Bitmain (claims 14505, 14376, 14378) — while leaving its actual revenue counterparties anonymous: first "2 customers" accounting for more than half of available energy (claim 14506), then "4 customers" covering the entire first facility plus 85MW at a second (claims 14412, 14374, 14363). Only in the March 2022 filings are those four identified as JointHash Holding Limited, Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14210, 14200, 14071, 14060). The seven-month gap between contracting and identification is the single most important feature of this history, and it resolves in an uncomfortable direction: JointHash is disclosed as a GMR subsidiary and Bitmain is the same entity described elsewhere as an equity investor (claim 14378), an equipment supplier of 200PH (claims 14504, 14416, 14360) and, through its Antpool affiliate, a joint-venture partner (claims 14234, 14212, 14203). Two of four named customers are therefore related parties, and GMR alone holds roughly 50% of the first facility (claims 14411, 14377, 14231, 14211, 14201, 14072, 14061). Coinmint should not be read as a customer at all; claims 14419, 14371 and 14357 describe APLD as the tenant in a co-hosting arrangement.

Separating executed from aspirational, the hosting and energy agreements are described as entered into, and the Antpool LLC agreement is dated and signed (claim 14234). The Texas 200MW wind facility is explicitly a letter of intent "subject to entry into definitive agreements" (claims 14204, 14074), and should carry no weight. The Antpool vehicle pairs a "up to 1.5GW" ambition (claim 14203) with $8,000 of disclosed capitalization (claim 14234) — a framework, not a build.

The revenue arithmetic does not reconcile because it is never presented. The record quantifies megawatts and percentages, never contracted dollar values or terms, so 185MW of claimed coverage cannot be tied to any revenue figure. Only GMR's share is sized; F2Pool and Hashing LLC remain unquantified residuals within a fully subscribed capacity claim.

Applied Digital's customer record through April 2022 is narrower than the announcement count suggests. Substantially the same disclosure block is restated across eight filing dates between 2021-08-12 and 2022-04-11, so claims 14506, 14412, 14374, 14363, 14232, 14210, 14200, 14071 and 14060 describe one evolving customer set rather than nine discrete events.

On naming, the pattern inverts the usual one: the strategic partners were identified early while the revenue counterparties were not. SparkPool, GMR and Bitmain appear by name as equity investors and service providers from the first disclosure (14505, 14376, 14377, 14378). The hosting customers were anonymous for seven months. August 2021 discloses agreements with two customers covering more than 50% of available energy without identifying either (14506). November 2021 raises that to four customers covering the first facility's full allocation plus 85MW at a second site, still unnamed (14412, 14363). Identification arrives only on 2022-03-20: JointHash Holding Limited, Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (14232). The operative language throughout is "entered into agreements," which is signed-contract phrasing rather than MOU or discussion language. The one explicitly non-binding item is the Texas 200MW wind facility, a letter of intent "subject to entry into definitive agreements by the parties" (14204, 14074).

Concentration did not improve as the count went from two to four. JointHash is disclosed as a GMR subsidiary (14232), and GMR is separately credited with a contract for roughly 50% of the 100MW (14411, 14231, 14061, 14211) — most plausibly the same commitment described twice under two names. Bitmain is simultaneously equity investor (14378), equipment supplier on the 200PH order (14416, 14360), lead generator for hosting prospects (14417) and named anchor customer (14232), while its affiliate Antpool is the joint venture counterparty in 1.21 Gigawatts LLC (14234, 14203). Coinmint runs the other direction entirely: there Applied Digital is the customer, not the host (14419, 14371).

The revenue arithmetic cannot be reconciled because no revenue is stated. The disclosures quantify megawatt allocations only — no contract values, rates or terms — so "total available energy" plus 85MW does not convert into dollars. The single hard figure on record is the $8,000 capitalization of the Antpool joint venture (14234), against a stated ambition of up to 1.5GW (14203).

The distinction that matters here is between agreements Applied Digital signed as a seller and everything else. The genuinely revenue-bearing signed contracts are the four customer agreements covering the first facility's 100MW and 85MW at the second (14412, 14374, 14363, 14232, 14210, 14200, 14071, 14060), plus the GMR contract for roughly half the first site's capacity (14411, 14231, 14201, 14211, 14061, 14072). Several prominent-sounding items run the other direction: the Bitmain 200PH order is Applied Digital buying equipment (14504, 14416, 14360), the Coinmint co-hosting arrangement is Applied Digital renting space (14419, 14371, 14357), and SparkPool's role is advisory software and design work rather than a revenue relationship (14503, 14418, 14358). The Texas 200MW wind facility is explicitly a letter of intent "subject to entry into definitive agreements" (14204, 14074) and should not be counted at all.

On naming, the counterparties went unidentified for roughly seven months. The August 2021 disclosure cited "agreements with 2 customers" representing over half the available energy without naming them (14506), though adjacent language implies GMR and Bitmain (14507); November's expansion to four customers stayed anonymous (14412, 14374, 14363). Names surface only in March 2022: JointHash Holding Limited, Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (14232, 14210).

That naming is what exposes the concentration. JointHash is disclosed as a GMR subsidiary, so the separately headlined GMR contract for approximately 50% of 100MW and one of the four named customers are the same exposure presented twice. Bitmain appears at once as hosting customer, equipment supplier, lead source, strategic equity investor (14505, 14378) and, through its Antpool affiliate, joint-venture partner (14234, 14212, 14203). Two related-party blocs therefore sit behind most of the contracted capacity, and four customers is a thinner diversification claim than it reads.

On arithmetic, there is nothing to reconcile, because no contract values, rates or terms appear anywhere in this set; capacity is the only unit given and revenue must be inferred. The single dollar figure is the $8,000 capitalization of 1.21 Gigawatts LLC against an announced ambition of up to 1.5GW (14234). The record supplied ends in April 2022.

The signed column is narrower than the announcement count suggests. Four customer contracts are the only agreements described in binding terms, and they were not named until the March 2022 filings, where they appear as JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (14232, 14210, 14200). Before that the same book of business was disclosed anonymously: two customers covering "greater than 50% of the available energy" in August 2021 (14506), then four customers covering the full first facility plus 85MW at a second by November (14412, 14374, 14363). The set did not expand between those disclosures so much as acquire names, and the Nov 1, Nov 21 and Nov 30 versions, like the March 20/27 and April 7/11 versions, are the same paragraph carried forward verbatim. The volume of claims overstates the amount of contracting.

The partners should be kept separate from the customers. SparkPool and Bitmain are described as strategic equity investors and service providers (14505, 14503, 14376, 14378, 14358), and SparkPool never appears in the named customer list at all; its role is software, hardware selection and site engineering (14418), not offtake. GMR is credited throughout with "a contract for approximately 50% of our 100MW capacity" (14411, 14377, 14231, 14201, 14061), but the named list places that contract at JointHash. These are most plausibly one contract described two ways rather than two. Coinmint runs the other direction entirely (14419, 14371, 14357), with Applied as the co-hosting customer. The Texas 200MW wind facility is a letter of intent and is labelled as such, "subject to entry into definitive agreements" (14204, 14074). The Antpool joint venture, 1.21 Gigawatts LLC, is an executed LLC agreement but capitalized at $8,000 against an ambition of up to 1.5GW (14234, 14203).

Concentration moved from two unnamed counterparties to four named ones, but two of the four are also equity investors, and Bitmain is simultaneously supplier (14504, 14416), lead source (14417) and joint-venture affiliate. The revenue arithmetic cannot be reconciled from this record. Every agreement is denominated in megawatts or percentages of capacity rather than contract value or price per MW, and the Energy Services Agreement cited covers only 100MW at the first facility (14233, 14202) while the customer contracts claim a further 85MW at a second.

The record divides cleanly along a binding/non-binding line, and the non-binding portion is small. Executed instruments include the GMR services and hosting contracts (14502, 14411, 14231, 14201), the Coinmint co-hosting agreement (14419, 14371, 14357), the SparkPool advisory agreements (14503, 14418), the Amended and Restated Energy Services Agreement for 100MW (14233, 14202), and the Antpool joint venture LLC agreement forming 1.21 Gigawatts LLC (14234, 14212, 14203). The only explicitly conditional item is the November 24, 2021 Texas letter of intent for a 200MW wind-powered facility, which the company itself qualifies as "subject to entry into definitive agreements" (14074, 14204); that 200MW should not be carried into any capacity or revenue build. Note also that the Bitmain 200PH order (14504, 14416, 14360) runs the other direction — it is procurement spend, not contracted revenue — as does the Coinmint arrangement, where Applied Digital is the hosted party rather than the host.

On identification, there is a clear before-and-after. From August through November 2021 the hosting counterparties are unnamed aggregates: "2 customers" representing more than 50% of available energy (14506), then "4 customers" covering the full first facility plus 85MW at the second (14412, 14374, 14363). Names appear only on March 20, 2022: JointHash Holding Limited, Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (14232, 14210, 14200, 14071, 14060). The disclosure is candid that JointHash is a GMR subsidiary, and GMR and Bitmain are both described as strategic equity investors (14505, 14378, 14377). So at least half the named book is related-party, and the "strategic partner" framing applied to SparkPool, GMR and Bitmain (14358, 14360, 14376) conflates equity holders, service providers and revenue counterparties in one sentence.

Concentration has not diluted; it has been re-labeled. Two counterparties above 50% became four counterparties at 100% of the first facility, with GMR alone holding roughly 50MW of 100MW (14411, 14231, 14072, 14061). Revenue arithmetic cannot be reconciled at all, because no contract value, rate per megawatt or term is disclosed anywhere in this set — only capacity. The capacity math is internally consistent for the first facility but the 85MW at the second has no corresponding energy agreement on record, and the Antpool vehicle carries a $8,000 capitalization (14234) against a stated ambition of up to 1.5GW (14203).

Applied Digital's counterparty disclosure divides cleanly into three tiers, and the distinction matters more than the company's uniform "agreement" language suggests. Genuinely signed and named is the GMR hosting contract for roughly 50% of the first site's 100MW, which recurs verbatim from claim 14411 through claims 14377, 14231, 14211, 14201, 14072 and 14061 — the same sentence redeployed across at least seven filings over eight months, which is itself evidence that no new anchor signing occurred in that window. The Coinmint co-hosting arrangement (claims 14419, 14371, 14357) is real but inverted: there Applied Digital is the customer, not the vendor, and it should not be netted into demand-side coverage. At the other end, the Texas 200MW wind facility is explicitly an LOI "subject to entry into definitive agreements by the parties" (claims 14204, 14074) and should carry zero weight.

On identification, the company operated for roughly seven months on counted rather than named customers. August 2021 discloses "2 customers" representing greater than 50% of available energy without naming them (claim 14506); by November this becomes "4 customers" covering the full 100MW plus 85MW at a second facility (claims 14412, 14374, 14363), still unnamed. Names arrive only on 2022-03-20 (claim 14232): JointHash Holding Limited, Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC. The reveal undercuts the diversification the count implied. JointHash is disclosed in the same sentence as a GMR subsidiary, and GMR, Bitmain and SparkPool are all named equity investors in the issuer (claims 14505, 14378, 14376, 14377). So at least half the contracted book sits with holders of the company's own stock, and Bitmain simultaneously supplies the 200PH equipment order (claims 14504, 14416), supplies hosting leads (claim 14417), and — through its Antpool affiliate — co-owns the 1.21 Gigawatts LLC joint venture (claims 14234, 14212, 14203). Concentration did not fall as the customer count doubled; it was re-labelled.

The revenue arithmetic does not reconcile because it is never presented. Every claim quantifies megawatts or petahash, none discloses contract value, rate per megawatt, or term, so booked capacity cannot be bridged to dollars from this record. The one hard number available is the $8,000 capitalization of the Antpool JV (claim 14234) against an ambition of up to 1.5GW (claim 14203) — a gap that should be read as option value, not backlog.

Applied Digital's customer record is more concrete than its early language suggested, but it is concentrated in counterparties that are also its investors and suppliers. The first disclosures in August 2021 referred to agreements with two customers that would take more than half of the available energy, without naming them (claim 14506). A parallel passage said GMR and Bitmain had entered into hosting contracts (claim 14507). By November 2021 the count had become four customers, covering all of the first facility's energy plus 85MW at a second site, still unnamed (claims 14412, 14374, 14363). Only in the March 2022 filings were the four identified: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14210, 14071, 14060). The progression from unnamed to named is a genuine improvement in disclosure quality.

Concentration is the more important issue. GMR alone was described as having signed for roughly 50% of the 100MW capacity (claims 14411, 14231, 14201). GMR and Bitmain were also described as strategic equity investors (claims 14505, 14377, 14378). Bitmain further supplied 200PH of mining equipment (claims 14504, 14416) and customer leads (claim 14417). Its affiliate Antpool co-formed the 1.21 Gigawatts LLC for up to 1.5GW of datacenters (claims 14234, 14212, 14203). At least half of contracted capacity therefore sits with related-party-adjacent counterparties whose economics run in several directions at once. The SparkPool and GMR advisory arrangements are vendor relationships, not revenue (claims 14503, 14418, 14359). The Coinmint agreement runs the other way, with Applied Digital as the customer (claim 14419).

The softer language should be separated from the signed agreements. The Texas 200MW wind facility is explicitly a letter of intent, subject to definitive agreements (claims 14204, 14074). The Antpool vehicle is a joint-venture formation, not a customer contract, and its only disclosed figure is $8,000 (claim 14234).

The revenue arithmetic cannot be reconciled from this record. No contract values, tenors, per-megawatt pricing or take-or-pay terms appear. The Energy Services Agreement covers 100MW (claim 14233), while customers are said to absorb 100MW plus 85MW at a second facility whose power arrangements are not documented here (claim 14200). Readers should treat capacity percentages as allocation statements, not evidence of revenue.

Applied Digital's record separates into three tiers. The first holds signed commercial contracts with named counterparties. The Coinmint LLC co-hosting agreement (14419, 14371) is one, though there Applied was the customer, not the vendor. The hosting agreements are the others. They were first described as contracts with 'GMR and Bitmain' (14507) and '2 customers' taking more than 50% of energy (14506). By March 2022 they were identified as JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (14232, 14210). The second tier holds services and equity relationships rather than revenue contracts. SparkPool, GMR and Bitmain are described as 'strategic equity investors' (14505) that supply advisory, procurement and equipment access (14503, 14418, 14417). The third tier is explicitly non-binding or structural. The Texas 200MW wind facility is a letter of intent 'subject to entry into definitive agreements' (14204, 14074). The 1.21 Gigawatts LLC with Antpool is a joint-venture formation document, not a customer order (14234, 14212). Its claim heading carries an '$8,000' figure with no stated unit or context.

To the company's credit, the unnamed '2 customers' and '4 customers' language (14506, 14412) was eventually resolved into named entities. Two of the four are affiliates of the company's own equity investors and suppliers, however. GMR alone holds roughly 50% of the 100MW first facility (14411, 14201). Bitmain is simultaneously an investor, an equipment vendor (14504), a source of 'leads for potential hosting customers' (14417), a hosting customer (14232), and, through Antpool, a JV partner (14234). The customer count rose from two to four between August and November 2021 (14506, 14412). Effective concentration nonetheless remained high, and it is substantially related-party in character.

The revenue arithmetic cannot be reconciled from this record. No claim discloses contract values, pricing per MW, term length or expected revenue. The only quantitative anchors are capacity shares. These are roughly 50% of 100MW for GMR (14411), 'the total available energy' at the first facility plus 85MW at a second site across all four customers (14363), and a 200PH equipment order (14416). The allocation of the remaining ~50MW among Bitmain, F2Pool and Hashing is never broken out. The 85MW second-site commitment is also not tied to a disclosed energy agreement comparable to the 100MW Amended and Restated Energy Services Agreement (14233). Diligence should therefore treat the capacity allocations as signed but unpriced, and the Texas and Antpool items as pipeline rather than backlog.

Applied Digital's customer record is best read as a sequence of commitments that grew in scale and specificity, while the list of counterparties stayed small and closely tied to its own investors. In August 2021 the company said it had agreements with two unnamed customers covering more than half of the energy under its Energy Services Agreement (14506), and separately said GMR and Bitmain had signed hosting contracts to pre-fill the first 100MW site (14507). By November 2021 the count had become four customers taking all of the first facility's energy plus 85MW at a second facility (14412, 14374, 14363), still without names. The four were named only in March 2022: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies, F2Pool Mining and Hashing LLC (14232, 14210, 14071). These are described as entered agreements, so they read as signed contracts rather than MOUs. Several other items on the record are not customer revenue at all. The Texas 200MW wind facility is a letter of intent that is explicitly subject to definitive agreements (14204, 14074). SparkPool's arrangements are services and advisory agreements (14503, 14418), and the SparkPool disclosure is cut off at "in the process of s" (14358), so its status is not established by the record. The Coinmint co-hosting deal runs the other way, with Applied Digital as the hosted party (14371). The 1.21 Gigawatts LLC is a joint-venture formation with Antpool (14234, 14212), not an offtake contract. Concentration is the main diligence issue. GMR alone is said to hold roughly 50% of the first 100MW (14411, 14201). GMR and Bitmain are also strategic equity investors (14505, 14377). The Bitmain group is at once investor, equipment supplier through the 200PH order (14504, 14416), hosting customer, and, through its affiliate Antpool, JV partner (14203). That means at least two of the four named customers are related to shareholders, and together they likely account for most of the contracted capacity. On revenue arithmetic, no contracted dollar values, per-MW pricing, term lengths or recognized hosting revenue appear on the record, so the announced capacity cannot be reconciled with any revenue figure. The 85MW second-facility allocation also refers to a site that had not yet been built when it was disclosed.

Applied Digital's customer record from August 2021 through April 2022 shows a real progression from unnamed counterparties to named hosting contracts. The signed hosting base is still narrow and heavily tied to the company's own equity holders. In August 2021 the company disclosed agreements with two unidentified customers covering more than 50% of available energy (claim 14506). It separately said GMR and Bitmain had entered hosting contracts to pre-fill the first planned 100MW site (claim 14507). By November the count had grown to four customers covering all of the first facility's energy plus 85MW at a second site, still without names (claims 14412, 14374, 14363). Only in the March 2022 filings are the four identified: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14210, 14200, 14071, 14060). These are the only counterparties on record described as signed hosting customers.

Much of the remaining partnership language is not customer revenue. SparkPool's role is services, design and advisory work, including advice on an Ether mining pool (claims 14503, 14418, 14358). Bitmain's 'preferred access' and hosting 'leads' describe supply and referrals (claim 14417). The Coinmint arrangement ran the other way, with Applied Digital as the hosted party (claim 14419). The Amended and Restated Energy Services Agreement is power supply, not demand (claims 14233, 14202). The Texas 200MW wind project is a letter of intent expressly subject to definitive agreements (claims 14204, 14074). The 1.21 Gigawatts LLC with Antpool is a joint-venture formation agreement; its 'up to 1.5GW' scope is aspirational capacity, not contracted load (claims 14234, 14203, 14212).

Concentration is the central issue. GMR alone is said to hold roughly 50% of the 100MW first facility (claims 14411, 14201, 14231). GMR and Bitmain are also strategic equity investors (claims 14505, 14377, 14378), and a Bitmain affiliate is the joint-venture partner (claim 14234). At least half, and plausibly most, of first-site capacity therefore sits with related or affiliated parties. How the remaining capacity and the second site's 85MW divide among Bitmain, F2Pool and Hashing is not disclosed.

The revenue arithmetic cannot be reconciled from this record. No contract values, per-megawatt pricing, tenors or revenue guidance appear in any claim. The only dollar figure, $8,000 in the heading of claim 14234, is attached to the joint-venture formation and should not be read as customer revenue. Diligence should obtain contract economics, terms and termination rights for the four named agreements before crediting the capacity claims with revenue.

Applied Digital's customer record through April 2022 is short, recycled heavily across filings, and dominated by counterparties that are also its shareholders. The August 2021 disclosures describe hosting contracts with GMR and Bitmain (claim 14507) and agreements with two unnamed customers covering more than 50% of available energy (claim 14506). By November the count had grown to four unnamed customers taking all of the first facility's energy plus 85MW at a second site (claims 14412, 14374, 14363). The four were not identified until March 2022: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14210, 14200). The only specific contract with a stated size is GMR's roughly 50% of the 100MW capacity (claims 14411, 14201). That figure is repeated nearly verbatim in at least seven filings, which makes it look like more activity than a single agreement. Softer arrangements are identified as such. The Texas 200MW wind facility is a letter of intent explicitly "subject to entry into definitive agreements" (claims 14204, 14074). The Antpool vehicle, 1.21 Gigawatts LLC, is a joint-venture LLC agreement rather than a customer contract (claims 14234, 14212). SparkPool is a services provider and adviser, not a revenue counterparty (claims 14503, 14418). The Coinmint co-hosting agreement runs the other way, with Applied Digital as the hosted party (claim 14419). The utility behind the Amended and Restated Energy Services Agreement is never named in these excerpts (claim 14233). Concentration is the central diligence point. GMR and Bitmain are strategic equity investors (claims 14505, 14377, 14378). Two of the four named hosting customers are GMR or Bitmain entities, and the JV partner Antpool is a Bitmain affiliate. Naming the customers therefore broadened the list mostly in name, since roughly half or more of contracted capacity sits with shareholder-affiliated parties. The revenue arithmetic cannot be reconciled from this record. No contract values, per-MW pricing, terms or revenue figures are disclosed. The only dollar amount is the $8,000 attached to the 1.21 Gigawatts LLC formation (claim 14234), a capitalization figure rather than revenue. The Bitmain 200PH order (claims 14504, 14416) is equipment Applied Digital buys, not a sale. Capacity statements are internally consistent: 50% for GMR fits within a four-customer 100MW allocation. Whether that capacity converts into revenue depends on energization and pricing, and the filings do not address either. The record provided ends at April 2022 (claim 14059), so later customer developments are outside this read.

Applied Digital's customer record through April 2022 contains signed hosting agreements, but the counterparties behind them were named late, and several prominent relationships are not customer contracts at all. In August 2021 the company said it had agreements with 2 customers covering more than 50% of available energy without naming them (claim 14506), although a companion passage identified GMR and Bitmain as hosting-contract parties (claim 14507). By November 2021 the count had grown to 4 customers covering all energy at the first facility plus 85MW at a second, still unnamed (claims 14412, 14374, 14363). Only in March 2022 were the four identified: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14210, 14200, 14071, 14060).

Much of the partnership language should not be read as customer demand. SparkPool is a services and advisory provider, not a hosting customer (claims 14503, 14418), and a later filing's partially quoted passage suggests the relationship was changing (claim 14358). Coinmint is a co-hosting host for the company's own rigs (claim 14419). The Bitmain 200PH order is an equipment purchase (claims 14504, 14416). The Energy Services Agreement is power supply (claims 14233, 14202). The Texas facility is only a letter of intent, expressly "subject to entry into definitive agreements" (claims 14204, 14074). The 1.21 Gigawatts LLC with Antpool, a Bitmain affiliate, is a signed entity agreement, but its "up to 1.5GW" scope is aspirational framework language rather than contracted capacity (claims 14234, 14203).

Concentration is high and has not diversified. GMR alone holds about 50% of the 100MW first facility, a figure repeated verbatim across every filing (claims 14411, 14231, 14201). GMR and Bitmain are also strategic equity investors (claims 14505, 14377, 14378), so two of the four customers, and likely most of the contracted load, are shareholder-affiliated. Bitmain's role as both customer and source of "leads for potential hosting customers" (claim 14417) deepens that dependence.

The record does not support any revenue reconciliation. No contract values, hosting rates, terms, or revenue figures appear in these claims. The only disclosed dollar figure is the $8,000 attached to the 1.21 Gigawatts formation (claim 14234), which is nominal. Capacity allocations in megawatts are not revenue, so reconciling the announced deals would require filed financials and the hosting agreements themselves.

Applied Digital's early customer record is more concrete than most development-stage hosting stories, but the named counterparties overlap heavily with the company's own shareholders. The signed-agreement language is firm from the first filing. In August 2021 the company said it had entered into agreements with two customers covering more than 50% of available energy (claim 14506), and that GMR and Bitmain had signed hosting contracts to pre-fill the first 100MW site (claim 14507). By November 2021 this became four customers covering all of the first facility plus 85MW at a second (claims 14412, 14374, 14363). Those four were not identified until March 2022, when the filings named JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14232, 14210, 14200). For roughly seven months, then, half of the customer base was described only by count. Softer language sits alongside these agreements. The Texas 200MW wind facility is a letter of intent explicitly 'subject to entry into definitive agreements' (claims 14204, 14074). The 1.21 Gigawatts LLC with Antpool, a Bitmain affiliate, is a joint-venture formation rather than a hosting sale, and the record ties it to only $8,000 (claim 14234). The SparkPool and GMR relationships are advisory and services arrangements, not revenue contracts (claims 14503, 14418, 14359). The Coinmint co-hosting agreement runs the other direction, with Applied Digital as the customer (claim 14419). Concentration is the central issue. GMR alone holds about 50% of the first 100MW (claims 14411, 14231). GMR and Bitmain are both strategic equity investors and hosting customers (claims 14505, 14377, 14378), and Bitmain is also the equipment supplier and a source of customer leads (claim 14417). Adding the unnamed customers only brought in F2Pool and Hashing, so arm's-length demand remains a minority of contracted capacity. The revenue arithmetic cannot be reconciled from this record. None of the claims discloses contract pricing, term, take-or-pay provisions, or expected revenue. The only figures that tie out are capacity figures, and the progression from more than 50%, to 100%, to 100MW plus 85MW is internally consistent. Diligence should obtain the per-MW economics and termination rights of the related-party contracts before treating the 185MW as contracted revenue.

Applied Digital's record from August 2021 to April 2022 describes hosting capacity pre-sold to a small group of counterparties. Most of those counterparties are also its shareholders. The filings repeat the same few statements many times, and the repetition makes the base of commitments look broader than it is. Only a subset of the language describes signed customer agreements. SparkPool's arrangements are services and advisory agreements in which Applied Digital is the buyer of expertise (14503, 14418). The Coinmint agreement makes Applied Digital the co-hosting customer, not the provider (14419, 14371). The Bitmain 200PH order is an equipment purchase (14504, 14416). The Texas 200MW wind facility rests on a letter of intent expressly subject to definitive agreements (14204, 14074). The Antpool vehicle, 1.21 Gigawatts LLC, is a joint-venture LLC agreement, not a hosting contract (14234, 14203).

The customer claims themselves became more specific over time. In August 2021 the company cited agreements with two unnamed customers covering more than 50% of available energy (14506). The same document said GMR and Bitmain would use the capacity of the first 100MW site (14507). By November 2021 this had become four unidentified customers taking all of the first facility's energy plus 85MW at a second site (14412, 14374, 14363). Only in March 2022 were the four named: JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (14232, 14210, 14200). GMR's contract is described consistently as roughly 50% of the 100MW (14411, 14231, 14201). At least half of first-site capacity therefore sits with one counterparty.

Concentration is worse than a count of four suggests. GMR and Bitmain are both disclosed as strategic equity investors (14505, 14376, 14377, 14378). Bitmain is also the equipment supplier and a source of hosting leads (14417). Its affiliate Antpool is the joint-venture partner (14234). A large share of the contracted base is therefore related-party or investor-affiliated. F2Pool and Hashing LLC are the only named customers with no disclosed equity tie, and their capacity shares are never given.

The revenue arithmetic cannot be reconciled from this record. No claim discloses contract value, price per megawatt or kilowatt-hour, term, take-or-pay provisions, or deposits. The only quantities are megawatt allocations. On megawatts the figures are broadly consistent, with GMR at about 50MW of the 100MW. The August 2021 statements are in mild tension with each other, since two customers above 50% sit beside GMR and Bitmain filling the whole site. The record provided also ends mid-sentence in April 2022, and later disclosures are not covered here.

Applied Digital's early customer record is thinner than its partnership vocabulary suggests. The instruments that read as signed are the hosting contracts, the Bitmain equipment order and the Antpool joint venture. In August 2021 the company disclosed agreements with two unnamed customers covering more than 50% of available energy (claim 14506). By November that had become four customers taking the full first-facility allocation plus 85MW at a second site (claims 14412, 14374, 14363). The Texas wind project is plainly softer: a November 2021 letter of intent that is expressly subject to definitive agreements (claims 14204, 14074). The 1.21 Gigawatts LLC formed with Antpool in January 2022 is an executed operating agreement, but it creates a development vehicle for data centers, not a revenue contract (claims 14234, 14203). Coinmint appears as a co-hosting counterparty, but there Applied Digital was the customer, not the provider (claims 14419, 14357).

On identification, the four hosting customers went unnamed through three successive November 2021 filings. They were first named in March 2022 as JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining and Hashing LLC (claims 14232, 14210). SparkPool, GMR and Bitmain were named throughout, but as "strategic partners" providing advisory, software and equipment-access services (claims 14502, 14503, 14417). SparkPool never appears as a hosting customer.

Concentration is the central diligence point. GMR alone held roughly 50% of the 100MW first site, a figure repeated verbatim from November 2021 onward (claims 14411, 14231, 14201). GMR and Bitmain together were described as pre-filling that site (claim 14507). Both are also strategic equity investors (claims 14505, 14377, 14378), and Antpool, the joint-venture partner, is a Bitmain affiliate (claim 14212). So at least half, and plausibly most, of contracted capacity sat with shareholders who were also service providers to the company. Widening the roster from two to four customers diluted this only modestly, since two of the four are affiliated with investors.

The revenue arithmetic cannot be reconciled from this record. None of the claims discloses contract pricing, term, take-or-pay provisions or expected revenue per megawatt. The capacity figures (100MW, 85MW, 50% to GMR) cannot be converted into a revenue run-rate. The $8,000 figure attached to the Antpool entity (claim 14234) is shown without units or context. The record supplied also ends mid-entry in April 2022, so later deals and any subsequent revenue disclosure fall outside this read.

Applied Digital's record through August 2022 contains a small number of genuinely signed, named agreements. The same language is repeated across several filings, which can make the deal flow look larger than it is. The core signed customer base is a set of hosting agreements with JointHash Holding Limited (a GMR subsidiary), Bitmain Technologies Limited, F2Pool Mining, Inc. and Hashing LLC (claims 14210, 14014, 14037). By August 2022 a second GMR subsidiary, Spring Mud, LLC, had been added, taking the count to five (claims 13731, 13781). GMR alone is described as having signed for approximately 50% of the 100MW first facility (claims 14201, 13773). The five-year, 200MW hosting contract with Marathon Digital Holdings, dated July 12, 2022, is the largest named, signed customer agreement on record (claims 13855, 13780).

Several items carry weaker language or leave the counterparty out. The Texas facility with 200MW of wind power is only a letter of intent with no named counterparty, and it is expressly subject to definitive agreements (claims 14204, 14074). The Amended and Restated Energy Services Agreement covering 100MW, and a separate agreement with a utility, are signed, but neither counterparty is identified in the quoted text (claims 14202, 14013). The 1.21 Gigawatts LLC with Antpool is an executed joint-venture agreement, not a customer contract. Its figure of up to 1.5GW describes intended development capacity, not contracted demand (claims 14212, 13730). The March 2021 Services Agreement with GMR, SparkPool and Valuefinder is an advisory arrangement (claims 14019, 13783). The strategic partnership language around SparkPool, GMR and Bitmain describes equity holders, not buyers (claims 14018, 13771). SparkPool has since ceased operations and forfeited its shares (claim 13772).

Concentration moved sharply. At first the entire 100MW base was held by five entities, and two of them were GMR subsidiaries controlling roughly half the capacity. GMR and Bitmain were also equity partners, so the early book was largely related-party. The Marathon contract alone is twice the size of the first facility. If contracted megawatts are additive, Marathon now accounts for about two-thirds of roughly 300MW, which shifts dependence from insiders to a single arm's-length counterparty.

The revenue arithmetic cannot be reconciled from this record, because no contract values, pricing or recognized revenue are disclosed alongside the deals. Contracted megawatts also already exceed the only facility capacity described as under development, so the Marathon commitment depends on capacity not yet evidenced. The record supplied appears truncated after claim 13770.

Applied Digital's customer record through September 2022 is mostly signed agreements rather than MOU, LOI, or discussion language. It is concentrated among counterparties who are also the company's shareholders, advisers, or joint-venture partners. The provided record ends partway through claim 13613, so this read covers only the fiscal 2022 disclosures.

The firmest agreement is the five-year hosting contract with Marathon Digital Holdings for 200 MW, dated July 12, 2022, which the company describes as entered into (13855, 13733, 13780). Five hosting customers are named individually: JointHash Holding and Spring Mud (both GMR subsidiaries), Bitmain Technologies, F2Pool Mining, and Hashing LLC (13731, 13754, 13673). The company also names an earlier co-hosting agreement with Coinmint (13735) and an executed Services Agreement with GMR, SparkPool, and Valuefinder (13768, 13769, 13770). Softer language appears in two places. Bitmain is described as a "strategic partner" with no contract terms quoted (13750, 13771). The 1.21 Gigawatts LLC with Antpool is a signed joint-venture agreement, but its "up to 1.5 GW" figure describes development capacity, not contracted demand (13748, 13730). No counterparty is left unnamed or described only as an "anchor customer."

Concentration was high from the start and became more related-party in character. The five named customers were to account for 100% of first-facility hosting revenue (13756). GMR alone, through its two subsidiaries, held about 50% of the 100 MW capacity (13732, 13755). GMR also received shares under the Services Agreement (13680), and Bitmain's affiliate Antpool is the joint-venture partner (13766). So customer, adviser, shareholder, and partner roles overlap. SparkPool ceased operations and forfeited 4,965,432 shares in June 2022 (13772, 13681), which shows that these counterparties can fail. Marathon's 200 MW is twice the first facility's 100 MW, so it shifts concentration toward a single unaffiliated customer. That capacity depends on facilities still to be built, as does the 85 MW promised to the five customers at the second facility, which is conditioned on the facility being "built and able to provide such energy" (13646).

The revenue arithmetic cannot be reconciled from this record. None of the quoted claims give contract values, per-MW pricing, or recognized revenue. The "$8,000" in the label for 13747 is not supported by its verbatim quote. The megawatt commitments exceed built capacity, so any revenue run-rate implied by the announced deals should be treated as contingent on construction.

Applied Digital's customer record, as captured here, is drawn almost entirely from filings dated between late August and late September 2022, and the same handful of agreements is restated across them. Readers should treat the volume of claim ids as repetition, not breadth. The input also ends mid-quote at claim 13616, so any later disclosures fall outside this read. Nothing in the record uses MOU, LOI or 'discussions' language. The weaker instruments are different in kind. The 1.21 Gigawatts LLC agreement with Antpool, a Bitmain affiliate, is a signed joint venture document (claims 13729, 13782, 13644). Its headline capacity, however, is aspirational: the vehicle 'will develop, acquire, construct' datacenters of 'up to 1.5 gigawatts' (claims 13730, 13748). That figure should not be read as contracted demand. The March 2021 Services Agreement with GMR, SparkPool and Valuefinder is an advisory arrangement paid in equity, 7,440,148 shares (claim 13680), not a revenue contract. Bitmain is described as a 'strategic partner' (claims 13736, 13771) without disclosed commercial terms beyond its separate hosting agreement. Every counterparty is named, and none is hidden behind 'anchor customer' phrasing. The named signed hosting contracts are five agreements covering JointHash and Spring Mud (both GMR subsidiaries), Bitmain, F2Pool and Hashing LLC (claims 13731, 13754, 13760), plus a five-year, 200 MW contract with Marathon Digital dated July 12, 2022 (claims 13733, 13757). Concentration is the central diligence issue. At the first facility, those five customers were to provide 100% of hosting revenue (claim 13756). GMR alone held roughly half of the 100 MW (claims 13732, 13773), and the GMR and Bitmain/Antpool groups overlap as customers, advisers, shareholders and JV partners. Counterparty durability is already visible as a risk: SparkPool ceased operations and forfeited 4,965,432 shares in June 2022 (claims 13772, 13681). The Marathon contract, at twice the first facility's capacity, would move concentration from the GMR/Bitmain cluster toward a single US miner rather than diversify it. On revenue arithmetic, the record discloses capacity, not dollars, so reconciliation is not possible. The 100 MW first facility, the 85 MW promised to the five customers at a second facility 'once it is built' (claim 13646) and the 200 MW Marathon commitment together exceed any operational capacity described. No pricing, term revenue or backlog figure is given. The lone '$8,000' attached to claim 13747 carries no unit or context in the quoted text and cannot be tied to any deal.

Applied Digital's record separates into three tiers, and most of the volume is repetition rather than new business. The signed, named hosting agreements are real. Five counterparties are identified at the first facility: JointHash and Spring Mud (both GMR subsidiaries), Bitmain Technologies, F2Pool Mining and Hashing LLC (claims 13645, 13615, 12603). GMR is separately described as holding roughly 50% of the 100 MW Jamestown capacity (13647, 12570). The Marathon Digital five-year contract for 270 MW, later 309.6 MW, is the most concrete disclosure in the file (12633, 12569, 12587). By contrast, the five customers at the second facility are tied to 85 MW 'once it is built' and are not named in the quoted text (13646). The 'first major AI customer' behind a contract 'worth up to $180 million' is never identified (12640).

The joint ventures belong in a different category. The 1.21 Gigawatts LLC with Antpool is a signed entity agreement, but its 'up to 1.5 GW' of datacenters is prospective development language, not contracted capacity (13644, 13558, 12551). Highland Digital with GMR 'may acquire' hardware (12725). SAI Computing is 98% owned by the company, which makes Foundry closer to a minority participant than a customer (12634). The March 2021 Services Agreement paid GMR, SparkPool and Valuefinder in stock (12636 to 12638). SparkPool later ceased operations and forfeited its shares (12606). This means GMR is simultaneously a shareholder, advisor, JV partner and roughly half of Jamestown hosting, and readers should weigh that related-party overlap.

Concentration has shifted rather than diversified. In 2022 the book was dominated by GMR-linked entities at a 100 MW site. By the 2023 filings Marathon's 309.6 MW is about three times the entire Jamestown facility, making a single counterparty the center of gravity. One unnamed AI customer then became the headline growth driver. Many 2023 claims (12571, 12614, 12615, 12616) restate 2022 text verbatim and should not be read as incremental wins.

On arithmetic, Marathon's facility split of 90 MW in Texas and 180 MW in North Dakota reconciles to the original 270 MW (12605). The additional 39.6 MW is not allocated in the quoted text. No record here discloses contract pricing, recognized revenue or backlog for the hosting agreements. The only dollar figure is a ceiling ('up to $180 million') attached to an unidentified counterparty. The announced deals therefore cannot be reconciled to reported revenue from this record, and the AI figure should be treated as a maximum rather than contracted value.

Applied Digital's record through mid-2023 contains a small core of signed, named hosting contracts, surrounded by joint ventures, advisory arrangements and unnamed AI deals that should not be read as equivalent. The firmest items are the five-year Marathon Digital hosting contract for 270 MW, split into 90 MW in Texas and 180 MW in North Dakota, plus a later 39.6 MW add-on at Jamestown (claims 12569, 12605, 12588). Also firm are the Jamestown agreements with five named customers: JointHash and Spring Mud, both GMR subsidiaries, plus Bitmain, F2Pool and Hashing LLC (claims 12615, 12603). Many claim ids repeat these same few facts across overlapping filings, so they are not independent confirmations.

The joint ventures belong in a different category. Highland Digital with GMR 'may acquire' hardware (claim 12725). The 1.21 Gigawatts LLC with Antpool describes datacenters of 'up to' 1.5 GW (claims 12616, 12591). That is development intent, not contracted load. SAI Computing with Foundry is 98% owned and consolidated (claim 12593), so it is essentially an internal vehicle rather than a counterparty. The March 2021 services agreements with GMR, SparkPool and Valuefinder were paid in stock and are not customer revenue (claims 12636-12638). SparkPool ceased operations and forfeited 4,965,432 shares (claim 12606).

The AI pivot reverses the naming pattern. The first AI customer, 'up to $180 million over 24 months', and the second, 'up to $460 million over 36 months' and described as the largest customer, are both unidentified (claims 12640, 12409, 12399). Both figures are ceilings rather than committed minimums. At full value they imply about $90 million and $153 million a year. One wording point is worth checking: claim 12409 calls Sai Computing Holdings 'wholly-owned', while SAI Computing LLC is 98% owned. These may be different entities, but the record does not settle it.

Concentration has moved from a named, GMR-heavy crypto base to unnamed AI buyers. GMR sits on every side of the relationship: it is a joint-venture partner, an advisory counterparty paid in stock, and, through two subsidiaries, roughly half of Jamestown's 100 MW (claim 12639). The revenue arithmetic does not fully reconcile. Marathon's 90 plus 180 MW correctly totals 270 MW. However, about 50 MW for GMR plus Marathon's 39.6 MW leaves only about 10 MW for Bitmain, F2Pool and Hashing inside a 100 MW facility (claim 12602). No realized revenue is disclosed against either AI ceiling. The claim record provided also ends mid-entry at claim 12390.

Applied Digital's record separates cleanly into two eras, and they differ in how much is disclosed. The crypto-hosting book is built on signed, named agreements. The AI book that followed rests on anonymous counterparties and ceiling-value language. The firmest contract is the five-year Marathon Digital hosting agreement of July 2022 for 270 MW, split 90 MW in Texas and 180 MW at a second North Dakota site, plus a later 39.6 MW at Jamestown (claims 12613, 12569, 12605, 12588). At Jamestown the company names five customers: JointHash and Spring Mud, both GMR subsidiaries, Bitmain, F2Pool and Hashing LLC. It also discloses a 100 MW energy services agreement (12571, 12603, 12602). Many claim ids repeat the same filing text and should be read as single disclosures rather than corroboration.

The joint ventures are a different category. 1.21 Gigawatts with Antpool "will develop" up to 1.5 GW (12632, 12551), and Highland Digital with GMR "may acquire" hardware (12725). Both are framework vehicles with no disclosed offtake. SAI Computing is 98% owned and consolidated (12634, 12593), so Foundry is a minority partner, not a customer. GMR is also a related party. It received 7,440,148 shares under a 2021 advisory agreement, is a JV partner, and holds roughly 50% of Jamestown capacity (12636, 12614). A co-advisor, SparkPool, ceased operations and forfeited shares (12606). That makes GMR's customer status a concentration risk as much as a validation.

The AI contracts are where identification stops. The "first major AI customer" is worth up to $180 million over 24 months (12640, 12399), and a "second AI customer" is worth up to $460 million over 36 months (12409, 12422). Neither is named, and the second is described as the company's largest customer. Concentration has therefore moved from named miners, chiefly Marathon and GMR, to a single unidentified counterparty.

The arithmetic does not fully reconcile. At the ceiling, the AI deals imply about $90 million and $153 million a year respectively. The record discloses no recognized revenue, prepayments or GPU deployment against which to test those figures, and "up to" should be read as a maximum, not a backlog. On capacity, Marathon's 39.6 MW plus GMR's roughly 50 MW leaves only about 10 MW of Jamestown's 100 MW for Bitmain, F2Pool and Hashing. That is tight enough to warrant confirmation. The final claim on record (12390) is truncated and could not be assessed.

Applied Digital's record as of mid-2023 shows a signed, named hosting book in crypto mining. Its artificial intelligence (AI) pipeline, which draws most of the attention, is unnamed and described only by ceilings. The firmest agreement is the five-year Marathon Digital hosting contract dated July 12, 2022. It covers 270 MW, split into 90 MW in Texas and 180 MW in North Dakota, and Marathon later added 39.6 MW at Jamestown, for 309.6 MW in total (claims 12569, 12605, 12588). Jamestown also has five named counterparties: JointHash and Spring Mud, both GMR subsidiaries, plus Bitmain, F2Pool and Hashing LLC. That site is backed by an Amended and Restated Energy Services Agreement for 100 MW (claims 12571, 12603, 12602).

The AI side reads differently. The first AI Cloud customer is worth "up to $180 million" over 24 months (claim 12640). The second is worth "up to $460 million" over 36 months and is called the company's largest customer (claims 12409, 12422). Neither counterparty is identified, and both figures are maximums rather than committed minimums. At the ceiling they imply roughly $7.5 million and $12.8 million a month. The record includes no revenue recognition, backlog or prepayment disclosure that would confirm either figure, so these deals cannot yet be reconciled against reported results.

Several other items are frameworks rather than customer revenue. The Antpool/Bitmain vehicle 1.21 Gigawatts "will develop" facilities of up to 1.5 GW (claims 12590, 12591). Highland Digital with GMR "may acquire" hardware (claim 12729). SAI and Sai-Foundry are 98%-owned joint venture shells, not third-party contracts (claims 12593, 12420).

Concentration is significant and partly related-party. GMR holds about 50% of Jamestown's 100 MW (claim 12639). It is also a joint venture partner and received 7,440,148 shares under the 2021 advisory agreement (claim 12636), so it is customer, partner and shareholder at once. SparkPool, another advisory counterparty, ceased operations and forfeited 4,965,432 shares (claim 12606), a reminder that these relationships have already unwound once.

The Jamestown capacity math also needs a check. GMR's roughly 50 MW plus Marathon's 39.6 MW leaves only about 10 MW of the 100 MW energy agreement for Bitmain, F2Pool and Hashing combined. Either those allocations are very small or the reported figures overlap. Finally, the record provided ends partway through claim 12420, so later developments are not assessed here.

Applied Digital's record through mid-2023 splits into a named, signed crypto-hosting book and an AI Cloud book whose counterparties are not identified. On the hosting side, the agreements are executed and the counterparties are named. Marathon Digital signed a five-year contract for 270 MW, split 90 MW in Texas and 180 MW in North Dakota, and later added 39.6 MW at Jamestown, for a total of 309.6 MW (12569, 12605, 12588). The five Jamestown customers are listed by name: JointHash and Spring Mud, both GMR subsidiaries, plus Bitmain, F2Pool and Hashing LLC (12571, 12603). GMR is said to hold roughly 50% of the 100 MW site (12614). Many of these claim ids repeat one filing passage, so the volume of claims does not amount to independent confirmation. The AI Cloud contracts are different in kind. The first is worth "up to $180 million" over 24 months (12640) and the second "up to $460 million" over 36 months (12409, 12422). Neither customer is named, and "up to" describes a ceiling, not committed revenue. Supermicro is described as a party Applied Digital is "working with" (12406, 12419), which reads as a supply relationship, not a customer. The Antpool venture (12567, 12591) and the 98%-owned Foundry venture (12593, 12420) are equity structures, not revenue contracts. The 1.5 GW capacity figure is a development aim. Concentration has grown heavier and harder to see over this period. The early book depended on related parties. GMR, SparkPool and Valuefinder received shares under advisory agreements (12636, 12600), and SparkPool stopped operating and forfeited 4,965,432 shares (12606). GMR is both an equity recipient and the largest Jamestown tenant, and Bitmain is both a customer and the affiliate of a JV partner. Marathon then became the largest named counterparty. After that, an unnamed AI customer was described as the "largest customer" (12409). No revenue figures appear in these claims, so the revenue arithmetic cannot be reconciled. At the ceiling, the AI contracts imply about $90 million and $153 million a year. The capacity figures also raise a question. Marathon's 39.6 MW at Jamestown, added to GMR's roughly 50 MW, leaves only about 10 MW of the 100 MW site for Bitmain, F2Pool and Hashing. That points to capacity being reallocated or customers leaving, and the record does not explain which. The claim record is also cut off after claim 12412, so later disclosures are not covered here.

Applied Digital's record through its August 2023 filing is smaller than the claim count suggests, because most entries restate the same few agreements across successive filings. The supplied record breaks off partway through claim 12331, so any later deals are outside this read. The most concrete commitments are in crypto hosting. The Marathon Digital contract is a named, dated, five-year agreement (12569, 12547). One labelling problem needs fixing: the claim titles cite 309.6 MW, but the quoted text supports only 270 MW. Jamestown's 100 MW was contracted to five named counterparties (12571), and GMR entities took roughly half (12570, 12548). The customer list shifted within three months. By August 2023, JointHash and F2Pool no longer appear, while Arrakis and a Global Operating entity do (12340). That churn is worth confirming against the contracts themselves. Because 270 MW dwarfs the 100 MW Jamestown site, Marathon was the dominant hosting counterparty.

The AI cloud pivot moved concentration toward customers who are never named. The first 'major AI customer' agreement is worth 'up to' $180 million over 24 months (12390, 12399, 12412, 12308, 12320, 12328, 12341). The second is worth 'up to' $460 million over 36 months and is described as the company's largest customer (12409, 12422, 12309, 12321, 12329, 12342). These are signed agreements, but they are not identified counterparties. The 'up to' wording signals a ceiling rather than committed revenue, and the 'significant pre-payment' is never quantified. At full ceiling, the two contracts imply about $90 million and $153 million a year, roughly $243 million combined. The record contains no recognized revenue, backlog, or deferred-revenue figure to test that against, so the arithmetic cannot be reconciled from these claims. That gap is the central diligence question.

Several headline relationships are not customer agreements at all. Super Micro is framed only as 'working with' the company (12406, 12419, 12314, 12323, 12336, 12348). The HPE item is a vendor 'collaboration' to supply hardware (12315, 12324, 12337, 12349). Both are supply-side relationships. The 1.21 Gigawatts venture with Antpool describes capacity of 'up to 1.5 GW,' which is aspirational rather than contracted (12567, 12551). The Foundry partnership began as the SAI joint venture (12568, 12552) and was later restructured into Sai-Foundry, which Applied Digital owns 98% of and consolidates (12317, 12338, 12407, 12420). That leaves Foundry a nominal partner. The two colocation contracts are also supply-side, and those providers are unnamed (12316, 12331).

Applied Digital's record contains fewer distinct signed customer agreements than its claim count suggests, because the same few announcements recur across filings. The two AI cloud contracts from mid-2023 are the only dollar-denominated customer commitments. The first is worth up to $180 million over 24 months (claims 12320, 12308, 12328, 12341), and the second is worth up to $460 million over 36 months (claims 12321, 12309, 12329, 12342). Neither counterparty is named. Both values are ceilings rather than committed minimums. In the 2024 filings the dollar figures no longer appear, and the contracts are restated only as a 'first major AI customer' and a '36-month contract with a second customer' (claims 6095, 6096, 6129, 5983). That quiet removal deserves follow-up.

Most of the other named relationships are on the supply side rather than the revenue side. Super Micro and HPE began as a 'working with' arrangement and a 'collaboration' (claims 12314, 12323, 12336, 12348, 12315, 12324, 12337, 12349). By 2024 the filings call them 'partnerships' (claims 6126, 6089, 6127, 6090), but no new terms are disclosed. Dell appears through supply agreements (claims 6091, 5991), and the two colocation contracts have unnamed counterparties (claims 12316, 12331). The largest prospective deal is a 400 MW capacity lease with an unidentified 'US-based hyperscaler.' It was still described only as an exclusivity LOI in both April and August 2024 (claims 3233, 3246), so it should not be treated as contracted. Together AI is the only newly named AI customer, and no value or term is attached (claims 3234, 3243).

Concentration has moved in two directions. In 2023 the crypto hosting business named at least six counterparties, including Marathon, Bitmain and Spring Mud, a GMR subsidiary (claim 12340). By 2024 that had narrowed to 'four crypto-mining customers,' none of them named (claims 6128, 6097). GMR is also party to a 2021 advisory services agreement (claim 6088), which suggests a related-party overlap with a hosting customer. The AI cloud revenue base, meanwhile, depends on two anonymous contracts plus Together AI.

The revenue arithmetic cannot be reconciled from this record. At full ceiling value, the two AI contracts imply about $7.5 million and $12.8 million a month, or roughly $243 million a year combined. No recognized-revenue or backlog figures are included here to test that against. The later removal of the contract values makes it more important to compare these ceilings with reported segment revenue. The record supplied was also truncated after claim 5863, so later developments may be missing.

Applied Digital's customer roster presents substantial disclosure opacity that complicates revenue verification. The company names only one Cloud Services customer explicitly—Together AI (claims 3234, 3243)—while disclosing a "second customer" in the Cloud Services Business with a 36-month contract (claims 6129, 6096) without identification. Claim 5903 states Cloud Services accounts for approximately 17% of fiscal 2024 revenue across these two customers, implying significant per-customer concentration behind an unnamed counterparty. The crypto mining segment remains entirely unnamed: seven customers holding 3–5 year contracts (claims 5879, 5880), with claim 5880 describing this segment as accounting for "the majority of the revenue." The company provides no basis to verify which of the seven crypto counterparties, if any, have actually signed binding agreements versus pilot arrangements.

The company's supply partnerships with Super Micro, HPE, and Dell (claims 5877, 5876, 5878) are not customer agreements but hardware-supplier relationships labeled as strategic "partnerships." These generate no direct revenue to Applied Digital and should not be confused with customer wins.

A critical threshold issue: the "US-based hyperscaler" 400 MW capacity lease exists only as a letter of intent and exclusivity agreement (claims 3233, 3246, 5906), not a signed customer contract. The LOI language explicitly denies binding status.

Counterparty concentration has worsened, not improved. In May 2024 the company cited four crypto mining customers (claim 6128); by August 2024 this rose to seven (claims 5879, 5880). This growth came entirely from unnamed operators. In Cloud Services, Together AI is the sole named customer; the second customer remains undisclosed despite a claimed 36-month contract.

Revenue reconciliation: if crypto represents "the majority" and Cloud Services represents 17%, the residual segment allocation is unspecified. Without naming the seven crypto counterparties or the unnamed second Cloud Services customer, investors cannot perform counterparty credit, concentration, or attrition analysis. The LOI with the hyperscaler should carry zero revenue assumption. Together AI alone is the verifiable named customer win in the company's disclosure record.

Applied Digital's disclosed customer base splits sharply between named accounts, quantified-but-unnamed customers, and supplier partnerships masquerading as customer traction. Together AI stands as the sole publicly named Cloud Services customer (claims 3234, 3243), announced April 2024. A second Cloud Services customer entered a 36-month contract in June 2023 (claims 6129, 6096) but remains unnamed; collectively these two accounts generated approximately 17% of fiscal 2024 revenue (claim 5903). The company's crypto mining segment comprises seven customers under multi-year hosting contracts (claims 5879, 5897, 5880), though this count previously appeared as four (claim 6128), suggesting either restatement or reclassification. Notably, all seven crypto customers are described only by segment category, never by name.

The high-profile 400 MW hyperscaler arrangement is explicitly an executed letter of intent with a "US-based hyperscaler" (claims 3233, 3246, 5906), not a signed customer agreement. This deal remains in exclusivity but lacks execution, creating meaningful contingency risk. Meanwhile, HPE and Super Micro are characterized as partnerships announced in June and May 2023 respectively (claims 6127, 6126), yet both are framework arrangements for equipment provision to customers rather than direct revenue contracts. Dell's relationship is a supply agreement for servers (claim 6091), not a customer arrangement. These distinctions matter: HPE and Super Micro provide no revenue themselves; they enable delivery to the actual customers (Together AI and the unnamed second Cloud Services account).

Customer concentration has intensified rather than diversified. Cloud Services remains confined to two accounts generating 17% of revenue, while crypto mining represents the revenue majority across seven unnamed counterparties. The single named growth customer, Together AI, entered in April 2024, yet no subsequent customer announcements have been disclosed through August 2024. The gap between announced partnerships (HPE, Super Micro, Dell) and named customer wins (one) suggests promotional messaging may emphasize enabling infrastructure over actual customer acquisition. The unnamed hyperscaler LOI—positioned as transformational 400 MW capacity—remains unsigned as of late August 2024, making it aspirational rather than binding revenue. Applied Digital's public record thus reveals heavy concentration among unnamed customers, a single named growth customer, framework partnerships rather than customer relationships, and one major unsigned arrangement, creating visibility gaps around the company's actual customer momentum and revenue predictability.

Applied Digital's disclosed customer base reveals material gaps between executed agreements and partnership language. Together AI stands as the sole named Cloud Services customer (claims 3234, 3243), yet company disclosures refer to two Cloud Services customers generating approximately 17% of fiscal 2024 revenue (claim 5903), leaving the second customer entirely unnamed. The 400 MW hyperscaler arrangement is explicitly a letter of intent with exclusivity (claims 3233, 3246, 5861), not a signed capacity lease, creating contingency risk absent from simpler framing.

Supply and technology partnerships with Super Micro, HPE, and Dell (claims 6126, 6127, 6091) represent supplier or channel relationships rather than customer contracts; these vendors do not appear in the revenue attribution. The May 2023 services agreement with GMR and others (claim 6088) involves advisory services and strategic investment positioning, not Cloud Services or hosting revenue.

Crypto mining customer concentration has shifted materially. By May 2024, Applied Digital disclosed four named-but-unidentified crypto customers under multi-year contracts (claims 6128, 6097). By August 2024, that figure rose to seven crypto customers (claims 5869, 5879, 5880, 5897), representing a 75% increase without announcement or explanation. The company states this segment accounts for the majority of revenue (claim 5880), yet none of the seven customers are named. This concentration carries reputational and cyclicality risk that transparent customer identification would surface.

Revenue arithmetic presents opacity. Cloud Services at 17% of fiscal 2024 revenue across two customers, one of which is Together AI, suggests Together AI alone generates ~8.5% of total revenue or the two customers are unevenly weighted. Crypto mining's stated status as the majority revenue segment, paired with seven unnamed counterparties, means the vast majority of Applied Digital's disclosed revenue derives from customers not identified to investors. The upper Midwest utility providing contractual energy cost ceilings (claim 6092) is a cost-control mechanism, not revenue.

The company's disclosure strategy distinguishes sharply between infrastructure partnerships (named vendors, operational contracts) and revenue customers (Together AI named, crypto and second Cloud Services customers anonymized). This segmentation limits diligence visibility into customer concentration, churn risk, and contract stability across the company's largest revenue segment.

Applied Digital's customer disclosure record reflects significant opacity regarding actual signed revenue contracts versus aspirational partnership language. Of the named counterparties, only Together AI is explicitly identified as a paying Cloud Services customer; this single named customer, alongside one unnamed Cloud Services customer secured in June 2023, generated approximately 17% of fiscal 2024 revenue. The company's material revenue driver—crypto mining services—remains entirely unnamed across all claims; Applied Digital discloses serving seven such customers under three-to-five-year contracts (initially four as of May 2024, later revised to seven by August 2024), yet provides no customer identities, suggesting either confidentiality agreements or strategic vagueness regarding this majority-revenue segment. The purported partnerships with Super Micro, HPE, and Dell are not customer contracts but supplier or infrastructure relationships, though the language "partnership" and "deliver the Company's cloud services to our customers" risks conflating vendor relationships with end-customer traction. Most significantly, the 400 MW hyperscaler capacity lease announced in April 2024 remains at letter-of-intent stage through August 2024; the repeated emphasis on "exclusivity" and "LOI" throughout the claims record demonstrates this deal has not matured to a signed contract. The GMR arrangement from March 2021 is explicitly a services advisory agreement, not a customer contract. Counterparty concentration has shifted toward unnamed crypto mining (seven customers providing majority revenue) while Cloud Services remains heavily concentrated with two customers, one named and one opaque. Revenue reconciliation presents gaps: crypto mining dominates by volume; Cloud Services contributes 17%; the remaining revenue base is undisclosed and unattributed to named or even described counterparties. The company's investor disclosures distinguish between LOI status and signed contracts in technical language but obscure this distinction through repetitive partnership announcements lacking revenue signals. The shift from four to seven crypto customers between May and August 2024 without named additions suggests either previously undisclosed customers or consolidation of existing relationships, neither of which clarifies actual counterparty dependence. Together AI represents the sole transparent customer win; the remainder of Applied Digital's revenue model lacks the specificity necessary to assess deal quality or concentration risk.

Applied Digital's customer disclosures reveal significant ambiguity between signed contracts and preliminary agreements, with material counterparties remaining unidentified. The company names Together AI, Sai Computing, and an unnamed second Cloud Services customer as its three identified customers in that business segment, yet claims to serve seven crypto-mining customers under multi-year contracts without naming any of them—a critical gap that obscures revenue concentration (claims 5879, 5897). The flagship 400 MW capacity arrangement is explicitly a letter of intent with an unnamed US-based hyperscaler, not a signed lease, despite repeated framing as an "executed" LOI that constitutes exclusivity (claims 3233, 3246, 5890). The company discloses that Cloud Services represented approximately 17 percent of fiscal 2024 revenue while crypto mining accounts for the "majority," yet provides no itemization of revenue by customer, making it impossible to verify whether the unnamed seven crypto miners or the three named Cloud Services customers drive the larger portion (claims 5903, 5880).

The counterparty composition has shifted materially over the disclosure window: in May 2024, Applied Digital referenced four crypto-mining customers; by August 2024, this had grown to seven (claims 6128, 5879). The expansion is unexplained and undifferentiated—no announcement of new customer wins appears in the claim set, suggesting either internal reclassification or silent backfill of the customer base. Super Micro and Hewlett Packard Enterprise are described as "partnerships" for component and technology delivery rather than revenue customers, positioning them as enablers rather than end customers, though the language blurs the distinction (claims 5900, 5901). The Services Agreement with GMR dates to March 2021 and is advisory in nature, not a capacity or hosting contract (claim 6088). Notably absent from the named customer list despite being a technology co-investor is the identity of any major hyperscaler, which remains veiled as "US-based" and under LOI-only status—a material omission given the 400 MW capacity is central to growth guidance. The repetition of nearly identical claims across multiple filing dates (claims 5879, 5880, 5897 are substantively equivalent) suggests syndication across documents rather than fresh disclosure of new customer wins, limiting confidence in the underlying customer count and revenue correlation.

Applied Digital distinguishes sharply between signed customer contracts and non-binding arrangements, though disclosure of counterparty identity remains incomplete. The company operates three distinct revenue streams: crypto mining hosting, cloud services, and equipment supply relationships. In crypto mining, APLD serves seven customers under signed multi-year contracts (3–5 years), which the company identifies as its largest revenue segment, yet names none of these counterparties individually. The crypto customer base reportedly grew from four (claim 6128, April 2024) to seven (claim 5897, August 2024), suggesting either portfolio expansion or a restatement. The Cloud Services Business comprises exactly two named or nameable customers: Together AI, explicitly identified across claims 3234 and 3243 as a signed customer, and two unnamed parties, one characterized as the 'first major AI customer' secured in May 2023 and another under a 36-month contract executed June 2023 (claims 5983, 6129). APLD quantifies Cloud Services at approximately 17% of fiscal 2024 revenue (claim 5903), implying the unnamed customers represent meaningful ARR. The 400 MW hyperscaler relationship, repeatedly cited (claims 3233, 3246, 5890), remains a letter of intent under exclusivity, not a signed capacity lease—a material distinction, since LOIs carry no binding revenue obligation. Equipment partnerships with HPE (June 2023), Super Micro (May 2023), and Dell represent supplier arrangements, not customer contracts; these relationships provide infrastructure but do not constitute direct revenue relationships. Counterparty concentration has intensified: seven anonymous crypto miners and one named cloud customer (Together AI) generate the vast majority of disclosed revenue, with two additional unnamed cloud customers representing opaque exposure. The arithmetic presents a puzzle: if crypto mining is described as the 'majority' of revenue and Cloud Services accounts for 17%, roughly 83% should flow from crypto—yet the identity, revenue per contract, and contract profitability of the unnamed miners remains undisclosed. The LOI with the unidentified hyperscaler carries no current revenue impact and should not be conflated with signed customer commitments.

Applied Digital's customer disclosure pattern distinguishes between signed contracts and non-binding commitments, yet reveals material transparency gaps in revenue concentration. The company names Together AI as a Cloud Services customer (claims 3234, 3243) but reports only two total Cloud Services customers accounting for approximately 17% of fiscal 2024 revenue (claim 5903); the second customer operates under a signed 36-month contract executed in June 2023, yet remains entirely unnamed (claims 6129, 6096). The revenue majority derives from seven crypto-mining customers, all under signed multi-year hosting contracts (claims 5897, 5879, 5869), but without counterparty identification. This unnamed cohort expanded rapidly: from four customers disclosed in May 2024 (claims 6128, 6097) to seven by August 2024, representing accelerating growth within an opaque customer base. Applied Digital's flagship strategic announcement—a 400 MW capacity lease—is not executed. Claims 3233, 5890, and 5906 explicitly reference an "executed letter of intent" with an unnamed US-based hyperscaler, meaning the transaction remains contingent and unbound. Separately disclosed relationships with Super Micro (claims 6126, 5877), HPE (claims 6127, 5876), and Dell (claims 6091, 5865) are infrastructure partnerships or supply agreements, not customer revenue contracts. An Electric Service Agreement with an unnamed upper Midwest utility (claims 6092, 5885) addresses energy cost management, not customer acquisition. The disclosed revenue arithmetic reconciles internally—17% from Cloud Services, plurality from crypto mining—but provides no transparency into the unnamed counterparties driving financial performance. Together AI, Applied Digital's sole named customer, represents a secondary position within its own Cloud Services segment. The absence of named crypto-mining customers despite their revenue plurality, combined with the LOI-only status of the 400 MW hyperscaler deal, suggests that investors cannot assess counterparty concentration risk or revenue stability based on public disclosure.

Applied Digital's customer and partnership disclosures reveal significant opacity regarding revenue-generating relationships. Of the announced customer agreements, only Together AI is explicitly named (claims 3234, 3243), identified as a Cloud Services customer. The company discloses two Cloud Services customers generating approximately 17% of fiscal 2024 revenue (claim 5903), yet neither second customer is identified by name. Similarly, the company reports serving seven crypto-mining customers under multi-year contracts (claims 5869, 5880), none of which are named, despite this segment generating the majority of revenue. This concentration—approximately 87% of disclosed revenue from nine unnamed customers—presents substantial counterparty disclosure gaps.

The much-publicized 400 MW hyperscaler arrangement is consistently characterized as a 'letter of intent' with exclusivity (claims 3233, 5861, 5906), not a signed customer agreement. This structural distinction is material: LOI language indicates non-binding intent, distinctly different from the signed contracts claimed for crypto customers or the 36-month Cloud Services contracts referenced for unnamed parties (claims 6129, 5983).

So-called partnerships with Super Micro (May 2023, claims 5877, 5883), HPE (June 2023, claims 5876, 5901), and Dell (claims 5865, 5902) appear to be supply or infrastructure arrangements rather than customer relationships generating direct revenue. Super Micro and HPE disclosures use identical language across multiple filings (repetition across claims 5877–5900 range), suggesting SEC filing templating. Dell is explicitly characterized as a 'supply agreement' for servers, not a customer engagement.

Counterparty concentration has evolved modestly: crypto customers increased from four (claim 6128, May 2024) to seven (claim 5869, August 2024), suggesting customer wins but without naming. Together AI emerged as the sole newly named Cloud Services customer in 2024 (claim 3243), yet two Cloud Services customers were referenced as early as June 2023 (claim 6095). The lack of revenue attribution to specific named counterparties prevents reconciliation of deal announcements to financial performance. The repetitive verbatim language across multiple claims and filing dates underscores the limited genuinely new customer disclosures, masking potential stagnation in diversification despite headline announcements.

Applied Digital's customer roster presents a bifurcated disclosure profile: named customers versus unnamed counterparties, and clear commercial contracts versus forward-looking letters of intent. Together AI is the only explicitly named cloud services customer (claims 3234, 3243), while a second cloud services customer entered into a 36-month contract in June 2023 (claims 6096, 5983) but remains unidentified. The crypto-mining segment, by contrast, is disclosed as a count rather than roster—four customers were disclosed in May 2024 (claim 6128), escalating to seven by August 2024 (claims 5869, 5879), all with signed three-to-five-year hosting contracts. This evolution from four to seven crypto customers is material, yet the lack of named counterparties prevents assessment of customer concentration risk or whether these represent true diversification or portfolio consolidation.

The partnership layer introduces further ambiguity. Super Micro and HPE are characterized as "partnerships" (claims 6127, 6090, 5864) to deliver cloud services, yet no named end customers are affiliated with these arrangements. These read as equipment supply and go-to-market partnerships rather than revenue-generating customer agreements. Dell is identified as a "supply agreement" counterparty (claims 6091, 5991) for servers, not a customer. GMR is explicitly a "strategic equity investor" (claim 5898), not a customer. The 400 MW capacity lease with an unnamed US-based hyperscaler is disclosed only as a letter of intent (claims 3233, 3246, 5906), indicating non-binding optionality, not a committed revenue arrangement.

Revenue reconciliation shows concentration risk. The Cloud Services Business "accounted for approximately 17% of our revenue in fiscal year 2024" (claim 5903) while serving two customers. Crypto mining therefore represents approximately 83 percent of revenue across seven disclosed customers. This revenue-to-customer ratio skews heavily toward crypto, a cyclical, volatile segment. The arithmetic is internally consistent but highlights dependency: material revenue concentration in an unidentified second cloud services customer and in a crypto mining cohort lacking individual counterparty visibility. The undisclosed identities of six of seven crypto customers and one of two cloud services customers materially constrain assessment of customer diversification, contract stability, or customer-level default risk.

Applied Digital presents a marked contrast between disclosed customer relationships and actual contractual commitment levels. While the company enthusiastically announces Together AI as its 'newest Cloud Services customer' (claims 3234, 3243), this single named counterparty obscures material transparency gaps elsewhere. A 'second customer' in Cloud Services under a purported 36-month contract executed June 2023 (claims 6129, 6096, 5903) has never been identified by name across any filing analyzed. The company similarly discloses serving seven crypto-mining customers with contracts spanning three to five years (claims 5897, 5879, 5869)—representing the revenue majority—without naming any operator, and this count itself escalated from four in May 2024 (claim 6128) to seven by August 2024, suggesting undisclosed account acquisition or retroactive recounting. The hyperscaler relationship warrants scrutiny. Claims 3233, 3246, 5890, 5872, and 5906 describe a 400 MW capacity lease explicitly as a 'letter of intent' with an unnamed counterparty, noting 'exclusivity' but no signed binding agreement as of August 2024; this remains optionality, not confirmed revenue. Infrastructure 'partnerships' with Super Micro (May 2023, claims 6126, 6089, 5989, 5883, 5877, 5863, 5900) and HPE (June 2023, claims 6127, 6090, 5991, 5990, 5884, 5901, 5876, 5864) employ intent-signaling language—'Together, we aim to deliver'—lacking contractual clarity; these appear to be supplier or go-to-market arrangements, not customer acquisitions. The Electric Service Agreement with an unnamed upper Midwest utility (September 2023, claims 6092, 5885, 5868, 5899, 5880) and Dell supply agreements (claims 6091, 5991, 5902, 5878, 5865) represent true signed contracts but generate no direct customer revenue. The Cloud Services segment—repeatedly positioned as strategic growth—accounts for only 17% of FY2024 revenue despite comprising two customers (claim 5903), while unnamed crypto miners drive the majority. Counterparty concentration remains extreme and largely opaque, fundamentally undermining visibility into customer quality, churn risk, and revenue sustainability.

Applied Digital's customer and partnership disclosures reveal significant gaps between announced deals and executed agreements. Together AI stands as the only specifically named Cloud Services customer (claims 3234, 3243), while the company references a 'first major AI customer' (claims 6095, 5983) and 'second customer' (claims 6129, 6096) in Cloud Services without identification. These two unnamed customers generated approximately 17% of fiscal 2024 revenue (claim 5903), yet their identities remain undisclosed. The crypto-mining customer base, by contrast, is disclosed numerically—growing from four customers (claim 6128) to seven (claims 5897, 5879)—with all holding signed multi-year contracts ranging three to five years.

The 400 MW hyperscaler capacity commitment, repeatedly cited across multiple filings (claims 3233, 3246, 5890, 5872, 5906, 5861), carries critical structural ambiguity. Applied Digital explicitly describes this as a letter of intent with a US-based hyperscaler under exclusivity, not a signed contract. This distinction matters: the deal is conditional and generates no current revenue, yet occupies prominent position in investor communications.

Infrastructure partnerships with Super Micro, HPE, and Dell (claims 6126-6127, 6089-6090, 5863-5902) represent supplier or technology relationships, not customer revenue agreements, though their naming might suggest strategic customer wins. These are delivery mechanisms, not counterparty commitments to purchase services.

Counterparty concentration has shifted measurably: the crypto-mining segment dominates revenue despite serving seven named customers, while Cloud Services—positioned as the strategic growth narrative—relies on two customers, only one publicly identified. The Electric Service Agreement with an unnamed upper Midwest utility (claims 6092, 5885, 5868) provides cost certainty but remains anonymous.

The disclosed revenue arithmetic reconciles poorly with announcement volume. Together AI is the sole named Cloud Services customer; two others remain unnamed despite generating 17% of annual revenue. The 400 MW LOI contributes zero current revenue despite repetitive disclosure. The company's true revenue concentration lies with seven crypto-mining customers operating under signed contracts—a mature, undercapitalized segment by positioning standards—not with the forward-looking AI and hyperscaler narratives emphasized in public filings.

Applied Digital's public disclosures distinguish named customers from unnamed counterparties in ways that obscure revenue concentration and customer identity. In Cloud Services, Together AI is the sole named customer (claims 3234, 3243), yet the business formally reports serving "two customers" that contributed approximately 17 percent of fiscal-year-2024 revenue (claim 5903). The second customer, entered as a 36-month contract in June 2023 (claims 6096, 6129), remains unidentified. Earlier claims reference a "first major AI customer" secured in May 2023 (claim 6095), introducing ambiguity about whether this represents a third customer, an antecedent arrangement, or a reporting inconsistency.

Crypto-mining operations, which account for the residual 83 percent of revenue, exhibit similar anonymity. The company reports serving seven crypto-mining customers with three-to-five-year contracts (claims 5897, 5879, 5869), up from four cited earlier (claim 6128). No individual customer names are disclosed. This expansion in customer count without revenue attribution—and without clarity on whether the increase reflects market diversification or concentration among a smaller set of dominant operators—prevents independent assessment of concentration risk or customer quality.

Infrastructure partnerships with Super Micro, HPE, and Dell (claims 6126, 6127, 6091, reiterated through August 2024) are positioned as supply and delivery enablers, not direct revenue customers. The flagship 400 MW hyperscaler capacity announcement, reiterated across claims 3233, 3246, 5890, 5906, 5872, and 5861, remains a letter of intent under exclusivity since April 2024—materially different from a signed contract. The energy cost agreement entered in September 2023 (claims 6092, 5885, 5868, 5899) likewise names no counterparty, identifying only an "upper Midwest utility."

Applied Digital's disclosed customer base comprises one named Cloud Services customer (Together AI), one unnamed Cloud Services customer, seven unnamed crypto-mining customers, and an unsigned 400 MW LOI. Revenue concentration—with crypto representing approximately 83 percent of total revenue and no disclosed per-customer contribution—rests entirely on unnamed counterparties. This structure prioritizes announcement volume over counterparty transparency, limiting external validation of revenue claims and customer-concentration metrics.

Applied Digital's disclosed customer relationships exhibit substantial opacity regarding counterparty identification and binding commitment status. Among the company's claims, only Together AI is explicitly named as a signed Cloud Services customer (claims 3234, 3243). The critical distinction between signed and non-binding arrangements reveals significant gaps: the company references a second Cloud Services customer under a 36-month contract executed in June 2023 (claims 6129, 6096, 5983), yet this counterparty remains unidentified. Similarly, Applied Digital claims to serve seven crypto-mining customers under multi-year contracts (3–5 years) as of August 2024, escalated from four customers cited in May 2024 (claims 6128 versus 5880, 5897), but none are named. The 400 MW capacity lease with an unidentified US-based hyperscaler is explicitly described as a letter of intent (LOI) with exclusivity (claims 3233, 3246, 5890), indicating non-binding intent rather than a signed customer contract and thus should not be counted as revenue-generating. Counterparty concentration has shifted markedly: early filings emphasize the unnamed crypto segment, while August 2024 disclosures introduce Together AI and a second unnamed Cloud Services customer, suggesting a strategic pivot toward higher-margin AI workloads. Partnerships with HPE, Super Micro, and Dell (claims 6127, 6126, 6091) are equipment or service supply arrangements, not customer contracts generating direct revenue. Revenue arithmetic provided in claim 5903 states that the Cloud Services Business 'currently serves two customers and accounted for approximately 17% of our revenue in fiscal year 2024,' which broadly aligns with crypto customers dominating the revenue base, though the unnamed status of both customer segments prevents independent verification of deal quality or concentration risk. The unidentified hyperscaler LOI, despite its prominence in capital-deployment narratives, represents optionality rather than confirmed revenue.

Applied Digital's customer and partnership disclosures reveal a stark distinction between executed agreements and uncommitted frameworks, with material revenue concentration among unnamed counterparties. The company has named only one cloud-services customer—Together AI (claims 3234, 3243)—while consistently referencing a second cloud-services customer that entered a 36-month contract in June 2023 (claims 6096, 6129) without public identification. More significantly, the company discloses that seven crypto-mining customers account for the majority of its revenue, all under three-to-five-year contracts (claims 5869, 5879), yet none are named individually. This unnamed concentration conflicts with the narrative of diversification: claim 5903 specifies that the Cloud Services Business—which includes Together AI and the unidentified second customer—represented only approximately 17 percent of fiscal 2024 revenue, meaning crypto mining dominates at roughly 83 percent. The revenue arithmetic thus reconciles with announced deals, but the structure underscores operational risk. Three announced partnerships—Super Micro (claims 5989, 5900), HPE (claims 5901, 5864), and Dell (claims 5902, 5865)—function as supply or technology arrangements rather than customer revenue contracts; they do not appear in the revenue base. Most notably, the 400 MW hyperscaler capacity lease has remained at letter-of-intent stage since April 2024 (claims 3233, 3246, 5861, 5906, 5890, 5872), despite repeated citation in quarterly filings. The hyperscaler itself remains unnamed and the LOI does not constitute a binding commitment. Across 16 months of disclosure (April 2024 through August 2024), Applied Digital has named two customer counterparties (Together AI and an unnamed utility for power supply), while describing the majority of revenue generation through unnamed crypto-mining operators and an uncommitted 400 MW framework. This disclosure pattern obscures true customer concentration and the binding status of purported growth drivers.

Applied Digital's disclosed customer base exhibits material distinction between signed revenue contracts and preliminary frameworks. The company has named precisely two paying customers with explicit contractual commitments: Together AI (claims 3234, 3243), welcomed mid-2024 as the Cloud Services segment's third customer, and GMR Limited/SparkPool (claim 6088), bound to a March 2021 advisory services agreement rather than capacity lease. The substantive revenue engine—seven unnamed crypto mining customers under three-to-five-year contracts (claims 5879, 5897, 5869)—remains undisclosed by counterparty identity. Claims 6128 and 6097 reference four such customers in earlier disclosures, suggesting organic customer addition by August 2024.

The purported 400 MW hyperscaler commitment requires careful characterization. Claims 3233, 3246, 5890, 5861, and 5872 consistently frame this as an executed letter of intent with exclusivity, not a signed master service agreement. An LOI is a non-binding or conditionally binding preliminary instrument, absent the enforceability of executed contracts. The unnamed US-based hyperscaler counterparty remains unidentified across all citations, representing concentration risk against a critical revenue prospect.

Supplier partnerships with HPE (June 2023), Super Micro (May 2023), and Dell function as hardware procurement channels rather than customer revenue streams. These recur throughout the record as partnership announcements but lack customer revenue recognition language. Similarly, claim 5898 recasts GMR as a "strategic equity investor" rather than a service customer, introducing potential capital structure opacity.

Revenue arithmetic creates tension with counterparty disclosure. Claim 5903 specifies the Cloud Services Business served two customers generating approximately 17 percent of fiscal 2024 revenue. The unnamed crypto mining segment accounts for the "majority of revenue" per claim 5880. If two Cloud Services customers represent 17 percent of total revenue, the unnamed seven crypto mining customers must collectively generate substantially more, yet remain completely undisclosed. This asymmetry—named customers in early-stage Cloud Services, wholly unnamed customers dominating revenue—suggests material customer concentration in counterparties the company has declined to identify. Concentration among seven mining operators with multi-year contracts inherently carries execution and counterparty concentration risk that public disclosure language has minimized.

Applied Digital's announced customer base presents a stark contrast between named counterparties and unnamed bulk categories that obscure meaningful commercial risk. Only Together AI (claims 3234, 3243) is identified by name as a signed Cloud Services customer. The company claims a second Cloud Services customer under a 36-month contract (claims 6096, 6129, 5983), but provides no identification. More problematically, seven unnamed crypto-mining customers (claims 5879, 5880, 5897, 5869) reportedly account for "the majority of the revenue" under 3–5 year contracts, yet no counterparty names appear in any disclosure. This concentration in unnamed hosting customers creates material opacity around customer quality, tenure duration, and concentration risk.

The 400 MW hyperscaler capacity arrangement is characterized consistently as a letter of intent with exclusivity (claims 3233, 3246, 5890, 5872, 5861), not a signed customer commitment. The unnamed "US-based hyperscaler" remains unidentified, and LOI language falls well short of binding customer agreements. The company's disclosure of this transaction alongside actual customer wins conflates optionality with booked revenue, a distinction material to investors.

Supply and technology partnerships with Super Micro (claims 6126, 6089, 5900, 5877, 5883), HPE (claims 6127, 6090, 5901, 5876, 5884), and Dell (claims 6091, 5991, 5902) are supplier relationships, not customer agreements. GMR, mentioned in a 2021 services agreement (claim 6088), is later described as a "strategic equity investor" (claim 5898) rather than an operating customer. An unnamed upper Midwest utility holds an Amended and Restated Electric Service Agreement (claims 6092, 5899, 5868, 5885) that provides cost certainty, not revenue.

Revenue reconciliation shows Cloud Services at approximately 17% of FY2024 revenue (claim 5903) across two customers, while crypto hosting "accounts for the majority" (claim 5880). These disclosures create arithmetic tension: if crypto dominance exceeds 50% and cloud represents 17%, unaccounted revenue streams warrant explanation. The crypto customer base expanded from four to seven (comparing claims 6128 versus 5879), yet anonymity prevents validating churn, quality shifts, or concentration concentration among these accounts. Together AI represents the sole named customer win disclosed publicly, limiting ability to assess sales execution or pipeline quality beyond a single reference.

Applied Digital discloses two Cloud Services customers representing approximately 17 percent of fiscal 2024 revenue (claim 5903), yet only one—Together AI—is publicly named (claims 3234, 3243). The second customer remains undisclosed. Separately, the company reports serving seven crypto-mining customers under multi-year contracts (claims 5879, 5897, 5869) with no counterparties identified by name. This represents material growth from the four crypto-mining customers cited earlier (claim 6128), indicating customer base expansion, but anonymity obscures concentration risk and whether any single counterparty triggers SEC materiality disclosure requirements.

The most substantial announced engagement is a 400 MW capacity lease framed as an "executed letter of intent" (LOI) with "a US-based hyperscaler" (claims 3246, 5890, 5906). Applied Digital positions this as signed exclusivity, yet LOI language indicates pre-contract status, not a binding customer agreement. No revenue should be attributed to this arrangement.

Partnerships announced with Super Micro (May 2023), Hewlett Packard Enterprise (June 2023), and Dell are technology and supply relationships, not customer agreements (claims 5877, 5876, 5865, 5878). GMR, described as a "strategic equity investor" (claim 5898), executed a strategy and advisory services agreement in 2021 (claim 6088) rather than a hosting customer contract. These distinctions separate revenue-generating customer relationships from supplier and framework partnerships.

Revenue arithmetic reconciles with disclosed customer counts: two Cloud Services customers at 17 percent of revenue, seven unnamed crypto customers driving the majority. However, asymmetric disclosure patterns limit diligence visibility. A single unnamed crypto counterparty likely representing 10–25 percent of total revenue would typically require naming under SEC materiality thresholds; the absence suggests either subthreshold individual exposure or customer-confidentiality arrangements.

Counterparty concentration has evolved marginally—crypto mining grew from four to seven customers—but the lack of named relationships across both segments creates opacity. The hyperscaler LOI remains aspirational. Transparency gaps prevent assessing customer diversification, churn velocity, and whether the customer base reflects durable demand or transient capacity arbitrage. Only Together AI's identification provides a concrete reference point for customer quality or strategic value.

Applied Digital's customer portfolio disclosure distinguishes sharply between executed customer contracts and partnership frameworks, though the company's public filings blur this distinction through repetitive claims. The core customer base rests on two pillars: seven crypto-mining customers with signed three-to-five-year hosting contracts (claims 5879, 5897, 6128), and two Cloud Services customers (claim 5903). Only one Cloud Services customer is named—Together AI (claims 3234, 3243)—while the second customer, secured in June 2023 under a 36-month contract (claims 6129, 6096), remains unidentified. This opacity creates asymmetric disclosure: Together AI is explicitly welcomed as the "newest" customer (claim 3243), yet contemporaneous filings reference an unnamed prior customer of equal contractual weight, representing half of Cloud Services' disclosed revenue contribution.

The hyperscaler capacity lease—repeatedly cited through August 2024 (claims 5890, 5872, 5861)—is a letter of intent for 400 MW, not a signed customer agreement (claim 3233). This distinction is material: LOI signals intent under exclusivity but carries no binding revenue obligation. Applied Digital conflates partnerships with customer relationships. Super Micro (May 2023) and HPE (June 2023) are technology and supply partnerships, not end-customer arrangements (claims 5876, 5884). Dell supply agreements (claims 5865, 5878) serve infrastructure provisioning, not revenue generation from named customers. GMR appears characterized simultaneously as a services counterparty (claim 6088) and later as a "strategic equity investor" (claim 5898), muddying its commercial function.

Counterparty concentration has shifted modestly. Crypto customers increased from four (May 2024, claim 6128) to seven (August 2024, claim 5880) over four months—a 75% expansion without naming a single customer. The Cloud Services segment, representing 17% of FY2024 revenue (claim 5903), relies entirely on two customers, compressing concentration risk substantially.

Revenue reconciliation aligns with disclosed customer counts: crypto mining represents the majority (implicitly 70–80%), Cloud Services contributes ~17%, leaving ~3–13% from other segments. However, the unnamed Cloud Services customer represents half of this segment's revenue, creating material undisclosed concentration. The unexecuted hyperscaler LOI introduces forward aspiration but zero current contribution. Until the hyperscaler LOI converts to a definitive customer agreement and the second Cloud Services customer achieves public identification, Applied Digital's revenue base remains underdiversified in contractual visibility.

Applied Digital's disclosed customer base presents a significant contrast between binding agreements and contingent commitments. Of the company's announced Cloud Services customers, only Together AI (claims 3234, 3243) is specifically named; a 'first major AI customer' from May 2023 and a second Cloud Services customer under a 36-month contract (claims 6095, 6129) remain publicly unidentified. The crypto mining segment, which the company describes as generating the majority of revenue, comprises seven customers (claim 5869) with 3–5 year hosting contracts, but these counterparties are never disclosed by name. The most prominently announced capacity commitment—a 400 MW lease with an unnamed 'US-based hyperscaler'—is explicitly a letter of intent rather than an executed agreement (claims 3233, 3246, 5872, 5890). This LOI structure indicates exclusivity and intent but not binding commitment, and has been repeated verbatim across filings from April through August 2024 without progression to signed contract status.

The company's partnerships with Super Micro (May 2023, claims 5989, 5877) and Hewlett Packard Enterprise (June 2023, claims 5990, 5901) are framed as supplier relationships for delivering cloud services infrastructure, not customer contracts. Dell represents a supply agreement for hardware (claims 6091, 5902), similarly non-revenue-generating in the direct sense. An undisclosed utility in the upper Midwest (claim 6092) entered an Electric Service Agreement in September 2023, which establishes cost structure but is operational, not a customer contract. Concentration risk is pronounced: the Cloud Services Business, accounting for approximately 17% of fiscal 2024 revenue (claim 5903), relies on two named-or-unnamed customers; crypto mining's majority-revenue contribution depends on seven unnamed hosting customers. The arithmetic disclosed—17% Cloud Services revenue from two customers versus unnamed crypto revenue from seven customers—raises questions about customer concentration and sustainability, particularly given Together AI's recent entry and the absence of quantified revenue contribution per counterparty. The persistent republication of May–June 2023 partnerships without milestone updates suggests limited subsequent deal progression, a pattern particularly evident with the LOI, which has remained unsigned for over 16 months.

Applied Digital has disclosed only one named customer in its Cloud Services business: Together AI, welcomed in April 2024 (claims 3234, 3243). A second Cloud Services customer remains unnamed despite a disclosed 36-month contract executed in June 2023 (claims 6096, 5983). The crypto-mining segment—explicitly generating the majority of company revenue (claim 5880)—comprises seven entirely undisclosed customers under three-to-five-year contracts (claims 5879, 5880). This base expanded from four to seven between May and August 2024, yet no counterparty names have been released.

Partnerships announced in May and June 2023 require reclassification: Super Micro (claims 6089, 5863) and HPE (claims 6090, 5876) are infrastructure suppliers enabling Applied Digital to serve customers, not revenue-generating counterparties. Dell supply agreements (claims 6091, 5878) constitute procurement inputs, not customer contracts. The 400 MW hyperscaler relationship remains non-binding: only a letter of intent under exclusivity has been executed (claims 3246, 5861, 5872), not a signed capacity lease, rendering this pipeline entirely aspirational.

Counterparty concentration reveals substantial anonymity. Of two Cloud Services customers, only Together AI is named; the second remains entirely undisclosed despite material commitment. Crypto mining, the larger revenue segment, is wholly unnamed across seven entities. Between May and August 2024, Applied Digital added three new unnamed crypto miners, intensifying customer opacity rather than disclosing concentration metrics.

Revenue reconciliation is incomplete. Applied Digital discloses that its two Cloud Services customers generated approximately 17 percent of fiscal 2024 revenue (claim 5903), implying material contribution from the unnamed second customer. However, the company discloses neither customer concentration percentages for crypto mining—the majority-revenue segment—nor revenue thresholds for the unnamed second Cloud Services counterparty. The 400 MW LOI introduces further uncertainty: no signed lease, financial terms, ramp timeline, or revenue projection have been disclosed. Without top-customer revenue disclosure or identification of the unnamed second Cloud Services customer, the arithmetic cannot be independently verified against the announced pipeline.

Applied Digital's customer disclosures distinguish rigidly between identified accounts and anonymous counterparties, with material concentration risk inadequately surfaced. Together AI is the sole named Cloud Services customer (claims 3243, 3234). For the second Cloud Services account, APLD refers only to an undisclosed customer under a 36-month contract (claims 5983, 5903, 6129, 6095), acknowledged to represent approximately 17% of fiscal 2024 revenue—extraordinary yield from two accounts, only one publicly identified. The crypto-mining segment, disclosed as the company's revenue majority (claim 5880), comprises seven contracted counterparties (claims 5869, 5879, 5880), though prior disclosures cited four (claims 6128, 6097). This mid-year expansion from four to seven customers between May and August 2024 signals either material customer acquisition or significant upward revision of the deployed footprint, neither adequately explained in subsequent communications.

APLD's infrastructure partnerships—Super Micro, HPE, Dell—are supplier or systems-integration arrangements, not revenue-generating end-customer contracts (claims 5863, 5864, 5902), yet investor materials conflate them with customer wins. The 400 MW hyperscaler lease remains an executed letter of intent (LOI), explicitly not a binding customer agreement (claims 5861, 5872, 5906), representing contingent future revenue dependent on completion.

Revenue reconciliation fails on two axes. First, if Cloud Services represents 17% of fiscal 2024 revenue and crypto mining comprises the majority, two named/unnamed Cloud Services customers and seven crypto accounts generate extreme concentration per counterparty. No revenue per customer, average contract value, or annualized customer contribution is disclosed. Second, the crypto segment's volatility—four to seven customers in three months—paired with anonymous identification prevents assessment of portfolio durability or validation that announced multi-year terms align with actual financial contribution to reported metrics.

APLD's customer architecture is materially opaque. Together AI is the singular named customer win in a disclosed portfolio where the majority of revenue derives from anonymous crypto counterparties and a second unidentified Cloud Services customer. This opacity obscures whether revenue concentrates in distressed accounts or reflects systemic customer acquisition or retention stress.

Applied Digital discloses meaningful customer concentration with limited transparency. The company names one Cloud Services customer—Together AI, signed in 2024 (claims 3234, 3243)—but maintains anonymity for the second Cloud Services customer, entered June 2023 with a 36-month contract (claims 6129, 6096, 5983). Crypto-mining revenue, disclosed as the majority of FY2024 earnings, derives from seven unnamed customers under multi-year contracts (claims 5869, 5879, 5880, 5897), though this cohort grew from four customers cited in May 2024 (claim 6128). The company's revenue disclosures lack sufficient granularity to reconcile customer counts with earnings attributable to segments. Claim 5903 quantifies Cloud Services at approximately 17 percent of revenue for FY2024, implying crypto-mining represents approximately 83 percent, yet no customer-level economics are provided.

Applied Digital conflates supplier and equity relationships with customer relationships in its disclosure pattern. HPE (claims 6127, 6090, 5864, 5876, 5901), Super Micro (claims 6126, 6089, 5989, 5883, 5863, 5877, 5900), and Dell (claims 6091, 5991, 5865, 5902) are presented as partnerships for technology supply or services delivery; none qualify as customer contracts. GMR appears as both an equity investor and a party to an advisory services agreement from March 2021 (claim 6088), later re-disclosed as a strategic equity investor (claim 5898), not a revenue-generating customer relationship.

Most critically, Applied Digital's largest disclosed partnership—a 400 MW capacity lease with an unnamed US hyperscaler—remains at letter-of-intent stage (claims 3233, 3246, 5861, 5872, 5906). The LOI represents non-binding exclusivity, not an executed contract, yet it receives prominence in customer and partnership summaries. The hyperscaler's identity remains anonymous, preventing third-party assessment of creditworthiness or deal probability.

The disclosure pattern reveals material counterparty concentration risk masked by anonymity. Together AI is the sole named customer; the remaining customer base—representing the bulk of revenue—remains unidentified. Absent customer concentration disclosures, revenue contingency breakdowns, or contract renewal dates for the multi-year mining cohort, investors cannot assess customer retention risk or revenue sustainability.

Applied Digital's disclosure record reveals a sharp distinction between signed customer relationships and partnership positioning. Only Together AI is explicitly named as a Cloud Services customer (claims 3234, 3243), while the company references two cloud services customers generating approximately 17% of fiscal 2024 revenue (claim 5903). The second customer remains unnamed despite repeated disclosure references (claims 6129, 5983, 5903), a material gap in transparency. The 400 MW capacity commitment, cited repeatedly (claims 3233, 3246, 5890), is clearly marked as a letter of intent with exclusivity rather than a binding customer agreement—no named counterparty is disclosed. Crypto-mining revenue, by contrast, appears to constitute the plurality of sales, with seven named contract holders across three-to-five-year terms (claims 5897, 5880, 5879). However, these customers remain unidentified by name. The counterparty concentration has remained static: the company announced HPE (June 2023) and Super Micro (May 2023) partnerships repeatedly across claims 6127, 6090, 5901, 5876, and others, yet these are supply/technology partnerships, not revenue customers. Dell's presence (claims 6091, 5902, 5878) is similarly a supplier relationship for hardware delivery. A services agreement with GMR (claim 6088, dating March 2021) involved portfolio advisory work, not cloud capacity sales. The arithmetic appears internally consistent: two cloud services customers (one named) generating 17% of revenue, with crypto-hosting operations as the core business driver. Yet the disclosure pattern conflates partnerships, supplier relationships, and actual paying customers—applying partnership language to what are effectively vendor arrangements with HPE and Super Micro. The absence of customer names beyond Together AI, combined with the static customer count reported across eighteen months of filings, suggests limited expansion in the high-margin Cloud Services segment despite the repeated emphasis on partnership announcements.

Applied Digital's customer disclosures reveal significant gaps between signed agreements and contingent commitments, with heavy reliance on unnamed counterparties. The company names only one Cloud Services customer—Together AI (claims 3234, 3243)—as a signed customer, though it claims to serve two Cloud Services customers total (claim 5903), leaving the second customer and its revenue contribution unidentified. The anonymous first "major AI customer" in Cloud Services (claim 6095) represents undisclosed counterparty risk. Seven crypto mining customers generate the majority of revenue but remain completely unnamed across all disclosures (claims 5869, 5879, 5880, 5897, 5906), creating opaque concentration exposure despite multi-year contract claims. In contrast, the headline 400 MW hyperscaler deal is explicitly a letter of intent, not a signed commitment (claims 3233, 3246, 5861, 5906), yet appears in earnings narratives alongside signed customer wins. The distinction matters: Applied Digital has entered exclusivity and executed an LOI but has not yet closed a binding lease. Similarly, partnerships with HPE (claims 6127, 5990, 5884, 5901) and Super Micro (claims 6126, 6089, 5989, 5877, 5900) are framed as active arrangements, but the verbatim language—"aim to deliver"—suggests framework or pre-contractual status rather than binding supply or customer agreements. By contrast, Dell's GPU and AI server supply agreement (claims 6091, 5991, 5902) appears genuinely operational. The revenue picture barely reconciles: Cloud Services represents approximately 17% of FY2024 revenue with only two customers disclosed (claim 5903), while crypto mining, the majority segment, derives from seven unnamed counterparties. This creates two diligence risks: first, unnamed customers account for a material, majority revenue stream with no public visibility into concentration, customer health, or contract terms; second, the company conflates LOIs and partnerships with signed customer commitments, obscuring actual near-term revenue certainty. Material customers remain unidentified despite SEC disclosure obligations, and the LOI with the hyperscaler should be clearly labeled contingent rather than presented as a confirmed capacity acquisition. The lack of named customer disclosure outside Together AI undermines visibility into sales traction and customer diversification claims.

Applied Digital has executed signed long-term datacenter leases with four material counterparties, yet names only one publicly. CoreWeave is the sole identified named customer, contracted for 400 MW across Polaris Forge 1 under multiple signed lease agreements (claims 115, 116, 372) valued at approximately $11 billion. Oracle has been named as tenant at Polaris Forge 2 (claim 2249), though the nature of that tenancy is ambiguous—described as "currently leased to Oracle" without characterizing it as a signed commercial agreement comparable to the datacenter leases. Base Electron is named as a partnership counterparty for 1.2 GW of power generation capacity (claims 79, 105), and Applied Digital owns ~10% of the entity (claim 877), indicating equity alignment rather than arm's-length customer relationship. ChronoScale appears in a management advisory services agreement (claim 269), distinct from capacity leases. An unnamed single crypto mining customer represents the legacy data center business (claims 128, 109), with approximately 1.5 years remaining on contract.

Three additional U.S.-based investment-grade hyperscalers remain entirely anonymous despite holding signed lease agreements: 200 MW at Polaris Forge 2 (~$5 billion, claims 90, 68), 300 MW at Delta Forge 1 (claims 70, 120), and 210 MW at Delta Forge 2 (~$5.2 billion, claims 122, 1437). A fourth unnamed hyperscaler holds 300 MW at Polaris Forge 3 (~$7.5 billion, claim 337). Collectively, these four unnamed hyperscalers represent approximately 1,010 MW and roughly $23–24 billion of the $36.2 billion contracted portfolio as of May 31, 2026 (claim 123).

Concentration has paradoxically become more opaque even as the customer base diversified. Approximately 70% of contracted revenue now derives from investment-grade hyperscalers (claim 332), yet one anchor customer accounted for 59% of fiscal 2026 continuing operations revenue (claims 313, 323). The mathematics suggest the 59% anchor customer is not CoreWeave—whose $11 billion represents roughly 30% of total contract value—but rather one of the unnamed hyperscalers. The revenue-per-megawatt divergence across contracts ($25.6 million average per MW across the portfolio; CoreWeave ~$27.5 million per MW) is consistent with this interpretation. CoreWeave's public visibility thus masks the true revenue concentration behind contractual confidentiality clauses, while the disclosed $36.2 billion aggregate contract value appears internally reconcilable across named and unnamed counterparties. A Memorandum of Understanding with CoreWeave for ELN-04 assignment (claim 214) remains conditional on subsidiary investment-grade certification.

The record as supplied ends mid-claim in May 2022, so this read covers the crypto-hosting era and cannot yet assess the Polaris Forge, Delta Forge, Finland or Base Electron programs listed among the projects. The 2006 and 2008 entries (21053, 21054, 21067-21069) belong to the predecessor shell entity and say nothing about datacenter delivery. They should be excluded from any trajectory scoring.

Announced capacity inflated quickly. In August 2021 the company described a single 100 MW energy services agreement (14511), with 500 MW framed only as demand (14512). By March 2022 it set an 800 MW target for May 2023 (14228), 1.8 GW for May 2024 (14229, 14032, 24804), and 5 GW over five years (14230, 24805). Alongside these sat a 1.5 GW Antpool joint venture (14203, 14235) and a 200 MW Texas wind LOI (14236). Within about seven months the headline grew roughly fiftyfold on a base of one operating site.

Dates for the concrete Jamestown work held. First-facility completion by end of 2021 (14509, 14414) is verified. So is the remaining 45 MW for Q2 2022 (14227, 14069, 14009), which followed 55 MW online by February 2022 (14008). The portfolio-level dates are a different story. The 800 MW and 1.8 GW deadlines passed long ago yet remain tagged not_yet_verifiable. The three-site, 600 MW year-end 2022 promise (24808) is in the same state. The Bitmain 200 PH order (14504, 14416, 14378) was repeated four times without any delivery confirmation. Nothing in the record shows these targets being restated or withdrawn. That is the pattern of revision without acknowledgement. The Antpool vehicle also drops out of view after April 2022, and the Garden City LOI ends in a sale rather than a build.

The record contains no independent evidence such as permits, interconnection queue entries or local reporting. The only external validation is the verified status on the Jamestown milestones. The early record therefore supports a small, real operator wrapped in portfolio claims that went unverified.

Over the next 90 days, the priority is ingesting the post-May 2022 claims so the AI-campus commitments can be scored against this baseline. After that, three checks matter. First, whether MISO or SPP queue positions and North Dakota PSC filings exist for Harwood, Oliver County and the Base Electron gas plant at the capacities claimed. Second, whether Louisiana and Alabama permitting supports the Delta Forge construction status. Third, whether the expired 2023 and 2024 targets are ever formally reconciled in filings.

The capacity record for Applied Digital runs from its bitcoin-hosting origins through May 2022 and stops there. Nothing in this dataset covers Polaris Forge, Delta Forge, Finland or the Base Electron gas plant, so any read of the current AI-campus pipeline has to begin by noting that gap. The 2006 and 2008 entries (21053, 21054, 21067-21069) belong to the predecessor shell and say nothing about the datacenter business.

Within that window, ambition grew quickly. In August 2021 the company described a single 100 MW energy services agreement and 500 MW of expansion demand (14511, 14512). Within months it added an 85 MW second facility (14375, 14362). By March 2022 it had jumped to 800 MW by May 2023, 1.8 GW by May 2024 and 5 GW over five years (14228, 14229, 14230), and the 1.21 Gigawatts joint venture with Antpool promised up to 1.5 GW (14235, 14212). Those targets were repeated almost word for word across the April and May 2022 filings (14031, 14032, 14033, 21463, 21464). That is restatement, not progress.

The near-term promises held. The Jamestown co-hosting facility was due by end-2021 and is marked verified (14509, 14372, 14414). The remaining 45 MW there, due by Q2 2022, is also verified (14227, 14009, 14030). The multi-year targets are a different matter. Both the 800 MW and 1.8 GW deadlines have passed, and both remain not_yet_verifiable in this record. There is no later claim acknowledging, revising or abandoning them. The Garden City wind site was carried for months as a 200 MW letter of intent (14204, 14074, 14034) and now shows as sold. That is consistent with a strategy pivot, but the record contains no claim explaining it.

The only independent corroboration on file is the verification status of the Jamestown milestones. No permits, interconnection-queue positions or site reporting are attached to any gigawatt-scale target, so the announced pace is unsupported by outside evidence in this dataset.

Over the next 90 days, the first priority is to ingest post-2022 claims for the Polaris Forge and Delta Forge campuses. Without them, the projects listed as under construction cannot be tested against what was promised for them. The fiscal second-quarter filing expected around January should show whether the 5 GW by March 2027 target (14230) is still stated, quietly retired or replaced by lease-backed figures. Interconnection filings with MISO for the North Dakota sites, and Louisiana and Alabama utility dockets for the Delta Forge sites, would be the first independent check on whether construction pace matches the narrative.

The claim record for Applied Digital has two problems that limit any read of Polaris Forge or Delta Forge. It stops at a partial May 2022 entry (claim 26007). It also contains 2006 and 2008 items (claims 21053, 21054, 21067-21069) that look like they belong to a predecessor shell entity rather than the datacenter business, and they should be set aside. So the trajectory below is really the bitcoin-hosting era. The current AI-campus projects on the project list have no capacity or timeline claims attached yet.

Within that window, announced ambition grew much faster than anything built. In August 2021 the company described a 100 MW energy services agreement and 500 MW of expansion demand (claims 14511, 14512). By March 2022 the targets had jumped to 800 MW by May 2023, 1.8 GW by May 2024 and 5 GW over five years (claims 14228, 14229, 14230). An Antpool joint venture was sized at up to 1.5 GW (claims 14212, 14036), and 600 MW or more of additional sites was promised before the end of 2022 (claim 19161). What was actually delivered was far smaller. The Jamestown first facility met its Q4 2021 completion (claims 14364, 14372). It then went from 55 MW online (claim 14008) to the full 100 MW, with the remaining 45 MW verified against the Q2 2022 target (claims 14009, 14055, 14227). The 85 MW second facility was repeated across filings (claims 14362, 14040) without a date ever being attached.

The most notable pattern is not a slip but a silence. The May 2023 and May 2024 deadlines have passed and remain unverifiable. The record holds no restatement, withdrawal or acknowledgement of either target. The Garden City wind facility followed a similar arc: a 200 MW letter of intent (claims 14204, 14236), then a September 2022 groundbreaking promise (claim 21466) that was never confirmed, and the project now carries a sold status. The 200 PH Bitmain order (claims 14504, 14416) also has no delivery evidence. No permit, interconnection-queue or on-the-ground evidence appears in the record for any site, so the only independently corroborated pace is the original Jamestown buildout.

Over the next 90 days, the priority is closing the 2022-to-present gap with claims on Polaris Forge 1 through 3 and the Delta Forge sites. On the independent side, the useful checks are interconnection-queue positions for Harwood and Oliver County, and North Dakota Public Service Commission filings for the Base Electron gas plant. It will also be worth noting whether management ever reconciles the 2023 and 2024 gigawatt targets with what was built.

The record supplied here ends in May 2022, partway through claim 19159, so this read covers Applied Digital's crypto-hosting origins rather than its current Polaris Forge and Delta Forge buildout. Nothing in the record carries a capacity, cost or date for those projects. The two 2006 and 2008 claims (21053, 21067) appear to belong to a predecessor entity and are not relevant to this trajectory.

The early capacity claims are modest and consistent. A 100 MW energy services agreement anchors Jamestown (14511, 14373, 14013), and a second facility is pegged at 85 MW (14375, 14040). The scale then jumps in March 2022: 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), 5 GW over five years (14230, 24805), and a further 1.5 GW through the Antpool joint venture (14235, 14212). That is roughly a fiftyfold expansion in ambition within seven months, without new sites or interconnection agreements disclosed to support it.

On timing, the near-term Jamestown promises held. The Q4 2021 first-facility completion is verified (14509, 14364), 55 MW was reported live in February 2022 (14008), and the remaining 45 MW for Q2 2022 is verified (14227, 14009). The larger promises are another matter. The 800 MW and 1.8 GW deadlines have both passed but remain tagged not_yet_verifiable, and the record contains no later claim confirming, revising or retiring them. The same applies to the 600 MW of additional sites promised for end-2022 (19161) and the 200 MW Texas facility due in late 2022 (19159, 14236). Garden City is now listed as sold, with no claim marking that change. The Antpool joint venture stops appearing after April 2022 (14016, 14036). The repeated 200 PH Bitmain order (14504, 14360) was never verified.

The record contains no independent corroboration of any kind, such as permits, queue positions or site reporting. The verified tags apply only to Jamestown's first 100 MW.

Over the next 90 days, the priority is to extend the record. That means capturing APLD's Polaris Forge 2 and 3 energization dates, matching them against MISO interconnection entries, and tracking North Dakota PSC filings for the Base Electron gas plant. Permitting at Boyce and Brookwood should be checked, and any concrete specification for the Finland campus noted. The pattern to test is whether near-term, site-specific promises keep holding while headline gigawatt targets lapse without comment.

The record supplied here runs only through May 2022 and breaks off mid-claim at 3124, so this read covers Applied Digital's crypto-hosting origins, not the Polaris Forge and Delta Forge AI campuses now listed as under construction. The 2006 and 2008 entries (21053, 21054, 21067-21069) describe the predecessor shell's business and have no bearing on facility delivery.

The capacity story moves from small and contracted to large and aspirational within seven months. Through late 2021 the claims are anchored to one 100 MW energy services agreement at Jamestown (14502, 14511, 14413, 14361) and an 85 MW second site (14375, 14362). The March 2022 disclosures then add 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), 5 GW over five years (14230), a 1.5 GW joint venture with Antpool (14235, 14212), and a 200 MW Texas wind letter of intent (14236). Only the 100 MW agreement and the 85 MW reservation (14040) are tied to a counterparty power contract. The gigawatt figures rest on a JV and an LOI, and the five-year target appears with a deadline in 14230 and without one in 14033.

Near-term dates held. The Q4 2021 first-facility completion is verified across 14509, 14414, 14372 and 14364. The remaining 45 MW for Q2 2022 is verified across 14227, 14196, 14069, 14055, 14030 and 14009, after 55 MW was reported online in early February (14008). The long-range dates fared differently. The May 2023 and May 2024 deadlines have both passed, yet 14031 and 14032 still read not_yet_verifiable, and nothing in this record retires, revises or reaffirms them. The Bitmain 200 PH order (14504, 14416, 14378, 14360) is also unresolved. The Texas LOI site, Garden City, now shows as sold, and the record contains no claim that acknowledges that exit.

No permits, interconnection queue entries or on-the-ground reporting appear in the record. The verified Jamestown milestones seem to rest on company filings rather than outside confirmation. Before the gigawatt-era targets are scored, MISO queue positions for the North Dakota sites and ERCOT records for Garden City should be pulled.

Over the next 90 days, the first priority is to load the post-May 2022 claims so the Polaris Forge, Delta Forge and Finland capacity figures can be compared against this baseline. Beyond that, two questions matter most. The first is whether the next quarterly filing states energized megawatts at Polaris Forge 1 that move it from partially to fully operational. The second is whether the company ever formally reconciles the expired 800 MW and 1.8 GW targets or lets them lapse quietly.

The capacity record on file covers only through May 2022. That matters, because every project that now defines Applied Digital, from Polaris Forge 1 through 3, Delta Forge 1 and 2, the Finland campus and the Base Electron gas plant, has no capacity or timeline claim in this dataset. What follows is a read of the crypto-hosting era, not the AI-campus build. The 2006 and 2008 entries (21053, 21054, 21067-21069) describe a predecessor shell entity and say nothing about the current delivery trajectory. Within the window that is covered, the near-term promise held up better than the long-range ambition. The Jamestown 100MW buildout moved from 55MW online as of February 2022 (14008, 14054) to a remaining 45MW guided for Q2 2022 (14009, 14030, 14055), and that milestone is marked verified. The same-day filings on 2022-05-13 do not quite agree with each other. One says the full 100MW is online (26005), while the other says 20MW remains to come online (19151). That gap should be resolved before the Q2 claim is treated as cleanly met. The long-range targets tell a different story. Guidance of 800MW by May 2023 (14228, 14031), 1.8GW by May 2024 (14229, 14032) and 5GW within five years (14230, 14033, 19158) was repeated with confidence through spring 2022. Both dated deadlines have now passed and still sit as not_yet_verifiable, so they have never been adjudicated, let alone met in the record. The 1.21 Gigawatts joint venture with Antpool, sized at up to 1.5GW (14235, 14036, 14203), and the 200MW Garden City wind LOI (14236, 14204) both disappear from later claims. Garden City is now listed as sold, and the record contains no acknowledgement that the JV was abandoned. The 85MW second facility (14375, 14362, 14040) is the clearest bridge to the Ellendale and Polaris Forge site. The Bitmain 200PH order (14416, 14378) was never confirmed. Nothing in the record offers independent corroboration: no permits, no MISO or SPP interconnection queue positions, no local reporting. The announced pace therefore rests entirely on company statements. Over the next 90 days, the first priority is to backfill 10-Q and 8-K claims from mid-2022 onward so the AI-campus commitments can be scored at all. The 800MW and 1.8GW targets should then be formally adjudicated as missed or restated. Beyond that, watch for energization evidence at the remaining Polaris Forge 1 buildings, North Dakota PSC and air permits for Polaris Forge 3 and Base Electron, and Louisiana and Alabama interconnection filings for the Delta Forge sites.

The record supplied here is truncated around mid-May 2022 (it cuts off after claim 19151), so this read covers Applied Digital's crypto-hosting origins rather than the Polaris Forge, Delta Forge and Finland AI-factory era that now defines the project list. The 2006 and 2008 entries (21053, 21054, 21067-21069) belong to the predecessor shell and should be excluded from any trajectory scoring.

On capacity, the arc is one of rapid headline inflation layered over modest physical delivery. The 2021 filings anchor on a 100 MW energy services agreement at Jamestown (14413, 14373, 14361) and an 85 MW second site (14375, 14362), with a vaguer 500 MW expansion demand figure (14512). By March 2022 the ambition had escalated to 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW within five years (14230), alongside a 1.5 GW Antpool joint venture (14235, 14212) and a 200 MW Texas wind LOI (14236). The Texas site, Garden City, now carries a sold status, so that 200 MW never became durable owned capacity.

On timing, the near-term Jamestown promises held. First-facility completion in Q4 2021 is verified (14414, 14372, 14364), and the remaining 45 MW for Q2 2022 is verified across repeated restatements (14227, 14196, 14069, 14030). The sequence from 55 MW online (14008) to 20 MW remaining (19151) is internally consistent. The problem sits with the aggregate targets. The deadlines for 14228/14031 and 14229/14032 passed in 2023 and 2024, yet they remain tagged not_yet_verifiable. Within this excerpt there is no sign the company acknowledged revising them, and that gap should be closed rather than left open-ended. The same applies to the 200 PH Bitmain order and the equipment-delivery claims (14504, 14378, 14365).

On independent corroboration, nothing in the supplied record (no permits, interconnection-queue entries or site reporting) backs the gigawatt-scale pace. Only the Jamestown increments have external verification attached.

Over the next 90 days, the priority is to load the post-May-2022 claims, so that the Ellendale and Polaris Forge capacity statements can be tested against the 2022 targets. Expired targets 14228 and 14229 should be resolved to missed or superseded. North Dakota Public Service Commission and MISO filings for Polaris Forge 2 and 3 are worth tracking, along with any permitting activity for the Base Electron gas plant, since on-site generation would signal that grid interconnection is constraining the announced schedule.

Applied Digital's capacity story can be traced through this record only as far as May 2022. The current project list includes Polaris Forge 1 through 3, Delta Forge 1 and 2, a Finland campus and a dedicated gas plant, but none of those have claims on file here, so this read covers the hosting-era foundation. The 2006 and 2008 entries (21053 through 21069) appear to belong to the predecessor shell and have no bearing on facility delivery. Ambition expanded sharply within a few months. In late 2021 the company was describing a 100 MW Jamestown energy agreement and an 85 MW second site (14373, 14375, 14362). By March 2022 it had moved to 800 MW by May 2023, 1.8 GW by May 2024 and 5 GW over five years (14228, 14229, 14230), plus a 1.5 GW Antpool joint venture (14235, 14036). The near-term Jamestown promises held. The Q4 2021 first-facility completion (14364, 14414) and the Q2 2022 45 MW tranche (14227, 14009) are both marked verified. The May 2022 filings contain an internal inconsistency, however. Claim 26005 says the full 100 MW was online, while claim 19151, from the same date, still refers to 20 MW of remaining capacity. Both deadlines for the 800 MW and 1.8 GW interim targets passed without any restatement in this record, and by May 2022 only the five-year 5 GW figure was still being repeated (19158). The Antpool venture is not among the current projects. Garden City, first introduced as a 200 MW wind letter of intent (14236, 14204), is now listed as sold. Retiring targets without acknowledging them is the pattern most worth tracking. The 200 PH Bitmain equipment order (14504, 14378) also never moved beyond not-yet-verifiable. The record contains no independent evidence at all: no permits, no interconnection queue positions and no site reporting. Even the verified Jamestown milestones appear to depend on the company's own disclosures. Over the next 90 days, the main need is to fill in the missing claim history for Polaris Forge and Delta Forge, so that current capacity and energization dates can be tested against the 2022 baseline. Three items deserve particular attention: state siting and air-permit filings for Polaris Forge 3 and the adjacent gas plant in Oliver County; interconnection queue status for Harwood and the Louisiana and Alabama sites; and any lease-commencement or energization disclosures that would show whether Polaris Forge 1's partial operation is converting into full contracted load on schedule.

The record supplied here stops in May 2022, so it reads as a history of Applied Digital's crypto-hosting origins rather than of the Polaris Forge and Delta Forge buildout now on the project list. The 2006 and 2008 entries (21053, 21054, 21067-21069) appear to belong to a predecessor shell entity and say nothing about delivery capacity. Within the 2021-2022 window, announced ambition scaled far faster than physical capacity. The company began with a single 100 MW energy services agreement at Jamestown (14511, 14413, 14361) and an 85 MW second site (14375, 14040). By March 2022 it was targeting 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), 5 GW over five years (14230, 14033, 19158), and up to 1.5 GW through the Antpool joint venture (14235, 14203). The joint venture's recorded formation figure of 8,000 USD (14234) is a reminder of how lightly capitalized that headline was. Near-term dates held. The Q4 2021 first-facility completion (14509, 14364) and the Q2 2022 remaining-45 MW phase (14227, 14009, 14030) are both verified, and 26005 reports the full 100 MW online by mid-May 2022. That same filing date also carries 19151, which describes 20 MW of remaining capacity still to come, an internal inconsistency that should be reconciled before 100 MW is treated as fully energized at that point. The longer-dated targets are the problem. The May 2023 and May 2024 gigawatt milestones have passed and remain unverified, with no claim in this record acknowledging a revision. The Bitmain 200 PH delivery (14504, 14416, 14360) was never confirmed. The Garden City wind site (14236, 14204) ended up sold rather than built. The dataset contains no independent corroboration, such as utility interconnection filings, North Dakota PSC siting records or local construction reporting, so the only external check is the verification tagging, and that covers only the Jamestown dates. Over the next 90 days, the priority is closing the evidentiary gap on the current pipeline. That means megawatt figures and energization dates attached to Polaris Forge 1's partial operation, MISO or SPP interconnection entries and county permits for Harwood and Oliver County, air and siting permits for the Base Electron gas plant, and whether the next quarterly filing restates or quietly drops the gigawatt targets carried over from 2022.

The visible record for Applied Digital is front-loaded. Nearly every dated capacity and timeline claim comes from the 2021 to mid-2022 bitcoin-hosting era, and the input cuts off at claim 19161 in May 2022. The 2006 and 2008 entries (21053, 21054, 21067 to 21069) belong to the predecessor shell and say nothing about the datacenter program. The Polaris Forge, Delta Forge, Finland and Base Electron projects that now define the company carry no capacity, cost or dated claims in this extract, so the current AI build-out cannot be assessed against its own promises here.

Within what is on record, ambition grew quickly before any build caught up. Applied Digital started with a 100 MW Jamestown energy agreement (14511, 14373) and an 85 MW second site (14375, 14362). Within months it was talking about 500 MW of expansion demand (14512), then 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), and 5 GW over five years (14230, 14033). The 1.5 GW Antpool joint venture (14235, 14203, 24798) was layered on top. The 800 MW and 1.8 GW deadlines have both passed, yet they are still tagged not_yet_verifiable rather than resolved. That is the clearest gap in the file: the targets were never visibly restated or retired, only overtaken by the pivot to AI hosting. The Bitmain 200 PH equipment order (14504, 14416, 14378, 14360) sits in the same unresolved state.

The near-term operational promises held. The first Jamestown co-hosting facility (14509, 14414, 14372, 14364) and the remaining 45 MW for Q2 2022 (14227, 14196, 14009, 14030) are verified, as is the Garden City 200 MW facility for end-2022 (21465). Garden City has since been sold, so its capacity no longer counts toward the company's own portfolio. In short, small dated commitments landed and the large multi-year aggregates drifted.

The record contains no independent corroboration for the announced pace: no permits, no interconnection queue entries, no on-the-ground reporting. That matters most for Polaris Forge 2 and 3 and the Delta Forge sites, all listed as under construction with no supporting evidence attached.

Over the next 90 days, three things matter. First, ingest the post-2022 claims so the Polaris Forge timelines can be scored. Second, check whether the 5 GW target due March 2027 (14230) is still referenced or has been quietly replaced. Third, pull North Dakota, Louisiana and Alabama permit and queue filings to test whether energization dates match the construction status the company reports.

The capacity record we hold for Applied Digital ends in mid-May 2022. It does not cover the Polaris Forge, Delta Forge or Finland projects listed on the docket, so this read covers the company's crypto-hosting origins, not its current AI campus pipeline. The claims dated 2006 and 2008 (21053, 21054, 21067-21069) describe a predecessor shell's auditor, transfer-agent and DOT contracts. They have no bearing on facility delivery and should be treated as legacy noise.

Within the window we have, the headline numbers grew much faster than anything physical. The August 2021 baseline was a single 100 MW energy services agreement at Jamestown (14511, 14413), with a loosely stated 500 MW of expansion demand (14512). By March 2022 the company was quoting 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW over five years (14230). It also cited a 1.5 GW Antpool joint venture (14235, 14212) and a 200 MW Texas wind letter of intent (14236). That is roughly a fifty-fold increase in stated ambition in seven months, while energized capacity stayed at one 100 MW site plus an 85 MW second facility that existed only as customer reservations (14375, 14040).

On timing, Jamestown largely held. The Q4 2021 first-facility completion target is marked verified across every restatement (14509, 14414, 14372, 14364). The company reported 55 MW live by February 2022 (14008) and said the remaining 45 MW was due in Q2 2022 (14227, 14009). It then claimed the full 100 MW was online by May 13, 2022 (26005), ahead of that deadline. The weaker items are the 200 PH Bitmain equipment order and its Q4 2021 delivery date (14504, 14365). These were repeated unchanged four times without any later confirmation, which looks like restatement rather than progress. The 800 MW and 1.8 GW milestones have now passed. Nothing in this dataset shows them being met or formally withdrawn, so they should be classed as silently lapsed until filings show otherwise.

The record contains no independent corroboration of any kind: no permits, no MISO or ERCOT interconnection queue positions, and no site reporting. The Garden City wind letter of intent never progressed past that stage in our claims, and that project is now listed as sold.

Over the next 90 days, the priority is ingesting the 2023-2026 disclosures so the Polaris Forge 1 through 3 and Delta Forge capacity and energization dates can be tracked against this early pattern. That pattern is a reliable first site surrounded by aggregate targets that did not materialize. We should also pull MISO queue entries for Harwood and Oliver County, along with Louisiana and Alabama local permits, to test whether the current pace has physical backing that the 2022 targets never had.

The record supplied here breaks off in May 2022, so this read covers Applied Digital's crypto-hosting origins and not the Polaris Forge and Delta Forge buildout that dominates its current project list. The 2006 and 2008 entries (21053, 21054, 21067-21069) belong to a predecessor shell entity and should be separated from the facility history. Within the window we have, the capacity narrative scaled very fast. It began with a single 100 MW energy services agreement at Jamestown (14511, 14413) and an 85 MW second site (14375, 14362, 14040). By March 2022 it had become an 800 MW target for May 2023 (14228, 14031), 1.8 GW for May 2024 (14229, 14032), a five-year 5 GW ambition (14230, 14033), and a 1.5 GW Antpool joint venture (14235, 14203, 14036), all within about seven months. The near-term Jamestown promises held. First energization by end-2021 (14509, 14364), 55 MW online in February 2022 (14008), and the remaining 45 MW by Q2 2022 (14227, 14009) are all marked verified. The larger promises are where the record weakens. The May 2023 and May 2024 gigawatt milestones are still tagged not_yet_verifiable long after their deadlines, which means nobody has checked them, not that they are still pending. The Bitmain 200 PH equipment order (14504, 14360) was never confirmed either. The Texas 200 MW wind site carries contradictory tags for the same end-2022 promise: delayed in 3123 and verified in 21465. Garden City now shows as sold, and nothing in the visible claims explains that. The Antpool joint venture does not map to any project on today's list, and we have no record of it being formally withdrawn. Every visible claim is company-stated. There are no permits, no MISO or SPP interconnection queue positions, and no outside site reporting to test the pace against. That gap matters most for the newer campuses, whose capacity and cost fields are blank. Over the next 90 days, watch for confirmation that Polaris Forge 1 has moved from partial to full operation. Watch for energization evidence at Polaris Forge 2 in Harwood. Look for North Dakota Public Service Commission siting filings for Polaris Forge 3 and the adjacent Base Electron gas plant, and for local zoning or utility agreements at Delta Forge 1 in Boyce and Delta Forge 2 in Brookwood. Check whether the Finland campus gains a named site or power contract. Separately, the 2023 and 2024 milestone claims should be resolved and the Garden City tags reconciled before any trajectory score is published.

The capacity record for Applied Digital grows by an order of magnitude in eight months and is then never reconciled. It begins with a single 100 MW energy services agreement at Jamestown (14413, 14373, 14361) and an 85 MW second site (14375, 14362, 14040). By March 2022 the company was describing 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), and 5 GW over five years (14230, 14033), plus an Antpool joint venture, 1.21 Gigawatts LLC, sized at up to 1.5 GW (14235, 14203, 14036). The 2006 and 2008 entries (21053 through 21069) describe a predecessor entity's contracts and are not relevant to facility delivery.

The near-term Jamestown promises held. The first co-hosting facility was due by Q4 2021 and is verified across repeated restatements (14509, 14414, 14372, 14364). The remaining 45 MW was due by Q2 2022 and is also verified (14227, 14196, 14069, 14055, 14030, 14009). That record of execution is real but small relative to the headline targets. The larger promises are where the record weakens. The 800 MW and 1.8 GW deadlines have both passed and remain marked not yet verifiable, with no later claim in this record that revises or retires them. The 1.21 Gigawatts venture does not reappear after April 2022. The Bitmain 200 PH equipment order, promised by end-2021 (14504, 14416, 14378, 14360), was also never adjudicated. The Texas 200 MW wind facility carries conflicting statuses, delayed in 3123 and verified in 21465, against the same 2022 deadline, and the Garden City project is now listed as sold. Those conflicts should be resolved before either status is cited.

No independent corroboration is on file for any project: no permits, interconnection-queue positions, or site reporting. The current portfolio has shifted to AI-oriented campuses, including Polaris Forge 1 through 3, Delta Forge 1 and 2, a Finland campus, and a co-located gas plant. Yet the record supplied here carries no capacity or cost figures for any of them, and the claim list is truncated after 24810. The trajectory of the AI pivot therefore cannot be assessed from this record.

Over the next 90 days, watch for three things. First, whether the company's filings state energized megawatts at Polaris Forge 1 that can be matched to its partially operational status. Second, whether North Dakota Public Service Commission or regional grid-operator filings appear for Polaris Forge 2 and 3 and the Base Electron plant. Third, whether disclosures for Delta Forge in Louisiana and Alabama include utility service agreements rather than site announcements alone. Any formal restatement of the lapsed 2023 and 2024 gigawatt targets would also be material.

The supplied record breaks off in mid-May 2022, so this read covers only that window and cannot speak to the Polaris Forge, Delta Forge, Finland or Base Electron projects listed as on record. Within that span the capacity story inflates quickly. Through late 2021 the company described a 100 MW energy services agreement for its first site (14511, 14413, 14361), with 85 MW reserved at a second facility (14375, 14362, 14040). A 500 MW expansion figure (14512) was framed only as customer demand. By March 2022 the ambition had multiplied. The company set targets of 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032) and 5 GW over five years (14230, 14033). It also announced a 1.5 GW joint venture with Antpool (14235, 14203, 14036) and a 200 MW Texas wind letter of intent (14236, 14204). In May 2022 it added three more sites totalling 600 MW, promised by year-end (24808). Each step widened the gap between contracted power and headline capacity. The Texas site, Garden City, now appears as sold rather than built out. Near-term dates at Jamestown held. The first facility came online in Q4 2021 as promised (14509, 14364). It was reported at 55 MW as of February 2022 (14008), and the remaining 45 MW arrived on the Q2 2022 schedule (14227, 14009, 14030). That is a clean record on a single, utility-backed site. The larger dated promises are a different matter. The 800 MW and 1.8 GW deadlines passed long ago yet remain marked not yet verifiable, and nothing in this record shows the company formally retiring or restating them. The same is true of the December 2022 three-site promise and the Bitmain 200 PH delivery (14416, 14504). No permits, interconnection queue entries or site reporting are attached to any claim here. Verification rests only on the company's own subsequent disclosures, which support the Jamestown build and nothing beyond it. The pre-2021 items (21053, 21067-21069) belong to the predecessor shell and say nothing about delivery. Over the next 90 days, the most useful checks are whether later filings explicitly reconcile the lapsed 2023 and 2024 gigawatt targets and the Antpool venture, and whether the Polaris Forge and Delta Forge construction projects appear in grid operator interconnection queues and county permit records at the capacities the company now claims. Analysis should also be rerun on the complete claim history once the data feed is no longer truncated.

The supplied record ends partway through May 2022 (the last entry, claim 3119, is cut off), so it covers Applied Digital's crypto-hosting origins but none of the claims behind Polaris Forge, Delta Forge, Finland or the Base Electron gas plant. The 2006 and 2008 entries (21053, 21068) belong to a predecessor shell and say nothing about this buildout. What remains is the story of how the capacity narrative was set.

Capacity ambition grew very quickly. In August 2021 the company described a 100 MW services anchor (14511) alongside 500 MW of expansion demand (14512). By November it had added an 85 MW second facility (14375, 14362). By March 2022 the targets were 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032) and 5 GW over five years (14230, 14033, 26015), supported by a 1.5 GW Antpool joint-venture ceiling (14235, 14212, 24798). The company went from 100 MW contracted to 5 GW aspirational in roughly seven months. Nearly all of the growth sits in stated or not-yet-verifiable claims, not in executed interconnection.

The near-term Jamestown dates held. The Q4 2021 first-facility completion (14509, 14414, 14372, 14364) and the remaining 45 MW for Q2 2022 (14227, 14196, 14055, 14009) are both verified, consistent with 55 MW reported live as of February 2022 (14008). The larger targets fared worse. The 200 MW Texas wind facility, repeated across at least six filings (14236, 14214, 14204, 14063, 14034, 14015), carries a delayed flag against its end-2022 date (3123) and is now listed as sold. We see no acknowledged revision of the 800 MW or 1.8 GW deadlines in this record. The five-year horizon moved quietly from March to May 2027 between 14230 and 26015, and the Bitmain 200 PH order (14504, 14416, 14378, 14360) was never confirmed either way.

The record contains no permits, queue positions or site reporting. Independent support is limited to the verification of the Jamestown milestones.

Over the next 90 days, the priority is ingesting the post-2022 claims so the Polaris Forge 2 and 3 and Delta Forge construction schedules can be tested the way Jamestown was. Beyond that, watch for utility interconnection filings for Harwood and Oliver County, air and siting permits for the Base Electron plant, and whether the company formally retires the lapsed 800 MW and 1.8 GW targets or simply lets them fade.

The record for Applied Digital is lopsided. Every substantive capacity and timeline claim dates from August 2021 through May 2022, when the company was a crypto-hosting developer. Nothing in the record covers the projects that now define it: Polaris Forge 1 through 3, Delta Forge 1 and 2, the Finland campus, or the Base Electron gas plant. The 2006 and 2008 entries (21053, 21054, 21067-21069) appear to belong to a predecessor shell and have no bearing on delivery. Any read of the current pipeline therefore rests on an older, narrower base, and that gap is the most important finding here.

Within that window, announced capacity scaled very fast. A single 100 MW energy services agreement at Jamestown (14511, 14413) became 185 MW once an 85 MW second site was added (14375). By March 2022 the company was promising 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229, 24804) and 5 GW within five years (14230, 23467). In parallel came a 1.5 GW Antpool joint venture (14235, 14212) and a 200 MW Texas wind letter of intent (14236). Ambition grew roughly fiftyfold in seven months with no matching growth in disclosed interconnection or equipment commitments.

Only the near-term, site-specific dates held. Jamestown's first-phase completion by end of 2021 (14414, 14372) and the remaining 45 MW by mid-2022 (14227, 14009) are both verified, consistent with the 55 MW reported online (14008) and the later full 100 MW (26005). The portfolio-level targets are a different matter. The 800 MW and 1.8 GW dates passed in 2023 and 2024, yet they remain tagged not_yet_verifiable, and nothing in the record shows a restatement. That is a silent lapse, not a documented revision. The Antpool JV and the 200 PH Bitmain order (14504, 14416) also have no follow-up. Garden City, which began as the Texas LOI, now shows as sold, and that outcome was never reconciled against the original 200 MW framing. Beyond the verified Jamestown milestones, the record holds no independent corroboration, whether permits, queue positions or site reporting, for any of the scaled targets.

Over the next 90 days, the March 2027 deadline on the 5 GW target (14230) comes into view, which makes it a natural point for the company either to restate or to quietly retire the figure. The priority for this record is to ingest the Polaris Forge and Delta Forge capacity and energization claims, along with any interconnection or air-permit filings tied to the Oliver County gas plant. Until that happens, the credible signal is limited to one verified 100 MW site.

The capacity story in this record is mostly about how fast the ambition grew, and much less about how much was built. In August 2021 the anchor was a single 100 MW energy services agreement at Jamestown (14511, 14413), with a vague 500 MW of expansion demand alongside it (14512). Within seven months the framing had jumped to 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), and 5 GW over five years (14230, 14033). A 1.5 GW Antpool joint venture (14212, 14036, 3119) and a 200 MW Texas wind LOI at Garden City (14236, 14204) carried most of that expansion. Neither has a follow-on claim in this record showing land, power or energization. Garden City now appears as sold, which is a contraction the claim history never explains.

On timing, the near-term promises held. The first co-hosting facility was due in Q4 2021 and is verified (14364, 14414). Jamestown's remaining 45 MW was due in Q2 2022 and is also verified (14196, 14009), following 55 MW online by February 2022 (14008). The longer-dated targets are a different matter. The 800 MW and 1.8 GW deadlines have both passed, yet they are still tagged not_yet_verifiable. That makes them the clearest unresolved promises in the file, and nothing in the record shows the company formally restating or retiring them. The 1.21 Gigawatts JV also stops appearing after mid-2022 without any acknowledgement that it lapsed.

No independent evidence is attached to any of these claims: no permits, no interconnection queue positions, no site reporting. The verified Jamestown milestones rest on company disclosures alone. The record also has two integrity problems. The 2006–2008 claims (21053, 21054, 21067–21069) concern airport testing and shell-company administration and almost certainly belong to a predecessor entity, not the datacenter business. And the extract ends in May 2022 at claim 3123, so the Polaris Forge, Delta Forge, Finland and Base Electron projects listed as active have no claims visible here at all.

Over the next 90 days, the priorities are to backfill the Polaris Forge and Delta Forge claim history and to check it against county building permits in Dickey, Cass and Oliver counties in North Dakota, Rapides Parish in Louisiana, and Tuscaloosa County in Alabama. The matching utility interconnection filings should be checked as well. Air permitting for the Base Electron gas plant is the most concrete near-term marker of whether the Polaris Forge 3 power plan is real. Any restatement of aggregate capacity targets should be compared explicitly against the lapsed 2023 and 2024 figures.

Applied Digital's capacity record splits into two regimes, and the data on file captures only the first. In late 2021 the story was a single 100 MW Jamestown energy services agreement (14413, 14361) and an 85 MW second site (14375, 14362). By March 2022 that had become 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229, 24804), 5 GW within five years (14230, 23467), and a 1.5 GW Antpool joint venture (14235, 14212). That is a jump of roughly fifty times in four months, with no matching rise in contracted power. The claims from 2006 and 2008 (21053, 21054, 21067) come from a predecessor registrant and have no bearing on this trajectory.

The near-term, site-specific promises held. Jamestown's first phase was due by end of 2021 and is verified (14364, 14414). Its 55 MW energized in February 2022 (14008), and the remaining 45 MW due in Q2 2022 is also verified (14055, 14009). The portfolio-level targets fared differently. The May 2023 and May 2024 deadlines have both passed, yet they still sit at not_yet_verifiable with no restatement, reaffirmation or withdrawal on record. That looks like revision by silence. The Antpool vehicle and the 200 MW Garden City wind LOI (14204, 14236) were repeated almost word for word for two months. Garden City now shows as sold, and nothing in the claim history explains that outcome.

The dataset contains no independent corroboration of any kind, including permits, interconnection-queue positions and site reporting. The only external anchors are the verified Jamestown milestones. The current pipeline (Polaris Forge 1 through 3, Delta Forge 1 and 2, the Finland campus and the Base Electron gas plant) has no capacity, cost or date claims on file at all. As a result, the company's AI-era buildout cannot be tested against its own earlier promises from this record.

Over the next 90 days, the priority is ingesting post-2022 disclosures so the Polaris and Delta Forge projects carry stated megawatts and energization dates. The record should also mark the lapsed 800 MW and 1.8 GW targets explicitly rather than leave them in limbo. External checks worth pulling are utility and regional transmission queue entries for the Harwood and Oliver County sites, state siting or air permits for the Base Electron gas plant, and building permits at Boyce and Brookwood. Any of these would show whether the current pace is backed by power and permits or, like the 2022 gigawatt targets, mainly by announcements.

The first thing to note about Applied Digital's ledger is what it leaves out. The facility record ends in May 2022. Every project the company now leads with, including the Polaris Forge campuses, the Delta Forge sites, the Finland campus and the Base Electron gas plant, has no capacity or timeline claim attached. The 2006 and 2008 entries (21053, 21054, 21067) belong to the predecessor shell and say nothing about datacenter delivery. Read this as a trajectory of the crypto-hosting era, not of the AI-factory pivot.

Within that window, announced scale expanded much faster than delivery. A 100 MW energy services agreement at Jamestown (14413, 14361) and an 85 MW second facility (14375, 14040) sat alongside a 500 MW expansion-demand figure (14512). By March 2022 that had become 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229, 24804), 5 GW within five years (14230, 23467), and a separate 1.5 GW Antpool joint venture (14235, 14212). The GMR hosting relationship moved the other way, from 100 MW (14502) to 50 MW (14377, 14359), with no explanation in the record.

On dates, the small promises held. The first co-hosting facility was verified for Q4 2021 (14414, 14372, 14364), and the remaining 45 MW at Jamestown was verified for Q2 2022 (14227, 14055, 14009), consistent with the full 100 MW statement (26005). The large promises did not hold. The 800 MW and 1.8 GW deadlines have both passed while still marked not yet verifiable, and the record shows no reaffirmation or acknowledged revision of either. The Bitmain 200 PH equipment order (14504, 14416) was never verified. The 200 MW Garden City letter of intent (14236, 14074) ended in a sale. The Antpool joint venture never appears again.

The ledger contains no permits, interconnection queue entries or site reporting for any project. Independent corroboration of the current build pace is therefore absent from this record, not refuted by it.

Over the next 90 days, the main thing to watch is whether the 5 GW target (14230), which falls due in March 2027, is restated, quietly retired, or reframed around the Polaris Forge pipeline. Next is whether the fiscal second-quarter disclosure gives Polaris Forge 1 a dated, energized-megawatt figure that would move it from partially to fully operational. After that, watch for any state permitting or interconnection filings tied to Harwood, Oliver County and the adjacent gas plant, since those would be the first third-party evidence in this record for the post-2022 buildout.

The claim record supplied here effectively ends in May 2022. The last legible entry is 21466, and the earlier 2006 and 2008 items (21053, 21054, 21067 to 21069) belong to the predecessor shell rather than the datacenter business. Polaris Forge, Delta Forge, the Finland campus and the Base Electron gas plant therefore carry no capacity or timeline claims in this file. Any read of their trajectory has to rest on status fields alone, and readers should treat this as a read of the early Jamestown and crypto-hosting era, not the current AI-campus program.

Within that window, the scaling pattern is steep and front-loaded. A 100 MW energy services agreement at Jamestown (14511, 14413, 14361) and an 85 MW second site (14375, 14362, 14040) became, by March 2022, targets of 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032, 21463) and 5 GW over five years (14230, 14033, 23467). The Antpool joint venture was layered on top at up to 1.5 GW (14203, 14212, 14059). That is a roughly fiftyfold expansion of stated ambition in seven months, with no new site-level commitments disclosed to underpin it.

The near-term dates held. The Q4 2021 completion of the first co-hosting facility (14509, 14414, 14372, 14364) and the remaining 45 MW at Jamestown by Q2 2022 (14227, 14196, 14069, 14030, 14009) are both marked verified, and the 55 MW interim figure (14008, 14054) was repeated consistently. The aggregate targets are a different matter. The May 2023 and May 2024 deadlines have long passed but remain flagged not_yet_verifiable, and nothing in the record restates, revises or retires them. The 200 MW Garden City wind LOI (14236, 14204, 14063) never acquired a delivery date, and the site now sits as sold. The Antpool JV and the Bitmain 200 PH order (14504, 14416, 14378) have no follow-through entries either. This is a pattern of quiet abandonment rather than acknowledged revision.

The record contains no permits, interconnection-queue filings or site reporting, so the announced pace is unsupported by independent evidence here.

Over the next 90 days, the priority is backfilling claims for the Polaris Forge and Delta Forge projects so the current program can be assessed at all. Beyond that, watch for energization evidence on the remaining Polaris Forge 1 capacity and for North Dakota Public Service Commission and air-permit filings tied to the Base Electron plant. MISO or SPP queue positions for Polaris Forge 3 and the Delta Forge sites would also be telling. Separately, the pipeline should formally adjudicate the expired 2023 and 2024 capacity targets.

The record supplied here ends in May 2022, so this read covers Applied Digital's hosting-era build-out and cannot assess the Polaris Forge, Delta Forge, or Finland projects now listed as under construction or announced. Within that window, the scale of the ambition changed sharply. Through late 2021 the company described a single 100 MW energy services agreement at Jamestown (14511, 14413, 14373, 14361), an 85 MW second site (14375, 14362, 14040), and a loosely framed 500 MW of expansion demand (14512). By March 2022 that had become 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), and 5 GW within five years (14230). It sat alongside an Antpool joint venture for up to 1.5 GW (14235, 14212) and a 200 MW Texas wind LOI (14236). That is roughly a tenfold jump in stated capacity over about four months, with no new power contracts on record to support it beyond the original 100 MW agreement. Delivery on the small, concrete commitments held up. The Q4 2021 Jamestown completion (14509, 14414, 14372, 14364) and the remaining 45 MW by Q2 2022 (14227, 14196, 14009, 14030) are both verified, and the 55 MW online figure (14008) fits that sequence. The large targets are a different matter. The deadlines on 14228 and 14229 passed years ago and both remain unresolved in the record. The 5 GW goal carries a March 2027 date in 14230, but its restatement in 14033 drops the date entirely. The claim of 600 MW of construction starts by end-2022 (3125) is also unresolved. The Texas LOI site, Garden City, is now marked sold, and no claim acknowledges any change of course. Independent evidence in this record is thin. The verified tags are confined to Jamestown, and there are no permit or interconnection-queue entries covering the multi-gigawatt plans. The 2006 and 2008 claims (21053, 21054, 21067 to 21069) appear to come from a predecessor shell entity and should be separated out. Over the next 90 days, the priority is ingesting the post-2022 claims so that 14228, 14229, and 3125 can be scored against actual energized capacity. For the live projects, watch for utility and RTO interconnection filings, county permits, and energization dates at Polaris Forge 2 and 3 and at the Delta Forge sites. Watch too for whether the Base Electron gas plant gains a formal regulatory filing or remains only an announcement.

The claim record supplied here stops mid-entry in May 2022 (claim 19159). Everything below is a read of the crypto-hosting era plus the current project roster. It is not a read of the Polaris Forge or Delta Forge AI campuses, whose capacity and timeline claims are not in this extract. The 2006 and 2008 entries (21053, 21054, 21067-21069) belong to a predecessor shell and say nothing about datacenter delivery.

On scale, the announced ambition grew steeply within about seven months. The anchor was a 100 MW energy services agreement at Jamestown (14511, 14413, 14361), with a vaguely sourced 500 MW of expansion demand (14512). By March 2022 that had become 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), and 5 GW over five years (14230, 14033). Alongside sat the Antpool joint venture, 1.21 Gigawatts LLC, for up to 1.5 GW (14235, 14203, 14036), and a 200 MW Texas wind LOI at Garden City (14236, 14214). None of these targets was tied to a named interconnection agreement in the record. The 85 MW second facility (14375, 14362, 14040) was described only as reserved energy, never as a site.

On dates, the near-term Jamestown milestones held. The first facility was completed by end-2021 (14509, 14414, 14372, 14364). The 55 MW online figure was dated to February 2022 (14008), and the remaining 45 MW was promised for Q2 2022 and is marked verified (14227, 14196, 14009). The longer-dated aggregate targets are a different story. The 800 MW and 1.8 GW deadlines have both passed and remain flagged not_yet_verifiable. The extract contains no later claim that restates, revises, or retires them, which is the pattern to note: the targets stop being repeated rather than being reconciled. Garden City, the vehicle for the Texas wind LOI, now carries a status of sold. The 1.21 Gigawatts JV has no project attached on the roster at all.

The record contains no independent evidence, such as permits, MISO or SPP queue positions, or site reporting, for any of these claims. That gap is the central weakness in assessing pace.

Over the next 90 days, a reviewer should first pull the full post-2022 claim history for Polaris Forge 2 and 3, Delta Forge 1 and 2, the Finland campus, and the Base Electron gas plant. With that in hand, they should check whether interconnection filings and energization dates for Harwood and Oliver County match management's stated megawatt ramps. They should also watch whether the company ever formally closes out the 1.8 GW and 5 GW framings or quietly replaces them with AI-campus figures.

The claims on file stop in May 2022, while the project roster now includes Polaris Forge 1 through 3, two Delta Forge sites, a Finland campus, and an adjacent gas plant. None of those carry a capacity, cost, or dated claim here. So this read covers the crypto-hosting era that preceded the AI pivot, not the pivot itself. Claims 21053 through 21069 (a Volpe Center contract, airport testing, 2008 auditor and transfer-agent appointments) look like a predecessor-entity or entity-resolution artifact. They should be excluded from any APLD delivery score.

Announced capacity grew very quickly. The 100 MW Jamestown energy agreement (14413, 14373) became 185 MW once the 85 MW second facility was added (14375, 14040). That figure became 500 MW of planned expansion demand in claim 14512. By March 2022 the company had laid out a ladder of 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032, 24804), and 5 GW within five years (14230, 14033). On top of that sat a separately framed Antpool joint venture of up to 1.5 GW (14235, 14036). Repeated restatement across filings does not add capacity, and nothing in the record shows the 1.5 GW joint venture ever producing a sited project.

Near-term dates were met. Jamestown reached 55 MW by February 2022 (14008), and its remaining 45 MW is verified against the Q2 2022 target (14227, 14069). The 200 MW Garden City wind site is verified against end-2022 (24806). Garden City has since been sold, however, so delivered capacity left the portfolio. The longer-dated targets are the problem. Both the May 2023 and May 2024 deadlines have passed and are still tagged not_yet_verifiable. The Bitmain 200 PH order (14416, 14360) has the same status. No claim on file retires, revises, or acknowledges any of them, which amounts to an unacknowledged drift until shown otherwise.

The record contains no independent evidence: no permits, no interconnection queue entries, no on-the-ground reporting. The verified tags confirm the company's own completion statements, not third-party observation.

Over the next 90 days, the priority is ingesting post-2022 filings so that Polaris Forge and Delta Forge carry dated MW and energization claims. Those should be paired with MISO, SPP, and Southern Company queue positions and with county permits in Ellendale, Harwood, Oliver County, Boyce, and Brookwood. The 800 MW and 1.8 GW targets also need to be explicitly scored against what was actually energized.

The claim record supplied here ends abruptly in May 2022 (the last entry, 21466, is cut off mid-text), so this read covers Applied Digital's crypto-hosting era and cannot assess the Polaris Forge, Delta Forge, Finland or Base Electron projects, none of which carry a capacity or timeline claim in this set. The 2006 and 2008 entries (21053, 21054, 21067-21069) appear to belong to a predecessor shell and say nothing about datacenter delivery. Ambition scaled steeply within months. In late 2021 the company described a 100 MW energy services agreement at Jamestown and 85 MW reserved at a second site (14361, 14362). By March 2022 that had grown to 800 MW by May 2023, 1.8 GW by May 2024 and 5 GW within five years (14228, 14229, 14230), plus a separately framed Antpool joint venture of up to 1.5 GW (14235, 14212). These figures were repeated almost verbatim through April and May 2022 (14031, 14032, 21463) without a site-level build plan behind them. Near-term dates held. Jamestown's Q4 2021 completion (14414, 14372, 14364) and the Q2 2022 addition of the remaining 45 MW (14227, 14196, 14030) are verified, and the 55 MW online figure (14008) was consistent with that sequence. The program-level targets are a different story. The 800 MW and 1.8 GW deadlines passed in 2023 and 2024, yet both are still tagged not yet verifiable. That is a data gap rather than a vindication, and it should be closed. The 200 MW Garden City wind LOI (14236, 14204) ended in a sale, and the 1.21 Gigawatts JV has no follow-up claim at all. Both look like capacity that was announced and later quietly dropped from the narrative, not delivered. Nothing in this record supplies permits, interconnection queue positions or site reporting, so the 2022 gigawatt pace rests only on company statements. Over the next 90 days the priority is to backfill post-2022 claims for the Polaris Forge and Delta Forge campuses and to test them against MISO and SPP queue filings, North Dakota PSC dockets for the Base Electron gas plant, and Louisiana and Alabama local permitting. The 800 MW and 1.8 GW claims should also be resolved against reported energized capacity, and Polaris Forge 1's partially operational status should be checked against stated megawatts in service.

The record shows two different companies under one ticker, and the gap between them is the main finding. Through mid-2022 Applied Digital was a crypto-hosting operator. Its capacity story began with a single 100 MW energy services agreement at Jamestown (14511, 14413), plus a reserved 85 MW second site (14375, 14040). Within months it grew to a 1.5 GW Antpool joint venture (14203, 14235), an 800 MW target for May 2023 (14228), 1.8 GW for May 2024 (14229) and 5 GW over five years (14230). That is roughly a fiftyfold increase in ambition, set against about 55 MW actually energized (14008).

Near-term promises mostly held. The first co-hosting facility was verified against its Q4 2021 date (14414, 14372), and the remaining 45 MW at Jamestown was verified against Q2 2022 (14227, 14009). The longer-dated promises did not hold. Both the 800 MW and 1.8 GW deadlines have passed, yet the record still marks them not yet verifiable, and no later claim confirms, revises or retires them. The Texas 200 MW wind facility repeated unchanged across six filings (14236 through 14063), was later tagged delayed (3123), and now appears as sold. We found no claim acknowledging that change. The Antpool JV also stops appearing after spring 2022 without any stated wind-down.

The company's current portfolio of Polaris Forge 1 through 3, Delta Forge 1 and 2, the Finland campus and the Base Electron gas plant has no capacity, cost or dated timeline claims in this record. The supplied history also ends at mid-2022, so we cannot test the AI-era pace against the company's own statements. The record contains no permit filings, interconnection-queue entries or site reporting either. Verification here rests only on the company's own operational disclosures, which is not independent evidence. For now, the announced pace is unconfirmed rather than contradicted.

In the next 90 days, the first thing to check is whether the 5 GW by March 2027 target (14230) is restated, quietly dropped or folded into the Polaris and Delta Forge totals. Second, any ND Public Service Commission siting or air-permit activity for the Oliver County campus and the adjacent Base Electron plant would be the first outside check on Polaris Forge 3's timing. Third, utility interconnection filings for Harwood and Boyce would show whether those sites have committed power. Finally, the share of Polaris Forge 1 that moves from partially to fully operational will show whether the company's recent near-term dates hold the way its 2021–22 ones did.

The record supplied here is lopsided: it ends abruptly on 2022-05-13, inside the bitcoin-hosting era, while the project list now centers on Polaris Forge, Delta Forge and a Finland campus that have no dated capacity or cost claims on file. Any read of the current AI build-out is therefore provisional, and that gap is itself the first finding.

Within the visible window, ambition scaled much faster than delivery. Applied Digital began with a single 100 MW energy services agreement at Jamestown (14511, 14413) and an 85 MW second site (14375, 14362). Within seven months that became a 1.5 GW Antpool joint venture (14235, 14203), an 800 MW target by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW over five years (14230). Those intermediate deadlines have long passed and are still tagged not_yet_verifiable. No claim in the record confirms they were met, revised or withdrawn. That is the profile of a target set quietly retired rather than formally reset.

The near-term operational promises held up better. Jamestown's first co-hosting facility was targeted for Q4 2021 and verified (14414, 14372, 14364). The 55 MW reported online by February 2022 (14008) and the remaining 45 MW promised for Q2 2022 (14227, 14009) are also verified. The Texas wind site is the clearest slip. A 200 MW letter of intent repeated across at least six filings (14236, 14204, 14063) carried a 2022 operating date that is now marked delayed (3123), and Garden City now appears as sold. The Bitmain 200 PH equipment order (14504, 14360) and its Q4 2021 delivery (14365) were never independently confirmed.

Independent corroboration is thin. The verified tags attach only to Jamestown, and the record contains no permits, interconnection-queue entries or site reporting for any Forge campus or for the Base Electron gas plant.

Over the next 90 days, three things matter most. The first is whether Polaris Forge 1 moves from partially to fully operational with tenant-attributable megawatts stated in filings rather than in press language. The second is state siting and air-permit dockets for Polaris Forge 3 and the adjacent gas plant, along with grid-operator queue positions for the Louisiana and Alabama Delta Forge sites. The third is whether management ever reconciles the 2022 gigawatt roadmap with the current pipeline, or lets it lapse as it has so far.

The record supplied here effectively ends in May 2022, so it says a lot about Applied Digital's crypto-hosting origins and almost nothing about the Polaris Forge and Delta Forge AI campuses that now define the company. None of those later projects carries a capacity or cost figure on file, and that gap is the first thing a reader should weigh. The 2006 and 2008 entries (21053, 21054, 21067, 21068, 21069) appear to come from a predecessor shell entity and have no bearing on the datacenter trajectory.

On the narrow question of Jamestown, the company delivered. The first co-hosting facility was promised for Q4 2021 and is verified (14509, 14414, 14372, 14364). The 100 MW energy agreement was repeated consistently (14413, 14373, 14233, 24796). It was framed as 55 MW online by February 2022 (14008, 14054), with the remaining 45 MW due by end of Q2 2022, and that milestone is also verified (14009, 14227, 14196). Equipment-side promises were never independently confirmed, including the 200 PH Bitmain order (14504, 14416, 14378, 14360) and the Q4 2021 hardware delivery (14510, 14365).

The portfolio-level ambition is a different story. Planned demand of 500 MW in August 2021 (14512) became, by March 2022, targets of 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), and 5 GW over five years (14230, 14033). It was paired with an Antpool joint venture of up to 1.5 GW (14235, 14203, 14036). Both dated targets have long passed and remain unverifiable, and the record shows no acknowledgement or restatement of either. The Texas wind site illustrates the slippage. It recurred as a 200 MW letter of intent across at least seven filings (14236, 14204, 14063, 14015) and its 2022 operational date is marked delayed (3123). The project now sits on record as sold. The 85 MW second facility (14375, 14362, 14040) is never tied to a named site.

No permits, interconnection-queue entries, or site reporting are attached to any claim, so the announced pace rests on company statements alone. Over the next 90 days, the priority is to ingest post-2022 claims for Polaris Forge 1 through 3, Delta Forge 1 and 2, and the Base Electron gas plant. Each should then be checked against North Dakota, Louisiana, and Alabama permitting records and MISO and SPP queue positions. Energization dates for the Harwood and Ellendale buildings are the nearest test of whether the AI-era schedule holds better than the 2022 gigawatt targets did.

The record supplied here ends mid-entry in May 2022, so this read covers Applied Digital's crypto-hosting origins, not the Polaris Forge, Delta Forge or Finland projects that now define the company. The 2006 and 2008 entries (21054, 21053, 21069, 21067, 21068) predate the datacenter business and carry no delivery signal. The capacity trajectory is a sharp step-change rather than a gradual build. In August 2021 the company framed itself around a single 100 MW energy services agreement (14511, 14413) and an 85 MW second site (14375, 14362, 14040), against a loosely stated 500 MW of demand (14512). By March 2022, with roughly 55 MW energized at Jamestown (14054, 14008), the ambition had jumped to 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032) and 5 GW over five years (14230, 14033, 21464), alongside an Antpool joint venture sized at up to 1.5 GW (14235, 14036). That is a roughly ninefold increase in stated near-term capacity within seven months, announced while less than a sixth of the first site was live. Near-term dates held. The Jamestown first-facility completion (14509, 14414, 14364) and the remaining 45 MW for Q2 2022 (14227, 14196, 14009, 14030) are both verified. The larger targets are a different matter. The 800 MW and 1.8 GW deadlines have long passed and remain unverified. Nothing in this excerpt shows them being formally walked back, so whether they were revised without acknowledgement cannot be settled from what is here. The 5 GW goal carries a March 2027 deadline in 14230 but none in 14033 or 21464, a framing drift worth noting. The Garden City 200 MW wind LOI (14236, 14204, 14015) ended in a sale, not a buildout. The record contains no permits, interconnection-queue entries or site reporting, so the announced pace has no independent corroboration here. Over the next 90 days, watch for three things. First, whether Polaris Forge 1 moves from partially operational to fully operational with tenant-confirmed energization. Second, North Dakota Public Service Commission and interconnection filings for Polaris Forge 2, Polaris Forge 3 and the adjacent Base Electron gas plant. Third, whether Delta Forge 1 and 2 show utility or MISO and Southern Company queue evidence matching their under-construction status. The March 2027 5 GW marker is also near enough that any quiet restatement will be telling.

The supplied record breaks off at claim 19161 in May 2022, so this read covers Applied Digital's hosting-era build-out and cannot assess the Polaris Forge, Delta Forge, Finland or Base Electron projects listed on the roster. The 2006 and 2008 entries (21053, 21054, 21067-21069) come from the predecessor shell entity and have no bearing on facility delivery.

The capacity story inflates quickly from a small, delivered base. In August 2021 the company described a single 100 MW energy services agreement at Jamestown (14511, 14413) alongside a vague 500 MW of expansion demand (14512). By November it added an 85 MW second facility (14375, 14362). In March and April 2022 the ambition jumped by more than an order of magnitude: 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032, 21463), 5 GW over five years (14230, 14033), an Antpool joint venture of up to 1.5 GW (14235, 14212, 14036), and a 200 MW Texas wind LOI (14236, 14204). The Texas site appears as Garden City on the roster and has since been sold.

On dates, the near-term Jamestown promises held. Completion of the first facility in Q4 2021 is verified (14509, 14414, 14372, 14364). So is the remaining 45 MW in Q2 2022 (14227, 14196, 14055, 14030), following 55 MW online by February 2022 (14008). The larger targets are a different story. The 800 MW and 1.8 GW deadlines both passed years ago and remain marked not_yet_verifiable. Neither the 1.21 Gigawatts JV nor the 600 MW-plus additional-sites claim (19161) receives any follow-up within the visible record, so revision without acknowledgement cannot be ruled out. The Bitmain 200 PH equipment order (14504, 14416, 14378, 14360) is also unresolved. SparkPool kept being cited as a partner into late November 2021 (14376), which deserves a check against that pool's wind-down that autumn.

The record contains no independent corroboration of pace: no permits, interconnection queue positions or site reporting. Every capacity and cost field on the project roster is blank, including for the partially operational Polaris Forge 1.

Over the next 90 days, the first priority is to backfill post-2022 claims so the AI-campus trajectory can be scored. Three things to watch: whether the next 10-Q or 10-K explicitly reconciles the lapsed 800 MW and 1.8 GW targets; MISO and SPP queue and North Dakota PSC filings for Polaris Forge 2 and 3 and the Base Electron gas plant; and energization evidence that would move Polaris Forge 1 from partial to full operation.

The record supplied here is truncated mid-entry around May 2022 (claim 19161), so this read covers only Applied Digital's crypto-hosting era and cannot speak to Polaris Forge, Delta Forge, Finland or the Base Electron gas plant, none of which appear in the claims shown. The earliest entries (21053, 21054, 21067 through 21069, dated 2006 and 2008) concern airport testing, a transfer agent and an auditor. They read as predecessor-shell filings and should be treated as noise rather than delivery history. Within the usable window, ambition scaled sharply and quickly. In August 2021 the anchor was a single 100 MW energy services agreement with a loosely stated 500 MW demand pipeline (14511, 14512). An 85 MW second site followed by November (14375, 14362). By March 2022 the company had layered on an 800 MW target for May 2023, 1.8 GW for May 2024, 5 GW over five years and a 1.5 GW Antpool joint venture (14228, 14229, 14230, 14235). It added a further 600 MW of sites promised before the end of 2022 (19161). That is roughly a fiftyfold increase in stated ambition inside nine months, resting on one operating site. The near-term Jamestown dates held. Both the Q4 2021 first-facility completion (14364) and the Q2 2022 remaining 45 MW (14055, 14196) are verified, and the 55 MW energization reported in February 2022 (14008) is consistent with that sequence. The larger targets are a different story. The 800 MW and 1.8 GW deadlines lapsed in 2023 and 2024 yet still sit at not_yet_verifiable (14031, 14032, 21463). This reflects a gap in the claims record, not evidence of delivery, and no later claim in this excerpt revises or acknowledges either target. The Bitmain 200 PH order (14378) has the same unresolved status. The 200 MW Garden City wind LOI (14204, 14214) became a facility that was ultimately sold, which matters for any capacity tally. The material contains no permits, interconnection queue positions or site reporting, so independent corroboration of the announced pace is absent. For the next 90 days, the priority is to backfill and resolve the lapsed 2023 and 2024 targets against filed results, then ingest the Polaris Forge and Delta Forge claim history. Specific checks: North Dakota PSC and county filings for Polaris Forge 3 and the Base Electron plant, utility interconnection agreements for Harwood and the Louisiana and Alabama sites, and whether the next quarterly filing restates energized versus contracted megawatts.

The supplied record cuts off mid-claim in May 2022 (claim 19161), so this read covers only Applied Digital's bitcoin-hosting origins. The Polaris Forge, Delta Forge, Finland and Base Electron projects on the roster have no claims in this extract, and their trajectory cannot be judged from it. The 2006 and 2008 entries (21054, 21053, 21067-21069) describe airport testing, a Volpe Center contract and transfer-agent appointments. They look like predecessor-shell history and should not be read as datacenter promises.

The capacity arc is steep. It begins with a single 100 MW energy services agreement at Jamestown (14511, 14413) and a planned 85 MW second site (14375, 14040). By March 2022 the company had set an 800 MW target for May 2023 (14228), 1.8 GW by May 2024 (14229, 21463), 5 GW over five years (14230), up to 1.5 GW through the Antpool joint venture (14235, 14212), a 200 MW Texas wind LOI (14236), and 600 MW or more of additional sites before year-end 2022 (19161). That is roughly an eighteen-fold expansion of ambition in seven months, set against about 100 MW of contracted power.

The near-term dates held. Jamestown's first-phase completion in Q4 2021 is verified (14509, 14414, 14372, 14364), as is the remaining 45 MW by Q2 2022 (14227, 14196, 14055, 14009). The headline targets look different. The 800 MW and 1.8 GW deadlines passed years ago and remain marked not_yet_verifiable. The Bitmain 200 PH equipment order (14504, 14416, 14378) has never resolved either. Nothing in the record shows the company acknowledging the shortfall. The Antpool joint venture drops out of the claims without a closing statement. Garden City now appears as sold, which the extract never connects to the original 200 MW LOI.

The extract contains no permits, interconnection queue entries or site reporting. The only independent signal is the verified status of the Jamestown milestones, which supports the modest 100 MW build and nothing beyond it.

Over the next 90 days, the first priority is to complete the dataset with post-2022 claims so the Polaris and Delta Forge commitments can be scored. After that, watch for energization of the remaining Polaris Forge 1 capacity and for Polaris Forge 2 and 3 entries in MISO or SPP interconnection queues. Also watch for North Dakota Public Service Commission or air-permit filings for the Base Electron gas plant, and for any restatement of the 2023-2024 gigawatt targets in filings.

The record supplied here stops in May 2022, so this read covers Applied Digital's crypto-hosting era and cannot yet test the AI-campus pipeline (Polaris Forge 1 through 3, Delta Forge 1 and 2, Finland, and the Base Electron gas plant), none of which carries a dated claim in this extract. The 2006 and 2008 entries (21053, 21054, 21067-21069) appear to belong to a predecessor shell and say nothing about facility delivery.

On capacity, the trajectory through early 2022 escalates sharply. The opening frame was a 100 MW energy services agreement at Jamestown (14511, 14413) against 500 MW of stated demand (14512). Within seven months that became 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), 5 GW over five years (14230, 14033), and up to 1.5 GW through the Antpool joint venture 1.21 Gigawatts LLC (14235, 14203). These targets were restated almost verbatim across filings rather than refined, and the record never reconciles them. Both the 800 MW and 1.8 GW deadlines have passed, remain tagged not_yet_verifiable, and carry no visible acknowledgement of a miss or restatement. The Texas 200 MW wind LOI (14236, 14204) ends with Garden City marked sold, so that capacity left the portfolio rather than being delivered.

On dates, the near-term promises held. The first co-hosting facility was due by year-end 2021 (14509, 14414, 14364) and verified. The remaining 45 MW at Jamestown was due in Q2 2022 (14227, 14069, 14030) and also verified, after 55 MW was reported online as of February 2022 (14008). The Bitmain 200 PH equipment deliveries (14504, 14416, 14365) never reached verification. The pattern is reliable execution on the first site paired with long-range targets that were never revisited.

The record contains no independent corroboration beyond the verification tags: no interconnection queue entries, permits, or site reporting. It also gives no capacity or cost figure for any current project.

Over the next 90 days, the priority is to get the post-2022 claim history into the record. Specifically, that means dated energization targets for each Polaris Forge 1 building, the Polaris Forge 2 and 3 schedules, and whether Delta Forge sites have filed with MISO or Southern Company. It also means checking whether the expired 800 MW and 1.8 GW targets were ever formally retired.

Applied Digital's record shows a small, delivered core under a much larger stack of headline targets that the record never reconciles. The anchor is a 100 MW energy services agreement at Jamestown (14413, 14361), soon joined by an 85 MW second facility (14375, 14040). In March 2022 the ambition jumped by an order of magnitude: 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), and 5 GW over five years (14230, 14033). There was also a 1.21 Gigawatts joint venture with Antpool sized at up to 1.5 GW (14235, 14203) and a 200 MW Texas wind LOI (14236). That Texas site, Garden City, now carries a sold status, so part of the 2022 pipeline has already left the portfolio.

On timing, the near-term promises held. Jamestown's Q4 2021 completion is verified (14414, 14364), and so is the remaining 45 MW for Q2 2022 (14227, 14009). The longer-dated targets are a different matter. The 800 MW and 1.8 GW deadlines have both passed but remain not_yet_verifiable. The 1.8 GW goal was quietly restated from 'by May 2024' to 'end of FY2024' (21463). The five-year 5 GW horizon also rolled forward from March to May 2027 between 14230 and 26015, with no acknowledgement either time. The pledge of three more sites totaling over 600 MW under construction by end-2022 (21467) is likewise unresolved. The Bitmain 200 PH order and its related equipment deliveries (14504, 14416, 14365) were never confirmed.

The record contains no independent corroboration of any kind: no permits, no interconnection-queue entries, no site reporting. Even the 55 MW online as of February 2022 (14008) rests on the company's own word. Two data-hygiene problems also stand out. The 2006 and 2008 claims (21053, 21054, 21067-21069) appear to belong to a predecessor registrant and should be excluded from this trajectory. And the claim history in this extract ends in May 2022, while the project list now includes Polaris Forge 1 through 3, Delta Forge 1 and 2, a Finland campus, and an adjacent gas plant, none of which has a single capacity or timeline claim on file.

Over the next 90 days, the priority is closing that gap. That means ingesting post-2022 claims for the Polaris and Delta Forge builds and checking whether Polaris Forge 1 moves from partially to fully operational. It also means looking for state or ISO interconnection filings that support Polaris Forge 3 and the Base Electron gas plant, and for permitting activity at Harwood, Boyce and Brookwood. Until that evidence exists, Applied Digital's verified delivery is limited to the Jamestown-era megawatts, and its gigawatt-scale claims should be treated as unverified.

The record supplied here ends in May 2022, so it covers Applied Digital's crypto-hosting origins but not the Polaris Forge, Delta Forge, Finland or Base Electron projects that now dominate its project list. The 2006 and 2008 items (21053, 21054, 21067, 21068, 21069) predate the datacenter business and say nothing about delivery. Within that window, the pattern is clear. Small, site-specific commitments were met on schedule. Large portfolio targets grew faster than any evidence of sites to support them.

At Jamestown the numbers stayed disciplined. The 100 MW energy services agreement was restated without change across 14413, 14373, 14361 and 14013. The 85 MW second facility was repeated consistently in 14375, 14362 and 14040. Then in March 2022 the ambition jumped by an order of magnitude: 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032) and 5 GW over five years (14230, 14033). The 5 GW horizon drifted quietly from March 2027 in 14230 to May 2027 in 26015, which looks like a rolling five-year window rather than a fixed commitment. 21463 recast the 1.8 GW goal as end of fiscal 2024, a roughly equivalent framing, while 21467 added a nearer test: up to three sites totalling more than 600 MW under construction by the end of 2022.

On dates, Jamestown's first facility by Q4 2021 is verified (14414, 14372, 14364, 14509), as is the remaining 45 MW by Q2 2022 (14227, 14196, 14069, 14055, 14030, 14009). The Bitmain 200 PH equipment deliveries remain unadjudicated (14504, 14416, 14378, 14360, 14365). More consequentially, the 800 MW and 1.8 GW deadlines passed years ago and are still marked not yet verifiable. The 200 MW Texas wind LOI (14236, 14204, 14214) never carried a date and the site is now listed as sold. The 1.5 GW Antpool joint venture (14235, 14212, 14036) maps to no current project. Neither exit appears in the claims as an acknowledged revision.

Independent corroboration is thin. Verification exists only for Jamestown, and the record contains no permit, interconnection-queue or site-visit evidence for any project.

Over the next 90 days, the priorities are adjudicating the 2023 and 2024 capacity targets against energized megawatts, extending this record past 2022 to capture Polaris Forge commitments, and checking state siting and interconnection filings for Polaris Forge 3 and the adjacent gas plant. It is also worth watching whether the Garden City sale and the dormant joint venture are ever explicitly reconciled.

The supplied record breaks off partway through May 2022 (claim 21463 is cut mid-line), and it contains no claims for Polaris Forge, Delta Forge, the Finland campus or the Base Electron gas plant. This read therefore covers the bitcoin-hosting era that came before the AI pivot. The 2006 and 2008 entries (21053, 21054, 21067-21069) look like filings from a predecessor shell and say nothing about datacenter delivery.

The capacity story is one of rapid escalation off a small verified base. Through late 2021 the company framed itself around a 100 MW energy services agreement at Jamestown (14413, 14361) plus an 85 MW second site (14375, 14362, 14040). In March 2022 that became 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), a 1.5 GW Antpool joint venture (14235, 14203) and 5 GW within five years (14230, 14033). Two quiet revisions followed within weeks. The 1.8 GW deadline was restated as end of fiscal 2024 (21463), and the five-year window moved from March to May 2027 (26015). Neither change was acknowledged. Each pushed the yardstick later while the headline number stayed the same.

On dates, the company met the near-term promises tied to a named site. The first Jamestown facility arrived in Q4 2021 (14414, 14364), and the remaining 45 MW came online in Q2 2022 (14227, 14009). Everything outside Jamestown is weaker. The 200 PH Bitmain equipment order due by end of 2021 (14504, 14416, 14378, 14360) and the related deliveries (14510, 14365) were never verified. The 200 MW Garden City wind LOI (14236, 14204, 14034) never produced a dated timeline, and that site is now listed as sold. The pledge of three more sites totaling over 600 MW under construction by end of 2022 (21467) remains unresolved.

Independent corroboration in the record is thin. Only the Jamestown milestones carry verified status, and no permits, interconnection-queue entries or on-the-ground reporting are attached to any of the larger targets. The 800 MW and 1.8 GW deadlines expired long ago but are still marked not_yet_verifiable. That is an adjudication gap, not evidence of delivery.

Over the next 90 days, the first priority is closing out claims 14228, 14229, 14031, 14032 and 21467 against reported energized capacity. Next, extend ingestion to cover the Polaris Forge and Delta Forge claims, so the AI-era promises can be checked against the same pattern of retiming. Finally, look for county permits and utility or RTO interconnection filings for Harwood, Oliver County, Boyce and Brookwood, which would show whether construction status is backed by grid commitments.

The supplied record stops in mid-2022, partway through claim 19158. Any read of the current Polaris Forge and Delta Forge buildout therefore rests on project status fields, not on dated claims. That gap matters, because every AI-era project on record lists no capacity and no cost.

Within the window that is documented, announced capacity grew far faster than anything delivered. Applied Digital began with a 100 MW energy services agreement at Jamestown (14511, 14413, 14373) and an 85 MW second site (14375, 14362). Within roughly four months the company was guiding to 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032, 24804), and 5 GW over five years (14230, 14033, 19158). It also announced a separate Antpool joint venture of up to 1.5 GW (14235, 14203, 14036). The 1.8 GW figure was restated against fiscal year-end (24804) without a change in substance.

The near-term promises held. Jamestown's first co-hosting facility was verified complete on the end-2021 target, repeated four times (14509, 14414, 14372, 14364). The remaining 45 MW was verified for Q2 2022 across five restatements (14227, 14196, 14055, 14030, 14009), building on 55 MW reported online in February 2022 (14008). The long-dated targets behave differently. The 800 MW and 1.8 GW deadlines passed in 2023 and 2024 but still sit at not_yet_verifiable, and the supplied record contains no acknowledgement of whether they were met or withdrawn. The Bitmain 200 PH order (14504, 14416, 14378, 14360) is in the same state. The 200 MW Garden City wind letter of intent (14236, 14204, 14074, 14015) led to a project now marked sold, and the Antpool joint venture does not reappear after April 2022. Both point to quiet pivots rather than stated revisions.

The record contains no permits, interconnection queue entries, or site reporting. The only independent corroboration is the verified tags on Jamestown, so the pace of the gigawatt-scale campaign has no outside support here.

Over the next 90 days, watch the fiscal first-quarter disclosures for megawatt figures on the energized portion of Polaris Forge 1 and any tenant-linked delivery schedule. Look for state siting or interconnection filings that put numbers on Polaris Forge 2, Polaris Forge 3 and the Base Electron gas plant. Delta Forge 1 and 2 need a first public capacity commitment. Finally, check whether the company ever reconciles the elapsed 2023 and 2024 targets.

Applied Digital's capacity narrative escalated far faster than its delivered footprint. The early record is anchored to a single 100 MW energy services agreement at Jamestown (14511, 14413) and an 85 MW second site (14375). Within seven months the company was pointing to 500 MW of expansion demand (14512), then an 800 MW target by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), and 5 GW over five years (14230, 14033). Layered on top were a 1.5 GW Antpool joint venture (14212, 14235), whose stated formation figure of 8,000 USD (14234) is hard to square with that scale, and 600 MW-plus of additional sites promised before the end of 2022 (19161).

Near-term dates at Jamestown held. The first co-hosting facility met its Q4 2021 target (14509, 14414), 55 MW was reported online by February 2022 (14008), and the remaining 45 MW landed in Q2 2022 as promised (14227, 14009). Beyond that one site, the record weakens. The Texas 200 MW wind facility moved from a letter of intent (14236) to an operational-by-late-2022 promise (19159) that is marked delayed, and Garden City now appears as sold. The 800 MW and 1.8 GW deadlines passed long ago and are still logged as not yet verifiable rather than met or missed. The same is true of the 200 PH Bitmain order due by the end of 2021, which was restated four times (14504, 14416, 14378, 14360). None of these entries shows a later acknowledgement of revision. That pattern is a reconciliation gap, not proof of a miss, but it should be closed. Claims 21053 through 21069, dated 2006 to 2008, appear to belong to a predecessor shell and should be separated from this trajectory.

The record contains no independent evidence, including permits, interconnection queue positions, or site reporting, so the announced pace cannot be corroborated here. The supplied claim history also ends mid-record (3124), before the pivot to Polaris Forge, Delta Forge, Finland and the Base Electron gas plant. This read therefore does not judge the current AI build-out.

In the next 90 days, the first priority is the next quarterly report. Watch whether it states energized megawatts that can be tested against the lapsed 800 MW and 1.8 GW targets. Watch also whether Polaris Forge 1 moves from partially to fully operational. Interconnection and air-permit filings for Polaris Forge 3 and the adjacent gas plant in Oliver County are the clearest independent markers of real momentum, as is any permitting activity at Boyce and Brookwood.

The supplied record for Applied Digital is cut off in May 2022, in the middle of a claim. This read therefore covers the crypto-hosting era that seeded the Polaris Forge and Delta Forge pipeline, not the AI campuses themselves. The 2006 and 2008 entries (21053, 21054, 21067, 21068, 21069) predate the current business and look like inherited shell-company filings. They should carry no weight here.

Stated ambition grew far faster than contracted power. In August 2021 the company described 500 MW of demand against one 100 MW energy services agreement (14512, 14511). By March 2022 the goals had become 800 MW by May 2023, 1.8 GW by May 2024 and 5 GW within five years (14228, 14229, 14230), plus up to 1.5 GW through the Antpool joint venture (14235). By May 2022 the company was also promising more than 600 MW under construction across up to three sites by year-end (21467). That is roughly a tenfold expansion in nine months. Over the same period, contracted power stayed at Jamestown's 100 MW plus an 85 MW second site (14375).

The narrow, site-specific promises held. Jamestown's first phase was completed by end-2021 as promised (14364), 55 MW was online by February 2022 (14008), and the remaining 45 MW arrived in the promised Q2 2022 window (14227, 14009). The first Texas wind site was verified for 2022 (24806), though Garden City has since been sold. The portfolio targets fared worse. The 800 MW and 1.8 GW deadlines passed in 2023 and 2024 and are still unverified rather than restated. Neither the Antpool venture nor the 200 PH Bitmain order (14378) has a resolution on record. These are revisions by silence, not acknowledged resets.

The record contains no permits, interconnection-queue entries or on-the-ground reporting. Independent corroboration of the current pace therefore cannot be assessed from this data. The only verifications concern Jamestown and the first Texas site.

Over the next 90 days, the fiscal first-quarter report, normally due in October, should show how many megawatts are energized at Polaris Forge 1, which is still only partially operational. It should also show whether Polaris Forge 2 at Harwood is on its stated schedule. North Dakota PSC siting filings for Polaris Forge 3 and the adjacent Base Electron gas plant will be key third-party tests, as will Louisiana and Alabama utility or queue filings for the Delta Forge sites. Together they will show whether the current pipeline behaves like Jamestown or like the 2022 gigawatt targets.

The record supplied here ends partway through May 2022, so this read covers the bitcoin-hosting era rather than the Polaris Forge and Delta Forge AI-campus buildout that now defines Applied Digital. The 2006 and 2008 entries (21053, 21054, 21067 through 21069) predate the current business and say nothing about delivery capacity. Within the usable window, announced capacity grew quickly. In mid-to-late 2021 the company framed itself around a single 100 MW energy services agreement at Jamestown (14511, 14413, 14361) and an 85 MW second site (14375, 14362, 14040). It also cited 500 MW of expansion demand (14512). By March 2022 the ambition had moved to 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), and 5 GW over five years (14230, 14033). The 1.21 Gigawatts joint venture with Antpool was layered on top at up to 1.5 GW (14235, 14212, 26007). These large figures are pipeline targets, not committed builds. The one discrete expansion project, the 200 MW Garden City wind site (14236, 14204), was later marked delayed against its 2022 operating target (3123). The project list now shows it as sold. Near-term site deliveries held. The first Jamestown co-hosting facility met its end-of-2021 target across every restatement (14509, 14414, 14372, 14364). The remaining 45 MW reached its Q2 2022 deadline as promised (14227, 14196, 14069, 14055, 14009, 14030). The aggregate targets are a different story. The 800 MW and 1.8 GW deadlines have both passed, yet they still sit at not_yet_verifiable. Nothing in this record shows the company reconciling them or restating them with an acknowledgement. The Bitmain 200 PH equipment order (14504, 14416, 14378, 14360) is in the same unresolved state. The record contains no independent corroboration of any kind, such as permits, interconnection queue positions, or site reporting. Every capacity figure rests on company statements alone. Over the next 90 days, the priority is to backfill the post-2022 claims so the Polaris Forge 1 ramp can be assessed against its own stated schedule. Polaris Forge 2 and 3 and both Delta Forge sites are listed as under construction, and each should be checked against utility and regional transmission organization interconnection filings. The announced Base Electron gas plant next to Polaris Forge 3 should be tracked through North Dakota Public Service Commission siting dockets. The question of whether the expired 2023 and 2024 megawatt targets were ever formally walked back remains worth resolving.

The claim record supplied here stops in mid-2022 at claim 19160, so it covers only Applied Digital's crypto-hosting era. Nothing in it speaks to the Polaris Forge, Delta Forge, Finland or Base Electron projects listed above, and the read below should be taken in that light. The 2006 and 2008 entries (21053, 21054, 21067 to 21069) describe airport testing and transfer-agent work that bear no relation to this business and look like an entity-resolution error worth purging. The announced capacity curve inflated rapidly over roughly seven months. In 2021 the company framed itself around a single 100 MW energy services agreement at Jamestown (14511, 14413) plus an 85 MW second site (14375, 14040), with 500 MW of loosely defined expansion demand (14512). By March 2022 the ambition had jumped to 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), 5 GW within five years (14230) and an Antpool joint venture of up to 1.5 GW (14235, 14203). None of these targets is tied to a named site, and the 1.21 Gigawatts JV never appears as a project on record. Near-term, site-specific dates held. Jamestown's first phase was verified against its Q4 2021 target (14414, 14364), and the remaining 45 MW was verified for Q2 2022 (14227, 14009). Everything further out tells a different story. The Bitmain 200 PH delivery (14504, 14416, 14360) was never confirmed. The Texas 200 MW wind facility moved from LOI (14236) to delayed (3123) and eventually to sold, and nothing in the record acknowledges that it was abandoned. The 800 MW and 1.8 GW deadlines passed long ago yet still carry not_yet_verifiable tags, which is a ledger gap rather than an open question. Independent corroboration exists only for Jamestown. The record contains no permit, interconnection-queue or site-level reporting behind the gigawatt-scale targets. Over the next 90 days the priority is to backfill the post-2022 claims so that the HPC-era pivot can be assessed on the same footing. Within that window, watch for any written-off or restated version of the 5 GW target ahead of its March 2027 horizon (14230). Watch also for evidence that Polaris Forge 1 has moved from partially operational to full, for state utility-commission or air-permit filings tied to the Base Electron gas plant, and for MISO or SPP queue entries supporting Polaris Forge 2 and 3.

The capacity story here is one of rapid narrative inflation against a narrow contracted base. Through late 2021 the record is anchored to a single 100 MW energy services agreement at Jamestown (14511, 14413, 14373) and an 85 MW second site (14375, 14362, 14040). By March 2022 the ambition had jumped to 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), 5 GW over five years (14230, 14033), a 1.5 GW Antpool joint venture (14235, 14203, 14036) and 600 MW of additional sites by end-2022 (19161). That is a step from roughly 185 MW of identifiable agreements to multi-gigawatt aspirations in about seven months, with no new site-level contracts behind the larger figures. The 2006 and 2008 entries (21053, 21054, 21067-21069) predate the datacenter business and look like shell-era or mismatched records; they should be excluded from any delivery assessment.

On dates, the pattern splits cleanly. Near-term, utility-backed Jamestown milestones held: the first facility by Q4 2021 (14509, 14414, 14372, 14364) and the remaining 45 MW by Q2 2022 (14227, 14069, 14055, 14030, 14009) are both verified. The portfolio-scale targets have not been resolved. The 800 MW and 1.8 GW deadlines passed in 2023 and 2024 but still carry not_yet_verifiable status, as do the 200 PH Bitmain order (14504, 14416, 14378, 14360) and equipment deliveries (14510, 14365). Garden City now shows as sold, and the Antpool joint venture does not appear among current projects. Within this record, the shift from crypto hosting to the Polaris and Delta Forge AI campuses happened without any explicit retirement of the 2022 gigawatt targets.

The record contains no independent corroboration. There are no permits, interconnection queue positions or third-party site reporting. Even the 55 MW online figure (14008) is self-reported, and the verified grades appear to rest on the company's own later disclosures.

Over the next 90 days, the priority is closing data gaps. The extract ends in May 2022 and holds no capacity, cost or date claims for Polaris Forge 2 and 3, Delta Forge 1 and 2, the Finland campus or the Base Electron gas plant. Those claims should be ingested and the lapsed 2023 and 2024 targets formally graded. Externally, watch MISO and SPP queue entries for Harwood and Oliver County, the air permit filing for the gas plant with North Dakota environmental regulators, and Louisiana and Alabama building and utility filings. Energization disclosures moving Polaris Forge 1 from partial to full operation would be the clearest test of whether the current build pace is real.

The record supplied here stops in May 2022, partway through claim 23466. Nothing in view covers Polaris Forge, Delta Forge, Finland or Base Electron, so this read covers only the crypto-hosting era that preceded them. The 2006 and 2008 entries (21053, 21068) describe a predecessor shell and should be excluded from APLD's delivery record. Within that window, the headline capacity grew far faster than anything physically delivered. The anchor was a single 100 MW energy services agreement at Jamestown (14511, 14413), paired with an 85 MW second site (14375). Within seven months the narrative escalated to 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229, 23466) and 5 GW over five years (14230), plus a 1.5 GW Antpool joint venture (14203, 14036) and a 200 MW Texas wind LOI (14236). Some numbers quietly shrank. The GMR arrangement was framed as a 100 MW services agreement (14502), then restated as a 50 MW hosting contract (14377, 14359) without any explanation in the claims. The Texas LOI resolves to a Garden City project now listed as sold. Promised dates held where the scope was small and local. The first Jamestown facility by end-2021 (14372, 14364) and the remaining 45 MW by Q2 2022 (14227, 14009) are both verified. The more ambitious items were never closed out. The 200 PH Bitmain order (14378), the Q4 2021 equipment delivery (14365) and the 600 MW of additional sites by end-2022 (19161) all remain not_yet_verifiable. So do the 800 MW and 1.8 GW targets, even though their deadlines passed in 2023 and 2024. They should now be graded as missed or met, not left open. On independent evidence, the record holds no permits, interconnection queue positions or site reporting for any claim. Every capacity figure after the initial Jamestown ESA rests on company statements alone. In the next 90 days, the first priority is completing the data. That means ingesting the post-2022 claims and resolving the stale 2023 and 2024 capacity targets against reported energized megawatts. On the current portfolio, watch whether Polaris Forge 1 moves from partially to fully operational on its stated schedule. Watch for queue or utility filings for Polaris Forge 2 and 3 and the Delta Forge sites. And check whether the Base Electron gas plant has a state air permit application on file, which would be the first hard signal that its power supply is real rather than announced.

Applied Digital's facility record is a story of small, verifiable deliveries surrounded by large, unverifiable ambitions. The 2006 and 2008 entries (21053, 21054, 21067-21069) belong to the pre-pivot shell and say nothing about datacenter delivery. The supplied record also breaks off at claim 26005 in May 2022, so this read can't assess the Polaris Forge, Delta Forge or Finland projects. That gap matters more than anything below.

On capacity, the trajectory runs from concrete to expansive. The company opened with a 100 MW energy services agreement at Jamestown (14511, 14413) and an 85 MW second site (14375, 14362). Within months, a 500 MW demand figure (14512) became 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW over five years (14230). Some figures shrank quietly. GMR's commitment fell from 100 MW (14502) to 50 MW (14377) with no reconciliation on record. The 5 GW horizon slid from March 2027 (14230) to May 2027 (26015) because each restatement reset the five-year clock rather than holding a fixed date. The 1.5 GW Antpool joint venture (14235, 14212) and the 200 MW Garden City wind LOI (14236) were repeated across five filings without new detail, and Garden City now shows as sold.

On dates, the near-term Jamestown promises held. The first co-hosting facility by year-end 2021 (14509, 14364) and the remaining 45 MW by Q2 2022 (14227, 14055) are both verified, and 55 MW was online by February 2022 (14008). The longer-dated targets are the problem. The 800 MW and 1.8 GW deadlines have both passed, yet they still sit as not yet verifiable (14228, 14229), and no claim in this record acknowledges a revision. The same is true of the Bitmain 200 PH delivery promised for 2021 (14504, 14365).

Independent corroboration is thin. The record contains no permit filings, interconnection-queue positions or site reporting. The only external check is the verification status on Jamestown milestones, which held up.

Over the next 90 days, three things matter. First, the full post-2022 claim set for Polaris Forge 1, which is partially operational, should show whether energized megawatts match leased megawatts. Second, MISO and SPP queue entries and county permits would test whether Polaris Forge 2 and 3 and the Delta Forge sites are advancing at the announced pace. Third, the company should either reaffirm or retire the 5 GW 2027 target, which on this evidence it is nowhere near.

The record supplied here stops in mid-May 2022, at claim 19161, before any Polaris Forge, Delta Forge or Finland claims appear. This read therefore covers the crypto-hosting era that set Applied Digital's credibility baseline, not the AI campuses now listed as under construction. The 2006 and 2008 items (21053, 21054, 21067, 21068, 21069) belong to a predecessor shell entity and say nothing about datacenter delivery.

Ambition outran footprint quickly. The opening frame was a single 100 MW Jamestown energy agreement (14511, 14413, 14373) with 500 MW of claimed demand (14512). Within seven months that became 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032, 21463) and 5 GW over five years (14230, 14033). An Antpool joint venture for up to 1.5 GW (14235, 14212, 14036) was layered on top. None of these targets is tied to a named site in the record, and none is followed by a revision or retirement.

The near-term, site-specific promises held. First co-hosting completion in Q4 2021 is verified (14509, 14414, 14364). The 55 MW reported online in February 2022 (14008, 14029) was followed by the remaining 45 MW in Q2 2022, also verified (14227, 14009, 14030). The speculative layer is a different story. The Bitmain 200 PH order (14504, 14360, 14378) cannot be verified. The 85 MW second facility (14375, 14362, 14040) never acquires a location. Garden City began as a 200 MW wind letter of intent (14236, 14204, 14034) and now carries a sold status. The September 2022 groundbreaking for a second Texas site (21466) and more than 600 MW under construction by year-end 2022 (21467) have no independent confirmation. The gigawatt targets appear to have lapsed quietly rather than being restated.

For independent evidence, the only verifications in the record are Jamestown energization milestones. The record contains no permit, interconnection-queue or on-the-ground entry for any larger target. That gap matters most for Polaris Forge 2, Polaris Forge 3 and the Delta Forge sites, which this data cannot assess.

Over the next 90 days, the fiscal first-quarter filing should be checked for megawatts actually energized at Polaris Forge 1 versus megawatts contracted. Next is whether MISO, SPP or Southern Company queue positions and county permits exist for Harwood, Oliver County, Boyce and Brookwood at the scale announced. Last is whether the Finland campus gains a named site and grid agreement or stays an announcement. A complete claim export is needed before drawing conclusions on the current portfolio.

The claim record supplied here ends in May 2022, mid-entry, so it covers Applied Digital's crypto-hosting origins but none of the Polaris Forge, Delta Forge or Finland AI campus commitments that now define the company. This read covers that early arc. The newer projects need their own claim history before their delivery pace can be judged. The 2006 and 2008 entries (21053, 21054, 21067, 21068, 21069) appear to belong to a predecessor shell entity and carry no facility signal. They should be filtered from the profile.

On capacity, the company's ambition expanded faster than any evidence of delivery. In 2021 it described a single 100 MW energy services agreement (14511, 14413, 14361) and an 85 MW second site (14375, 14362, 14040). By March 2022 it was claiming 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), 5 GW within five years (14230, 14033, 24805), and an Antpool joint venture of up to 1.5 GW (14235, 14203, 14212, 21457). That is a roughly fifty-fold jump in stated ambition in about four months, while the operating base was still the 100 MW Jamestown site.

The near-term promises held. Jamestown's first co-hosting phase was due in Q4 2021 and is verified (14509, 14414, 14372, 14364). The remaining 45 MW for Q2 2022 is also verified (14227, 14196, 14069, 14055, 14030, 14009), following 55 MW online in February 2022 (14008). The large targets fared worse. The 800 MW and 1.8 GW dates have both passed but remain unresolved. The 600 MW multi-site construction start due by end-2022 (3125) and the 200 PH Bitmain order (14504, 14416, 14378, 14360) are also unverified. None of these is repeated in the visible record after spring 2022, which looks like quiet retirement rather than acknowledged revision. The Antpool joint venture follows the same pattern. The 200 MW Texas wind LOI (14236, 14214, 14034) became Garden City, which is now listed as sold.

The record contains no independent corroboration of any of these claims: no permits, interconnection queue positions or site reporting. The verified tags appear to rest on the company's own later disclosures.

Over the next 90 days, the priority is to backfill 2023–2026 claims for the Forge projects. The next quarterly filing should then be checked against Polaris Forge 1's partial-operational status. MISO and regional utility interconnection records for Harwood and Oliver County, plus Louisiana and Alabama permit filings for Delta Forge, are the obvious independent tests. Finland should stay in announced status until a grid connection agreement surfaces.

The record supplied here runs only from the 2021 hosting pivot through mid-May 2022. It says a great deal about Applied Digital's first North Dakota site and nothing yet about the Polaris Forge, Delta Forge or Finland projects that now define the company. Any read of the current AI-campus pipeline has to wait on those claims, so what follows covers the founding trajectory.

On capacity, the arc shows steep inflation at the portfolio level against steady delivery at the site level. Jamestown was framed at 100 MW from the first filing (14511, 14413), and that figure never moved (14233, 21454). The aggregate ambition, by contrast, jumped within seven months from 500 MW of planned expansion demand (14512) to an 800 MW target by May 2023 (14228), 1.8 GW by May 2024 (14229), and 5 GW over five years (14230). The Antpool joint venture added a further notional 1.5 GW (14235, 14212). Those headline numbers were restated almost verbatim across March and April 2022 (14031, 14032, 14033) without any milestone detail.

On dates, the near-term promises held. Completion of the first co-hosting facility by end-2021 is verified across every restatement (14509, 14414, 14372, 14364). The remaining 45 MW at Jamestown slated for Q2 2022 is likewise verified (14227, 14196, 14009, 14030), with 55 MW reported live in February (14008) and the full 100 MW claimed by May (26005). The Bitmain 200 PH equipment deliveries (14504, 14378) remain unresolved. More significant, the 800 MW and 1.8 GW deadlines have both passed while still sitting as not_yet_verifiable. That is a resolution gap in our own record, not evidence of delivery, and neither target appears to have been formally revised or withdrawn in the claims shown.

Independent corroboration in this extract is limited to the verification flags themselves. No permits, interconnection queue entries or site reporting are attached. The Texas 200 MW wind LOI (14236, 14063) never progressed beyond letter-of-intent language in the record, and Garden City is now marked sold.

Over the next 90 days, the first priority is to close out 14228 and 14229 against reported energized capacity. The second is to ingest the post-2022 claims for Polaris Forge 1 through 3 and the Delta Forge sites, so that tenant-lease megawatts can be checked against interconnection filings and county permits. Finally, watch whether the Finland campus moves from announcement to any grid or land filing.

The record supplied for this read effectively ends in mid-May 2022, with the final entry (claim 26015) cut off. The current project slate, Polaris Forge 1 through 3 and Delta Forge 1 and 2, has no capacity or timeline claims attached in what is shown. This read therefore covers the crypto-hosting era that preceded those projects and cannot assess them directly. The 2006 and 2008 entries (21053, 21054, 21067, 21068, 21069) describe an airport-testing contractor and shell-company housekeeping. They look like a predecessor-entity or matching artifact rather than Applied Digital facility history, and should be excluded from any trajectory scoring.

Within the 2021-2022 window the pattern is clear. Near-term site promises were kept while the headline ambition inflated quickly. The first Jamestown co-hosting facility was promised for year-end 2021 and is verified (14509, 14414, 14372, 14364). The remaining 45 MW there was promised for Q2 2022 and is also verified (14227, 14196, 14069, 14055, 14030, 14009), following 55 MW online in February 2022 (14008). Over the same months the stated pipeline grew from 100 MW (14511) and 500 MW of expansion demand (14512) to 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), and 5 GW over five years (14230, 14033, 24805). On top of that sat a 1.5 GW Antpool joint venture repeated across roughly ten filings (14235, 14203, 14212, 14073, 14059, 14036, 14016, 21457, 24798).

The 800 MW and 1.8 GW targets are still tagged not_yet_verifiable even though both deadlines passed years ago. That is a resolution gap in our own data, not a neutral status. The Antpool venture and the 200 MW Garden City wind LOI (14236, 14204, 14034) stop recurring. Garden City is now listed as sold. Nothing in the supplied record shows the company formally retiring either commitment, which is the textbook shape of an unacknowledged revision. The 600 MW multi-site construction start promised for end-2022 (3125) and the Bitmain equipment deliveries (14504, 14365) are likewise unresolved.

The supplied record contains no independent evidence, meaning no permits, interconnection-queue entries, or site reporting, so no announced pace here can be corroborated externally. Over the next 90 days the priorities are to ingest the post-2022 claim history and close out 14228, 14229, and 3125 against reported energized megawatts. The read should then test Polaris Forge 1's phased energization against North Dakota interconnection filings and look for queue or permit records behind the Polaris Forge 3 and Delta Forge sites before treating their under-construction status as confirmed.

The supplied record stops at May 2022 (claim 26005 is cut off mid-entry), so this read covers the crypto-hosting era and cannot yet assess Polaris Forge 2 and 3, Delta Forge 1 and 2, or the Finland campus. The 2006 and 2008 entries (21053, 21054, 21067-21069) describe airport testing and a transfer-agent change unrelated to datacenters, and look like a predecessor-entity or resolution error that should be split out before publication. On scale, the ambition rose sharply over seven months. In August 2021 the company described 500 MW of expansion demand on a single 100 MW energy services agreement (14511, 14512). By March 2022 that had become 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032, 23466), 5 GW over five years (14230, 14033), and an Antpool joint venture of up to 1.5 GW (14235, 14203, 14036). The GMR hosting commitment moved the other way, from 100 MW (14502) to 50 MW (14377, 14359), and later filings never explain the change. On dates, the near-term Jamestown promises held. The first co-hosting facility by year-end 2021 is verified (14509, 14414, 14372, 14364), 55 MW was online by February 2022 (14008, 14054), and the remaining 45 MW for Q2 2022 is verified across six restatements (14227, 14196, 14069, 14055, 14030, 14009). The Garden City 200 MW wind site by end-2022 is also marked verified (21465, 24806), although that asset now shows as sold. The 800 MW and 1.8 GW targets and the Bitmain 200 PH delivery (14504, 14510, 14365) are past their dates but still tagged not_yet_verifiable. Those tags should be resolved, because the record cannot currently show whether those promises lapsed without acknowledgement. Independent corroboration is thin. Nothing in this record cites a permit, an interconnection queue position, or site reporting, so even the verified tags rest on company disclosure, and the 8000 USD JV figure (14234) reads like an extraction fault. For the next 90 days, the fiscal Q1 FY2027 10-Q due around mid-October is the main test. Watch whether it reconciles energized megawatts at Ellendale and Polaris Forge 1 against lease commitments, and whether the Harwood and Oliver County projects show up as MISO or SPP interconnection entries. Louisiana and Alabama utility filings for the Delta Forge sites are worth checking too. Finally, rerun the claim extraction past May 2022 so the AI-campus era can be judged on the same footing.

Applied Digital's capacity story inflated far faster than its delivered megawatts. Through late 2021 the record describes a modest footprint: a 100 MW energy services agreement at Jamestown (14511, 14413, 14361) and an 85 MW second facility (14375, 14362). The one larger figure was a 500 MW demand estimate (14512). By March 2022 the ambitions had grown by an order of magnitude: 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), 5 GW within five years (14230), and an Antpool joint venture for up to 1.5 GW (14235). That joint venture's stated capitalization was 8,000 USD (14234). When these targets were set, the only capacity confirmed as online was 55 MW at Jamestown, energized February 2, 2022 (14008, 14054).

Delivery dates have not held, and the company has tended to repeat them rather than revise them. Completion of the first facility by Q4 2021 was restated unchanged in August and twice in November 2021 (14509, 14414, 14372, 14364), and that date was missed. The remaining 45 MW at Jamestown was promised for Q2 2022 in at least six filings between March and April (14227, 14206, 14069, 14055, 14009, 14030), and it also slipped. The 800 MW and 1.8 GW targets are both marked delayed (14031, 14032), and the record contains no explicit retraction or re-baselining. By July 2022 the narrative had moved from expansion to restoring permanent power at Jamestown by mid-2023 (13857), and that date was missed too. This suggests the flagship site was not running at nameplate capacity for an extended period.

The supplied record contains no independent corroboration: no permits, no interconnection queue positions, no site reporting. Garden City stayed a letter of intent across four consecutive filings (14236, 14204, 14074, 14015) with no move to a definitive agreement in the data. Claims 21053 through 21069 (2006–2008) appear to belong to a predecessor shell entity. They should be excluded from this project history.

The record breaks off at claim 13856 in mid-2022. As a result, Polaris Forge 1 through 3, Delta Forge 1 and 2, and the Finland campus have no capacity or timeline claims here, and they cannot yet be assessed against the pattern above. Over the next 90 days, we will watch for three things. First, whether those project-level claims carry dated energization commitments. Second, whether interconnection filings with MISO and SPP, or county permits, for Harwood, Oliver County, Boyce, and Brookwood keep pace with announcements. Third, whether the 5 GW target due March 2027 (14230) is restated, quietly replaced, or formally withdrawn.

The record supplied ends mid-sentence at claim 13855 in July 2022, so this read covers Applied Digital's crypto-hosting era and cannot yet assess the Polaris Forge, Delta Forge, or Finland projects listed on file. Claims 21053, 21054, and 21067 through 21069 date from 2006 and 2008 and appear to belong to a predecessor shell entity. They should be excluded from any delivery analysis of the current business.

Announced ambition grew quickly, while delivered capacity grew slowly. In August 2021 the company described a 100 MW energy services agreement and 500 MW of expansion demand (14511, 14512). By March 2022 that had become 800 MW by May 2023, 1.8 GW by May 2024, and 5 GW over five years (14228, 14229, 14230). It was layered on a 1.5 GW Antpool joint venture (14235) and a 200 MW Texas wind LOI at Garden City (14236). Both deadlines have since lapsed as delayed, and neither the Antpool vehicle nor the Garden City LOI advances past stated status in later restatements (14059, 14074, 14034).

On timelines, the pattern at Jamestown is restatement without revision. The first facility was promised for Q4 2021 in four separate filings (14509, 14414, 14372, 14364) and missed. Partial delivery did arrive: 55 MW was reported online as of February 2022 (14008, 14054). The remaining 45 MW was then pinned to Q2 2022 and repeated unchanged across at least six filings (14227, 14196, 14069, 14055, 14030, 14009) before slipping. The July 2022 power disruption produced fresh dates for temporary and permanent power restoration (13856, 13857), and both were also missed. The record shows dates being carried forward rather than reset with explanation.

Independent corroboration is absent from what was provided. There are no permits, interconnection queue positions, utility filings, or site reporting attached to any of these claims, and every project row lacks capacity and cost fields. The announced pace therefore rests entirely on company statements.

Over the next 90 days, the priority is ingesting the post-2022 claim history so the AI-campus pivot can be measured against this baseline. Beyond that, watch for interconnection or large-load filings tied to Harwood, Boyce, and Brookwood, and for county building permits that would confirm construction phasing. Also watch whether the next quarterly report restates, revises, or quietly drops the five-year 5 GW target (14230), which falls due in March 2027.

The claim record supplied here ends abruptly in July 2022, partway through claim 13856, so this read covers only the bitcoin-hosting era. None of the Polaris Forge, Delta Forge, Garden City build-out or Finland projects listed on the record has a dated capacity or timeline claim in the excerpt, and no conclusions about the current AI-campus pipeline should be drawn from it. The 2006 and 2008 entries (claims 21053, 21054, 21067-21069) describe airport testing and shell-company auditor and transfer-agent appointments. They appear to belong to a predecessor entity and should be separated before any scoring. The Ellendale and Polaris Forge 1 rows share an address and likely describe one site.

Within the window that is present, ambition scaled much faster than delivery. The August 2021 framing was a 100 MW services agreement against 500 MW of stated demand (claims 14511, 14512). By March 2022 this had become an 800 MW target for May 2023, 1.8 GW for May 2024 and 5 GW over five years (claims 14228, 14229, 14230). Alongside these sat a 1.5 GW joint-venture ceiling with Antpool (claim 14235) and a 200 MW Texas wind letter of intent (claim 14236). The 800 MW and 1.8 GW targets are both marked delayed, and the 5 GW horizon (2027) remains open.

The Jamestown dates show a recurring pattern. First-facility completion promised for Q4 2021 (claims 14509, 14414, 14372, 14364) became 55 MW online by February 2022 with the remaining 45 MW pushed to Q2 2022 (claims 14008, 14009). That Q2 date was restated almost verbatim across at least six filings without acknowledging the earlier slip, and it too is marked delayed. Claim 13857 then refers to restoring permanent power at Jamestown by mid-2023. This implies an operational setback that the earlier capacity claims never anticipated, and that date also slipped.

The supplied material contains no independent corroboration: no permits, no MISO or ERCOT interconnection queue positions, and no site reporting. The Garden City letter of intent in particular never progresses beyond intent within this record.

For the next 90 days, the priority is ingesting the post-July 2022 claims so that the Polaris Forge capacity and energization dates can be tracked. The working question is whether the 2022 habit of restating dates without acknowledgement persists into the AI-campus era. That should be checked against MISO queue filings for the North Dakota sites, local permitting for Harwood, Boyce and Brookwood, and any lease-commencement disclosures tied to named tenants.

The record supplied here stops partway through July 2022, with claim 13856 cut off, so this read covers Applied Digital's crypto-hosting era only. It cannot assess Polaris Forge, Delta Forge or the Finland campus, none of which carries a dated capacity or timeline claim in this extract. The 2006 and 2008 entries (21053, 21054, 21067 through 21069) concern airport testing and auditor appointments. They appear to belong to a predecessor entity and should be set aside. On capacity, ambition outran delivery quickly. In August 2021 the company framed a 100 MW energy services agreement against 500 MW of demand (14511, 14512). By November it had added an 85 MW second site (14375). By March 2022 it had moved to a 5 GW five-year target (14230), interim milestones of 800 MW by May 2023 and 1.8 GW by May 2024 (14228, 14229), a 1.5 GW Antpool joint venture (14235) and a 200 MW Garden City wind letter of intent (14236). The verifiable operating base at that point was 55 MW at Jamestown (14008), and both interim targets are now flagged delayed. On dates, the Jamestown completion promise of Q4 2021 was repeated unchanged across four filings (14509, 14414, 14372, 14364) and then missed. It was recast as 55 MW live with the remaining 45 MW due in Q2 2022 (14227, 14196, 14009, 14055), which also slipped. By July 2022 a permanent power restoration target of June 2023 appears (13857). That implies a disruption that earlier claims never flagged, which is a revision without visible acknowledgement. The Garden City letter of intent was restated near-verbatim through April 2022 with no date or progress marker. No independent evidence appears in this extract: no permits, no interconnection queue entries, no on-the-ground reporting. The Antpool joint venture and Garden City are the least substantiated items, and the 8,000 USD figure attached to the joint venture's formation (14234) looks like either a nominal capitalisation or a parsing artifact, so it should be checked against the source filing. Over the next 90 days, the priority is to backfill the post-2022 claim history. After that, check whether the 5 GW by 2027 target is still cited or has been quietly retired, and match any Ellendale or Harwood capacity figures against utility interconnection filings and county permits. Finally, see whether Jamestown and Garden City still appear in current disclosures at all, because silent disappearance would complete the pattern of revision without acknowledgement seen here.

The extract supplied for Applied Digital ends abruptly in July 2022, partway through claim 13855. None of the Polaris Forge, Delta Forge or Finland projects on record has a single attached claim. This read therefore covers only the company's crypto-hosting origin story, not its current AI-campus pipeline. The 2006 and 2008 items (21053, 21054, 21067-21069) concern airport testing, auditors and transfer agents. They look like predecessor-shell filings and say nothing about facility delivery.

Within that window, announced capacity inflated quickly. In mid-2021 the company described a 100 MW energy services agreement at Jamestown (14511, 14413) and an 85 MW second site (14375, 14362). By March 2022 the ambition had jumped to 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), 5 GW over five years (14230), a 1.5 GW Antpool joint venture (14235) and a 200 MW Texas wind letter of intent tied to Garden City (14236). These figures were restated nearly verbatim across later filings (14031, 14032, 14033, 14036). The 800 MW and 1.8 GW targets are now marked delayed. The joint venture and the Texas letter of intent never progress beyond stated in this record.

Dates have not held. The first co-hosting facility was promised for Q4 2021 across four separate filings (14509, 14414, 14372, 14364), and all are marked delayed. Jamestown did report 55 MW online by February 2022 (14008). The remaining 45 MW, guided to Q2 2022 in at least six claims (14227, 14196, 14069, 14055, 14009, 14030), also slipped. By July 2022 the narrative had shifted to power restoration: temporary power by September 2022 (13856) and permanent power by mid-2023 (13857), both delayed. The earlier filings carry no acknowledgement that an outage was coming.

The record contains no permits, interconnection queue entries or site reporting. None of the announced pace can be independently corroborated here. That gap matters most for the Ellendale and Polaris Forge 1 rows, which share an address and may be duplicates.

Over the next 90 days, the first priority is to recover the truncated claims and confirm whether any Polaris or Delta Forge capacity figures exist. Next, check MISO queue positions for Ellendale and Harwood and the Louisiana and Alabama utility filings for Boyce and Brookwood. Finally, watch whether the 5 GW target, due March 2027, is reaffirmed, resized or quietly retired.

The supplied record cuts off at claim 13855 in July 2022. This read therefore covers only Applied Digital's crypto-hosting era. Polaris Forge, Delta Forge and the Finland campus have no capacity, cost or dated claims in what was supplied. The 2006 and 2008 entries (21053, 21054, 21067-21069) cover airport testing, a Volpe contract and the appointment of an auditor and transfer agent. They appear to belong to a predecessor shell entity and should carry no weight in judging datacenter delivery.

On scale, the company's ambitions grew much faster than its operating base. The August 2021 filings rested on a 100 MW energy services agreement at Jamestown (14511, 14413) against 500 MW of claimed expansion demand (14512). By March 2022 the company had added several larger targets: 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), 5 GW over five years (14230), and up to 1.5 GW through the Antpool joint venture (14235). It also cited a 200 MW Garden City wind letter of intent (14236), which is still only a letter of intent in every restatement through April 2022 (14074, 14063, 14034). Both the 800 MW and 1.8 GW targets are now marked delayed.

Dates have slipped repeatedly, mostly through quiet restatement rather than acknowledgement. The Q4 2021 Jamestown completion (14509, 14414, 14372, 14364) became 55 MW online as of February 2022 (14008). The remaining 45 MW was pushed to Q2 2022 (14227, 14196, 14069, 14009), and that date was repeated across five filings with no explanation for the change. That deadline also passed. By July 2022 the site was waiting on temporary power restoration by September 2022 (13856) and permanent restoration by mid-2023 (13857), and both are now delayed. The 85 MW second facility (14375, 14040) never received a date at all.

No independent evidence is attached. There are no permits, MISO or SPP queue positions, utility filings or on-site reporting to support this pace, so every capacity figure here rests on the company's word alone.

Over the next 90 days, the first job is to complete the record. Pull the post-2022 claims for Polaris Forge 1 and 2 and Delta Forge. Check whether Ellendale and Polaris Forge 1, which share an address, are being double-counted. Then compare any announced energization dates against North Dakota PSC dockets, interconnection queue status and tenant lease disclosures. Until that is done, the current pipeline deserves no more trust than the 2022 targets did.

The visible record stops mid-2022 at claim 13857, so this read covers Applied Digital's crypto-hosting origin story and cannot yet speak to the Polaris Forge, Delta Forge or Finland projects, none of which carry a single capacity or timeline claim here. The 2006 and 2008 entries (21053, 21054, 21067-21069) concern airport testing, transfer agents and auditors and appear to belong to a predecessor shell entity; they should be excluded from any delivery assessment. What remains shows a company whose ambition scaled far faster than its delivery. In August 2021 the anchor was a single 100 MW energy services agreement at Jamestown (14511, 14413), with 500 MW of expansion demand floated alongside it (14512). By March 2022 that had become 800 MW by May 2023, 1.8 GW by May 2024 and 5 GW within five years (14228, 14229, 14230), plus a 1.5 GW Antpool joint venture (14235) and a 200 MW Texas wind letter of intent at Garden City (14236). Both dated targets are now marked delayed, and the 5 GW five-year horizon falls in March 2027 with no intermediate milestones on record. Dates have slipped repeatedly and quietly. The first co-hosting facility was promised for year-end 2021 across four separate filings (14509, 14414, 14372, 14364), all delayed. Jamestown reached 55 MW by February 2022 (14008), but the remaining 45 MW was restated for Q2 2022 at least six times between March and April (14227, 14206, 14196, 14069, 14055, 14009) without acknowledging that the original 100 MW year-end promise had already been missed, and a July 2022 power-restoration target (13856) also slipped. The repetition of identical language across filings is the pattern to note: the company reissued targets rather than revising them. There is no independent corroboration in this dataset, meaning no permits, no interconnection queue entries and no site reporting for any project, so the announced pace rests entirely on company statements. Over the next 90 days the priorities are to load the post-July-2022 claims so the AI-era Ellendale and Polaris Forge commitments can be tested against the same standard, to pull MISO and SPP queue positions for Ellendale, Harwood and Boyce, and to check county permit filings for Brookwood, since the record currently carries no evidence that would distinguish those campuses from the 2022 targets that slid.

Applied Digital's capacity story in this record is one of headline ambition outrunning site-level delivery. The earliest operational anchor is a single 100 MW energy services agreement at Jamestown (14511, 14413, 14361), restated almost verbatim across a dozen filings. By March 2022 the ambition had jumped to 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW over five years (14230). Alongside those targets sat a 1.5 GW Antpool joint venture (14235, 14212) and a 200 MW Garden City wind letter of intent (14236). The Garden City claim never moves past letter-of-intent language in any later claim. The aggregate targets are now tagged delayed, and no claim on record resizes or withdraws them. They simply stop being repeated.

Dates have slipped repeatedly and quietly. The first co-hosting facility was due by the end of 2021 (14509, 14414, 14364). By April 2022 the company reported 55 MW energized as of February (14008), with the remaining 45 MW pushed to Q2 2022 (14227, 14009), a date that was itself later marked delayed. The July 2022 power-restoration disclosures (13856, 13857) set a temporary-power date of September 2022 and permanent restoration by mid-2023, and both were also missed. The pattern is re-dating without explicitly acknowledging the prior miss.

Independent corroboration is absent from this record. No permit, interconnection queue position or on-the-ground report is attached to any claim, so the announced pace rests entirely on company statements. The 85 MW second facility (14375, 14040) is never tied to a location.

Two caveats matter. The 2006 and 2008 entries (21053, 21054, 21067-21069) concern a predecessor entity and should not be read as Applied Digital delivery history. More importantly, the claim list is truncated at mid-2022. None of the current flagship projects, Polaris Forge 1-3, Delta Forge 1-2 and the Finland campus, has a capacity, cost or timeline claim on record, so this read cannot assess the HPC pivot itself.

Over the next 90 days, watch whether Polaris Forge and Delta Forge filings commit to specific megawatts and energization quarters. Check whether MISO queue entries, utility interconnection filings and county permits for Ellendale, Harwood, Boyce and Brookwood surface to match those commitments. Finally, watch whether Garden City and the 5 GW target are formally retired or quietly revived.

The record supplied here stops mid-entry in July 2022, so this read covers Applied Digital's crypto-hosting origins and not the Polaris Forge, Delta Forge or Finland AI campuses that now dominate its project list. Those projects carry no capacity, cost or date claims in this extract, and any judgment on them would be speculation. The 2006 and 2008 entries (21053, 21054, 21067-21069) belong to the predecessor shell entity and say nothing about facility delivery. Across 2021-2022, announced capacity scaled far faster than anything visibly built. The first-site anchor of 100 MW (14511, 14413, 14361) was real enough to report 55 MW energized by February 2022 (14008, 14054). Over the same months, though, the stated ambition jumped from 500 MW of expansion demand (14512) to an 800 MW target for May 2023 (14228, 14031), 1.8 GW for May 2024 (14229, 14032), and 5 GW over five years (14230, 14033). That last figure was restated in April 2022 without its 2027 end date (14033 versus 14230), a quiet softening of the commitment. The 1.5 GW Antpool joint venture (14212, 14036) and the 200 MW Garden City wind LOI (14236, 14204) were repeated release after release with no milestone attached. Dates have slipped almost uniformly. The Q4 2021 Jamestown completion was reaffirmed four times through 30 November 2021 (14509, 14414, 14372, 14364) and missed. The replacement Q2 2022 target for the remaining 45 MW (14227, 14196, 14069, 14009, 14030) was missed in turn. By July 2022 the company was promising restoration of temporary power at Jamestown (13856), which reads as a reversal rather than progress. None of these misses appears to have been acknowledged as a revision; the targets simply rolled forward. Bitmain's 200 PH equipment order (14504, 14416, 14378, 14360) followed the same pattern. The supplied material includes no independent evidence: no permits, no MISO or ERCOT queue positions, no local reporting. The 55 MW figure is therefore company-sourced only. Over the next 90 days, the priority is to load the post-July 2022 claims and test whether the AI-era Polaris Forge capacity figures repeat the 2022 pattern of headline growth outrunning energized megawatts. The specific checks are North Dakota PSC siting filings for Harwood, MISO and SPP interconnection entries for the Ellendale and Harwood sites, and Louisiana and Alabama local permitting for the Delta Forge sites. Lease-commencement disclosures in the next 10-Q are the cleanest test of whether announced megawatts are actually being delivered.

The claims record supplied for Applied Digital is truncated in July 2022 (the last legible entry is claim 13857), so this read covers only the company's crypto-hosting era. It cannot assess the Polaris Forge, Delta Forge or Finland projects listed as announced. The 2006 and 2008 entries (claims 21053, 21054, 21067, 21068, 21069) concern a predecessor shell entity and look like an entity-matching artifact. They should be excluded from any delivery assessment.

On capacity, headline targets grew quickly while only one site had a signed power agreement. The Jamestown energy services agreement held steady at 100 MW from August 2021 through April 2022 (claims 14511, 14413, 14373, 14013), and an 85 MW second site appeared in November 2021 (claims 14375, 14362, 14040). By March 2022 the company had added an 800 MW target for May 2023, 1.8 GW for May 2024 and 5 GW over five years (claims 14228, 14229, 14230). It also announced a 1.5 GW Antpool joint venture and a 200 MW Garden City wind letter of intent (claims 14235, 14236). None of these is tied to a named interconnection or a permitted site in the record. The 200 MW Marathon hosting agreement in July 2022 (claim 13855) was the first large commitment with a named counterparty.

Dates slipped repeatedly, and the misses were not acknowledged. Four successive filings promised the first co-hosting facility by Q4 2021 (claims 14509, 14414, 14372, 14364). By April 2022 the company reported 55 MW online as of February 2, 2022 (claim 14008), presenting partial delivery without addressing the missed full-facility date. The remaining 45 MW was then promised for Q2 2022 (claims 14227, 14206, 14009) and is marked delayed. By July 2022 a mid-2023 target for permanent power restoration at Jamestown had appeared (claim 13857), which implies the site was not running on permanent power. Both the 800 MW and 1.8 GW targets are flagged delayed.

The record contains no independent corroboration of any of this: no permits, no interconnection queue positions, no on-the-ground reporting. Every figure is company-stated.

The priority for the next 90 days is to finish ingesting post-2022 claims so the Ellendale, Harwood and Delta Forge timelines can be judged on their own record. Those claims should be checked against MISO and SPP queue filings and county permits in North Dakota, Louisiana and Alabama. Ellendale and Polaris Forge 1 share an address and are probably one project recorded twice, which should be resolved before any capacity totals are published.

Applied Digital's capacity narrative has grown far faster than its delivered megawatts. The record begins in August 2021 with one utility arrangement for 100 MW at Jamestown (14511, 14413) and an 85 MW second site (14375, 14362). By March 2022 it had become a corporate ladder of 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW within five years (14230), plus a separate joint venture with Antpool for up to 1.5 GW (14235, 14212). The only verified delivery in that period is 55 MW at Jamestown as of February 2022 (14008), roughly half of the first site's agreed power. The Garden City wind LOI at 200 MW (14236) was repeated in at least five filings through April 2022 (14214, 14204, 14074, 14063, 14034), and none of them describes it moving to a definitive agreement. Repetition without progress is the pattern to note here.

Dates have slipped consistently. The first co-hosting facility and the Bitmain 200 PH order were both pinned to year-end 2021 (14509, 14504) and were restated without change through late November (14364, 14360) before missing. The remaining 45 MW at Jamestown, promised for Q2 2022 (14227, 14009), also slipped. The July 2022 claims then introduce a power outage at Jamestown. Temporary restoration was dated September 12, 2022 (13856) and is marked delayed. The 800 MW and 1.8 GW targets lapsed with no revised figure in the record, which reads as silent abandonment rather than formal revision. The 1.21 Gigawatts venture likewise drops out without a stated outcome.

This record contains no independent corroboration. No permit filings, interconnection-queue positions or site reporting are attached, so the stated pace rests entirely on company disclosures. The project list is also notable for what is missing. Polaris Forge 1 through 3 and Delta Forge 1 and 2, which together carry the current AI-hosting thesis, have no capacity or cost recorded, and the claim history supplied here stops mid-2022. That gap limits any judgment about the current build. The 2006 and 2008 entries (21053 through 21069) come from the predecessor shell and have no bearing on the datacenter program.

Over the next 90 days, the first thing to watch is whether Polaris Forge and Delta Forge receive stated megawatt and energization dates that can be checked against MISO and SPP queue entries and county permits in Harwood, Boyce and Brookwood. Second, check whether the 5 GW by March 2027 target (14230) is restated, revised or dropped as it nears. Third, look for any filing that formally closes out Garden City and the Antpool venture.

The capacity narrative in this record inflated quickly and then went quiet. In August 2021 the company described a 100 MW energy services agreement and 500 MW of expansion demand (14511, 14512). Within seven months the targets had become 800 MW by May 2023, 1.8 GW by May 2024 and 5 GW over five years (14228, 14229, 14230), plus an Antpool joint venture sized at up to 1.5 GW (14212, 14235). The physical base did not grow at the same rate. The only delivered capacity on record is the 55 MW reported online at Jamestown as of February 2022 (14008, 14054), against the 100 MW contracted there (14361, 14013). The 200 MW Garden City wind site was repeated across at least six filings (14236, 14204, 14034) and never moved past letter-of-intent language, and no 85 MW second-facility site was ever named (14375, 14040).

Dates have slipped consistently and without acknowledgement. The Jamestown completion target of Q4 2021 (14414, 14364) was carried forward, unchanged, through late November 2021 and is now marked delayed. Jamestown's remaining 45 MW was promised for Q2 2022 in five separate filings (14227, 14196, 14069, 14055, 14009), and each one is delayed. By July 2022 the record shifts from expansion to temporary and permanent power restoration at Jamestown (13856, 13857), which points to an operational interruption rather than a scheduling adjustment. The 800 MW and 1.8 GW milestones both lapsed, and no restatement appears in the record.

The record contains no permits, interconnection queue entries or site reporting that would corroborate any of these figures. It cannot independently confirm that the announced pace was achievable. The 2006 and 2008 entries (21053, 21067) appear to belong to a predecessor shell entity and should be excluded from any trajectory read. The input also cuts off partway through claim 13857, so any post-2022 claims are not reflected here.

The current flagship projects (Polaris Forge 1, 2 and 3, Delta Forge 1 and 2) carry no capacity, cost or date claims in this dataset. Over the next 90 days, the priority is to populate them from recent 10-Q and 8-K filings and to check stated megawatts against MISO and SPP queue positions and North Dakota and Louisiana permit dockets. It will also matter whether the company reconciles today's figures with the lapsed 2022 gigawatt targets or simply leaves them out.

Applied Digital's record is two separate stories, and only the first is documented here. The claims on file run from the 2021 hosting pivot through mid-2022 and stop at the Marathon hosting agreement (claim 13855). The newer Polaris Forge and Delta Forge projects have no capacity, date or cost claims attached. The 2006 and 2008 entries (claims 21053, 21054, 21067) describe a predecessor shell's business and say nothing about this buildout, so they should be set aside.

Capacity grew quickly in announcements. In late 2021 the company described a 100 MW Jamestown energy services agreement (claims 14413, 14361) and an 85 MW second site (claims 14375, 14362). By March 2022 it was promising 800 MW by May 2023 (claim 14228), 1.8 GW by May 2024 (claim 14229) and 5 GW within five years (claim 14230). It added a JV with Antpool for up to 1.5 GW (claim 14235) and a 200 MW Texas wind letter of intent at Garden City (claim 14236). Neither the JV nor the Texas site moved past those terms in any later claim. The 800 MW and 1.8 GW targets are both marked delayed.

Dates slipped repeatedly. The first co-hosting facility was due in Q4 2021 (claims 14414, 14364) and missed. In February 2022, 55 MW was reported online (claim 14008). The remaining 45 MW was promised for Q2 2022 in at least five filings (claims 14227, 14196, 14069, 14009, 14030), and the date never changed even as the quarter ran out. The July 2022 Jamestown power-loss disclosure set new targets for temporary and permanent restoration (claims 13856, 13857), and both of those slipped too. The pattern is that dates carried forward unchanged and were replaced rather than acknowledged as missed.

There is no independent evidence in the record for any project. It contains no permits, no MISO or other interconnection queue entries, and no on-the-ground reporting. Ellendale and Polaris Forge 1 share the same street address, which looks like a rename rather than two sites. It should be reconciled before either one is tracked.

For the next 90 days, watch for: - capacity and energization claims finally being attached to Polaris Forge 1, 2 and 3; - Cass County or North Dakota PSC filings for Harwood; - interconnection or utility filings for Delta Forge in Louisiana and Alabama; - whether the 5 GW by March 2027 target (claim 14230), now about five months from its deadline, is restated, redefined or quietly dropped.

This record has a gap that limits any read of Applied Digital's current delivery. The supplied claims stop in mid-2022, at claim 13855, so the Polaris Forge and Delta Forge projects on the roster have no capacity or timeline claims to assess. The 2006 and 2008 entries (21053, 21054, 21067 through 21069) appear to belong to the pre-merger shell entity and say nothing about datacenter delivery. What remains is the bitcoin-hosting era, and its pattern is still useful context.

Announced capacity scaled sharply in a short window. Through late 2021 the company described two concrete assets: a 100 MW energy services agreement at Jamestown (14413, 14361) and an 85 MW second facility (14375, 14362). By March 2022 the ambition had grown to 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), 5 GW over five years (14230), a 1.5 GW Antpool joint venture (14235), and a 200 MW Texas wind LOI at Garden City (14236). The April 2022 restatement of the 5 GW goal dropped its explicit horizon (14033). Both dated portfolio targets are now marked delayed.

Dates slipped repeatedly, and the slips were generally rolled forward rather than acknowledged. The first co-hosting facility was due in Q4 2021 (14509, 14414, 14372, 14364). By February 2022 only 55 MW was online (14008). The remaining 45 MW was promised for Q2 2022 across at least five filings (14227, 14196, 14069, 14009, 14030). That target was then overtaken by a July 2022 power disruption, with temporary restoration promised by September 2022 (13856) and permanent restoration by June 2023 (13857). Both of those dates are also marked delayed. None of the later claims reconciles the missed Q2 milestone.

No independent corroboration appears in the record: no permits, no interconnection queue positions, and no site reporting. The Garden City LOI is repeated verbatim across six filings (14236, 14204, 14214, 14063, 14015, 14034) without ever being upgraded to a definitive agreement. That repetition is itself a signal.

Over the next 90 days, the priority is to ingest the post-2022 claims for Polaris Forge 1, 2, and 3 and for Delta Forge 1 and 2, and to test them against utility interconnection filings, county building permits, and tenant lease-commencement disclosures. The shared Ellendale and Polaris Forge 1 address suggests a duplicate project entry that should be merged before any trajectory is published.

The record supplied here ends in July 2022, mid-entry at claim 13857, so this read covers only Applied Digital's crypto-hosting phase. None of the Polaris Forge, Delta Forge or Ellendale projects on file has an attached claim, so the AI-campus trajectory cannot be assessed from this dataset. The 2006 and 2008 entries (21053, 21054, 21067-21069), covering a Volpe Center contract, airport testing and proxy-solicitation vendors, describe a predecessor shell, not datacenter activity. They should be excluded or re-tagged before publication.

Within that window, the announced ambition grew quickly while delivery lagged. Through late 2021 the company described a 100 MW first facility at Jamestown (14413, 14373, 14361) and an 85 MW second site (14375, 14362). By March 2022 that had become 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW over five years (14230, 14033). Separately, it announced a 1.5 GW Antpool joint venture (14235, 14203) and a 200 MW Texas wind letter of intent at Garden City (14236, 14204). Both milestone targets are now marked delayed, as is every dated promise in the record.

The date record is consistent: deadlines were repeated rather than formally revised. Jamestown's Q4 2021 completion (14414, 14372, 14364) carried through three filings without change before it passed. The remaining 45 MW was then pinned to Q2 2022 (14227, 14196, 14069, 14030, 14009). The 55 MW reported online as of February 2022 (14008, 14054) is the only figure in the record that reads as delivered rather than promised. By July 2022 the Jamestown site was awaiting temporary power restoration (13856), with permanent restoration pushed to mid-2023 (13857). That is a material operational setback, introduced without explicit reconciliation against earlier schedules. Garden City never progressed beyond letter-of-intent language in any claim.

The record contains no independent corroboration: no utility filings, interconnection queue positions, permits or site reporting. The capacity figures rest entirely on company statements, and the 85 MW second site is never located.

Over the next 90 days, the priority is backfilling post-2022 claims for Ellendale and the Polaris Forge sites. After that, the record should be matched against MISO queue entries and North Dakota permitting records, and checked for any Garden City or Antpool joint-venture disclosure that confirms abandonment or progress.

The record supplied here runs only to mid-2022 and cuts off partway through claim 13855, so this read covers Applied Digital's hosting-era Jamestown and Garden City claims rather than the later Polaris Forge and Delta Forge program. The 2006 and 2008 entries (21053, 21054, 21067-21069) describe transit-testing contracts and proxy logistics that appear to belong to a predecessor shell entity. They should not be read as datacenter commitments.

The capacity trajectory shows rapid escalation that outran delivery. In August 2021 the anchor figure was a 100 MW energy services agreement at the first site (14511, 14413), with a loosely framed 500 MW demand pipeline (14512). By November a second 85 MW site appeared (14375, 14362). By March 2022 the ambition had jumped to 800 MW by May 2023, 1.8 GW by May 2024, and 5 GW over five years (14228, 14229, 14230), alongside a 1.5 GW Antpool joint venture (14235, 14212) and a 200 MW Texas wind LOI at Garden City (14236). Both dated interim targets are marked delayed. The JV and the Garden City LOI were repeated through April 2022 (14016, 14034) without any progression to a definitive agreement or a date.

Dates have slipped consistently, and the slippage was absorbed quietly rather than acknowledged. The first co-hosting facility, promised for end-2021, was restated four times on that same deadline (14509, 14414, 14372, 14364). It then reappeared as 55 MW online with the remaining 45 MW due in Q2 2022 (14008, 14009, 14206), and that target was also missed. The July 2022 power-restoration claims (13856, 13857) push permanent Jamestown power to mid-2023, which implies an operational interruption that earlier capacity statements did not anticipate.

The dataset contains no independent corroboration: no permits, interconnection queue positions, or site reporting. All eight listed projects carry blank capacity and cost fields. On this record, the announced pace rests entirely on company statements and counterparty agreements.

Over the next 90 days, the most useful work is backfilling the post-2022 claims so that Ellendale and Polaris Forge commitments can be tested against this pattern. It is also worth pulling MISO and SPP queue entries and North Dakota PSC filings for the Ellendale and Harwood sites, and checking whether Garden City and the 1.21 Gigawatts JV were ever formally abandoned or simply stopped being mentioned.

The record shows ambitions growing much faster than delivered capacity. In August 2021 Applied Digital described a single 100 MW energy services agreement (14511) alongside 500 MW of planned expansion demand (14512). By November it had added an 85 MW second facility (14375, 14362). In March 2022 the targets jumped to 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW over five years (14230). The company also announced an Antpool joint venture for up to 1.5 GW (14235, 14212) and a 200 MW Texas wind letter of intent tied to Garden City (14236, 14214). The two dated targets are both marked delayed, and the joint venture and Garden City claims never moved beyond stated intent in the filings shown here.

The Jamestown dates set the pattern. The first co-hosting facility was promised for Q4 2021 in at least four separate disclosures (14509, 14414, 14372, 14364). The date was restated each time and never revised, and the target was missed. The same goes for the 200 PH Bitmain equipment order (14504, 14416, 14378, 14360). By February 2022, 55 MW was reported online (14008). The remaining 45 MW was promised for Q2 2022 in seven claims (14227, 14206, 14196, 14069, 14055, 14030, 14009), and that slipped too. After the July 2022 power loss at Jamestown, the company set a September 2022 temporary restoration (13856) and a June 2023 permanent fix (13857). Both are marked delayed. In this history, delays tend to surface as quiet carry-forwards rather than as explicit revisions.

The record contains no independent corroboration: no permits, interconnection queue positions or site reporting. The pace claims therefore rest entirely on company disclosures. There are also data-integrity problems that limit any read. The 2006 and 2008 claims (21053, 21054, 21067-21069) appear to belong to a predecessor entity, not to the datacenter business. Ellendale and Polaris Forge 1 share an address and are probably one site listed twice. The claim list breaks off mid-2022 at the Marathon hosting agreement (13855). None of the Polaris Forge or Delta Forge projects has capacity, cost or a dated claim attached.

Over the next 90 days, the priority is to fill that gap. That means extracting the Polaris Forge and Delta Forge capacity and energization dates from recent filings and calls. It also means checking whether the Polaris Forge 2 (Harwood), Delta Forge 1 (Boyce) and Delta Forge 2 (Brookwood) sites appear in MISO, SPP or Southern Company interconnection queues and in local permitting records. The test is whether this cycle's dates get explicitly revised or, as before, simply repeated until they lapse.

The record supplied here ends abruptly in July 2022, partway through claim 13855, so this read covers Applied Digital's crypto-hosting origins and cannot yet speak to Polaris Forge, Delta Forge, or the later Ellendale and Harwood buildouts. The 2006 and 2008 entries (21053, 21054, 21067-21069) belong to a predecessor shell entity and have no bearing on the datacenter trajectory.

Within the window available, announced capacity grew far faster than delivered capacity. The August 2021 baseline was a single 100 MW energy services agreement at Jamestown (14511, 14413), with a vaguer 500 MW of expansion demand (14512). By November an 85 MW second facility had appeared (14375, 14362). By March 2022 the framing had jumped to 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), 5 GW over five years (14230), a 1.5 GW Antpool joint venture capitalized at a nominal 8,000 USD (14234, 14235), and a 200 MW Texas wind letter of intent at Garden City (14236). That is a roughly fifty-fold expansion of ambition in seven months, resting on one partially energized site.

The dates did not hold. Completion of the first Jamestown facility, promised for Q4 2021 in at least four filings (14509, 14414, 14372, 14364), carried into 2022. By April the company reported 55 MW online as of February 2 (14008, 14054) and re-promised the remaining 45 MW for Q2 2022 (14227, 14206, 14009). That 45 MW was still listed as pending when the July 2022 entries record a Jamestown power loss, with temporary restoration targeted for September 2022 (13856) and permanent restoration pushed to mid-2023 (13857). Both the 800 MW and 1.8 GW targets are flagged delayed, and nothing in the record shows the company formally retiring them. They appear to have quietly fallen away rather than been revised.

The record contains no independent corroboration: no permits, interconnection queue positions, or site reporting for Jamestown, Garden City, or the Antpool venture. The Garden City LOI recurs unchanged across five filings (14204, 14214, 14034, 14063, 14074) without progressing to a definitive agreement.

For the next 90 days, the priority is loading the post-July-2022 claims so Polaris Forge 1 and 2 can be assessed against MISO and SPP queue filings, North Dakota PSC dockets, and county building permits. The key test is whether the HPC-era capacity figures show the same pattern seen here, where the headline number moves well ahead of energized megawatts.

Applied Digital's capacity narrative grew far faster than its contracted base. In August through November 2021 the concrete commitments were a 100 MW energy services agreement at Jamestown (14511, 14413, 14361) and an 85 MW second facility (14375, 14362), with a 500 MW demand figure framed only as expansion interest (14512). By March 2022 the company was citing 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229, 14032) and 5 GW within five years (14230, 14033). Most of that increment rested on non-binding or early-stage vehicles: the Antpool joint venture's 'up to 1.5 GW' (14235, 14212, 14036) and a 200 MW Texas wind letter of intent for Garden City (14236). That LOI is repeated across at least five filings (14214, 14063, 14034, 14015, 14074) without any recorded move to a definitive agreement.

The dates have not held. The Q4 2021 completion target for the first Jamestown facility was restated almost verbatim through late November (14509, 14414, 14372, 14364) and then missed. The next filings did not acknowledge the miss directly. They reframed the project as 55 MW online by February 2022 (14008, 14054) with the remaining 45 MW pushed to Q2 2022 (14227, 14206, 14009), which also slipped. By July 2022 Jamestown had a power problem: temporary restoration was promised by September 2022 (13856) and permanent restoration by mid-2023 (13857), and both are marked delayed. The 800 MW and 1.8 GW milestones were never met. A two-quarter slip on a 100 MW site does not support an eighteen-fold scale-up inside two years.

The supplied record contains no independent corroboration, meaning no permits, interconnection queue positions or site reporting, for any announced pace. The 2006 and 2008 entries (21053, 21054, 21067-21069) appear to belong to a predecessor entity and should not be read as datacenter history. The record also ends mid-2022, after the Marathon 200 MW hosting claim (13855). That leaves the current flagship roster unsupported by dated claims. Polaris Forge 1, 2 and 3, Delta Forge 1 and 2, and the Ellendale campus all appear with no stated capacity or cost.

Over the next 90 days, the priority is to capture dated capacity and energization claims for the Polaris Forge and Delta Forge sites and test them against outside evidence. That means North Dakota PSC and county filings for Ellendale and Harwood, MISO and SPP queue entries, and Louisiana and Alabama permitting records. Given the Jamestown pattern, a further test is whether new targets are restated unchanged across consecutive filings and then quietly reframed rather than revised in the open.

Applied Digital's capacity narrative escalated far faster than its delivered megawatts. The record opens in August 2021 with a 100 MW energy services agreement (14502, 14511) and an aspirational 500 MW of expansion demand (14512). By March 2022 that had become 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), 5 GW within five years (14230), and up to 1.5 GW through the Antpool joint venture (14235, 14212). In the same window, the only capacity described as actually energized was 55 MW at Jamestown as of February 2022 (14008, 14054). Both the 800 MW and 1.8 GW targets are tagged delayed. The 200 MW Garden City wind LOI (14236, 14204, 14074) and the 85 MW second facility (14375, 14362, 14040) were restated across at least five filings without a date, a location for the latter, or any sign of conversion to a definitive agreement.

On dates, the Jamestown record is the clearest test, and it did not hold. Full completion was promised for Q4 2021 in four successive filings (14509, 14414, 14372, 14364). It was then quietly reframed as 55 MW online with the remaining 45 MW due Q2 2022, a target repeated at least seven times through April 2022 (14227, 14206, 14196, 14069, 14055, 14030, 14009). The next Jamestown entries in July 2022 are not a revised 45 MW date but a power-restoration schedule (13856, 13857), with permanent restoration pushed to mid-2023. Both are also tagged delayed. That sequence suggests revision without explicit reconciliation of the earlier target.

No independent evidence appears in this record. It contains no permits, no interconnection queue positions, and no third-party site reporting, so the announced pace rests entirely on company statements. The 2006 and 2008 entries (21053 through 21069) predate the datacenter business and carry no weight here. The supplied claims also end in mid-2022 (13855 is truncated), so the Polaris Forge and Delta Forge projects, which carry no capacity or cost figures on record, cannot yet be assessed against any promise.

Over the next 90 days, the key question is whether filings attach specific megawatts and energization dates to Polaris Forge 1, 2 and 3 and to Delta Forge 1 and 2. A second question is whether those figures can be matched to utility or ISO interconnection filings and county permits for Ellendale, Harwood, Boyce and Brookwood. A third is whether any of the 2022-era targets is formally retired rather than allowed to lapse.

The record supplied here is incomplete. It cuts off mid-entry at claim 13857 in July 2022, and none of the Polaris Forge or Delta Forge projects listed on file has a single attached claim. This read therefore covers Applied Digital's crypto-hosting era rather than its current AI campus pipeline. The 2006 and 2008 entries (21053, 21054, 21067) concern airport testing and auditor appointments under a predecessor shell and have no bearing on facility delivery.

Announced capacity scaled faster than anything that was physically delivered. In August 2021 the company described a 100 MW energy services agreement at Jamestown (14511) alongside a loosely framed 500 MW of expansion demand (14512). By November an 85 MW second facility had been added (14375). In March 2022 the ambition jumped to 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW over five years (14230), stacked on a 1.5 GW Antpool joint venture (14235) and a 200 MW Garden City wind LOI (14236). Both dated targets are now marked delayed. The Garden City LOI was repeated across at least five filings (14204, 14034, 14063, 14015, 14074) without ever acquiring a date or a capacity milestone.

Dates at Jamestown slipped serially, and each slip was handled by restatement rather than acknowledgement. The Q4 2021 completion target was repeated four times (14509, 14414, 14372, 14364) before it lapsed. It was then quietly replaced by a 55 MW online figure as of February 2022 (14008), with the remaining 45 MW pushed to Q2 2022. That new date was restated in at least six documents (14227, 14196, 14206, 14030, 14055, 14069) and was also missed. By July 2022 the narrative had shifted to temporary and permanent power restoration at Jamestown (13856, 13857), which implies an operational setback that the earlier capacity claims never anticipated.

Nothing in this record offers independent corroboration. There are no permits, no utility interconnection or queue references, and no site reporting, so the announced pace rests entirely on company statements.

Over the next 90 days, the priority is closing the data gap. That means ingesting the post-2022 claim history and attaching Polaris Forge 1, 2 and 3 and the Delta Forge claims to their projects. Once that is done, the useful test is whether Ellendale and Harwood energization dates hold across successive quarterly filings or follow the Jamestown pattern of silent re-dating. In parallel, look for North Dakota Public Service Commission and utility interconnection records that could independently anchor the claimed megawatts.

The record supplied here ends mid-entry in July 2022, so this read covers Applied Digital's crypto-hosting era and cannot assess the Polaris Forge or Delta Forge AI campuses, which carry no capacity, cost or dated claims in the visible data. The 2006 and 2008 entries (21053, 21054, 21067) describe airport testing and shell-company administration unrelated to datacenters and look like a predecessor-entity mismatch that should be reviewed before publication. Ellendale and Polaris Forge 1 share an address and are probably one site recorded twice.

Announced capacity grew sharply within seven months. In August 2021 it was a single 100 MW energy services agreement (14511) against 500 MW of stated demand (14512). An 85 MW second facility followed (14375). By March 2022 the company targeted 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW over five years (14230), alongside a 1.5 GW Antpool joint venture (14235) and a 200 MW Garden City wind letter of intent (14236). Only the first 100 MW was tied to a signed utility agreement. The larger figures rest on a joint venture and a letter of intent, and the record shows no conversion of either into a site, a contract or a schedule.

The dated promises slipped. The first Jamestown facility was promised for Q4 2021 (14509, 14414, 14364) and reached 55 MW on February 2, 2022 (14008), a modest miss. The remaining 45 MW, promised for Q2 2022 (14227, 14196, 14069, 14009), was repeated unchanged in every filing through April without new detail. By July it had become a power-restoration problem. Temporary power was promised by September 2022 (13856) and permanent restoration later (13857), and the earlier Q2 commitment was never explicitly withdrawn. The 800 MW and 1.8 GW targets are both marked delayed, and the Garden City letter of intent repeats without progress across five filings (14204, 14074, 14063, 14034, 14015).

The record contains no independent evidence, such as permits, interconnection queue entries or site reporting, for any announced pace. This is the main gap.

Over the next 90 days, editors should extend the claim history past July 2022 and attach capacity, cost and in-service dates to the Polaris Forge and Delta Forge entries. They should also look for interconnection filings and county permits for Harwood, Boyce and Brookwood. Finally, they should check whether any post-2022 disclosure acknowledges the missed 2023 and 2024 targets or quietly drops them.

The supplied record ends partway through a July 2022 entry (claim 13857), so this read covers Applied Digital's crypto-hosting era and says nothing yet about Polaris Forge 1, 2 or 3, or about Delta Forge 1 or 2. Those projects appear on the project list with no capacity, cost, or dated claims attached. The 2006 and 2008 items (21053, 21054, 21067 through 21069) belong to a predecessor shell and have no bearing on facility delivery.

Within that window, announced capacity grew far faster than delivered capacity. In 2021 the company described a 100 MW Jamestown energy agreement and an 85 MW second site (14413, 14375). By March 2022 the framing had jumped to 800 MW by May 2023, 1.8 GW by May 2024, and 5 GW within five years (14228, 14229, 14230). Alongside those targets sat a 1.5 GW Antpool joint venture (14212, 14036) and a 200 MW Texas wind letter of intent at Garden City (14236). In the same period, roughly 55 MW was reported live at Jamestown (14008).

Dates did not hold. The first Jamestown facility was promised for Q4 2021 in at least four separate disclosures (14414, 14372, 14364, 14509) and was described as partially online only in February 2022. The remaining 45 MW was then pinned to Q2 2022 across seven nearly identical restatements between March and April (14227, 14206, 14196, 14069, 14055, 14030, 14009), and every one is flagged delayed. By July 2022 the narrative had shifted to restoring power at Jamestown (13856, 13857), with no visible reconciliation against the earlier Q2 promise. Both the 800 MW and 1.8 GW targets are marked delayed (14031, 14032), and the Garden City arrangement was repeated as a letter of intent at least six times without moving to a definitive agreement (14204, 14074, 14034, 14015).

The record contains no independent corroboration of any of these figures: no permits, no interconnection queue positions, no local reporting. Everything above rests on company statements, including the power disruption itself.

Over the next 90 days, the priority is to complete ingestion of post-2022 claims so that the Polaris Forge and Delta Forge commitments can be tested against this history of repeated restatement. Specific items to look for are North Dakota Public Service Commission and county filings for Ellendale and Harwood, utility interconnection filings for the Louisiana and Alabama sites, and any tenant lease disclosures that tie capacity to energization dates. Those would show whether the current pipeline has independent footing that the 2022 targets never did.

The excerpt supplied ends mid-2022, partway through claim 13856, so this read covers Applied Digital's hosting-era record and cannot assess the Polaris Forge or Delta Forge programs, which have no dated claims here. The 2006 and 2008 entries (21053, 21054, 21067-21069) concern transit testing and corporate service providers and appear to come from a predecessor entity, so they are set aside. Within the usable record, announced capacity scaled far faster than delivered capacity. The company began with a single 100 MW energy services agreement at Jamestown (14511, 14413) and an 85 MW second site (14375). By March 2022 it was citing 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), a 1.5 GW Antpool joint venture (14235), and 5 GW within five years (14230). The only delivered figure on record is 55 MW at Jamestown as of February 2022 (14008, 14054). Dates have not held. The Q4 2021 Jamestown completion target was repeated unchanged across four filings (14509, 14414, 14372, 14364) and is marked delayed. It then reappeared as 55 MW online with the remaining 45 MW due in Q2 2022 (14068, 14069). That reframing turned a missed completion into a partial success, and no claim in the record acknowledges the slip. The Q2 2022 tranche also slipped (14196, 14009, 14030). By July 2022 the company was discussing temporary and permanent power restoration at Jamestown (13856, 13857), with the permanent fix pushed to mid-2023. This points to an operational disruption rather than incremental growth. The 800 MW and 1.8 GW milestones are both flagged delayed. The Garden City 200 MW wind LOI (14236, 14204, 14034) was restated repeatedly but never advanced to a binding agreement within this record. Independent corroboration is absent from what was provided. No permits, interconnection queue positions, or site reporting are attached, and every project row lacks capacity and cost, so the announced pace cannot be checked against outside evidence. Over the next 90 days, the priority is to obtain the post-2022 claim history and compare Polaris Forge figures for Ellendale and Harwood against MISO and North Dakota PSC filings. Delta Forge 1 in Boyce and Delta Forge 2 in Brookwood should be checked against utility interconnection records. Polaris Forge 3 still has no location on file. The 5 GW target (14230), dated March 2027, is a test of whether the company restates, revises, or quietly retires its oldest headline commitment.

Applied Digital's capacity narrative on this record is one of rapid headline escalation built on a slow-moving physical base. The earliest anchor was a single 100 MW energy services agreement at Jamestown (14511, 14413, 14361). Within about seven months, framing had jumped to 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), and 5 GW over five years (14230, 14033). Alongside this sat an Antpool joint venture for up to 1.5 GW (14235, 14203, 14036) and a 200 MW Garden City wind letter of intent (14236, 14204, 14015). The ambition scaled by more than an order of magnitude. The contracted and energized footprint visible here never exceeded the original 100 MW site plus an 85 MW second facility (14375, 14362, 14040). The Antpool and Garden City items never move past joint-venture or LOI language in these claims.

Dated promises have not held. The first co-hosting facility was due in Q4 2021 and was repeated four times before slipping (14509, 14414, 14372, 14364). By early 2022, 55 MW was reported online (14054, 14008), but the remaining 45 MW was pushed to Q2 2022 and slipped again (14227, 14196, 14055, 14009). By July 2022 the site was discussing power restoration, with permanent power targeted for mid-2023 and also marked delayed (13856, 13857). Both the 800 MW and 1.8 GW milestones are delayed. The record shows the same Q2 2022 date restated across successive filings without visible acknowledgement that earlier targets had passed. The 5 GW target (14230) remains formally open to March 2027.

The supplied record contains no permits, interconnection-queue entries, or on-the-ground reporting, so the announced pace cannot be independently corroborated. The dataset also ends mid-entry in July 2022. The Polaris Forge and Delta Forge projects therefore carry no visible claims, and their trajectory cannot be assessed from this input. Polaris Forge 1 also shares an address with the Ellendale record and is likely a duplicate. The 2006–2008 items (21053, 21054, 21067–21069) appear to belong to a predecessor entity and should not be read as datacenter claims.

Over the next 90 days, watch whether the 5 GW framing is restated, resized, or quietly dropped ahead of its 2027 horizon. Check MISO and SPP queue positions for the Harwood and Ellendale sites, and utility or parish filings for Boyce and Brookwood, for evidence that load is contracted rather than announced. Finally, check whether the record can be re-ingested in full, so the post-2022 Forge-era claims can be scored on the same basis.

The supplied record effectively ends in July 2022 (the last entry, claim 13856, is cut off mid-line), so this read covers Applied Digital's hosting-era build-out and says little about Polaris Forge or Delta Forge, which appear only as project names with no capacity or cost attached. The 2006 and 2008 entries (21053, 21054, 21067-21069) concern a predecessor shell's auditors, transfer agents and an airport testing contract. They look like entity-history contamination and should be excluded from any delivery assessment.

On capacity, the trajectory is one of rapid headline inflation against modest delivered megawatts. Through late 2021 the story was a 100 MW energy services agreement at Jamestown (14511, 14413, 14361) and an 85 MW second site (14375, 14362). In March 2022 the ambition jumped to 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), 5 GW over five years (14230), a 1.5 GW Antpool joint venture (14235, 14203) and a 200 MW Garden City wind LOI (14236). Over the same window, delivered capacity was 55 MW (14008, 14054). Partner commitments also quietly contracted. The GMR arrangement appears at 100 MW in 14502 and at 50 MW in 14377 and 14359, with no explanation in the record.

Dates have slipped repeatedly, usually by restatement rather than acknowledgement. The Q4 2021 Jamestown completion (14509, 14414, 14372, 14364) became a partial 55 MW energization on February 2, 2022. The remaining 45 MW was then re-promised for Q2 2022 across at least five filings (14227, 14196, 14069, 14030, 14009), all now marked delayed. By July 2022 the company was promising Jamestown power restoration rather than expansion (13856, 13857). Both the 800 MW and 1.8 GW milestones lapsed. The 5 GW target lost its explicit end date between 14230 and 14033.

No independent evidence appears in this record: no permits, no interconnection queue positions, no on-the-ground reporting. The Garden City LOI never converts into a definitive agreement within the window shown.

Over the next 90 days, the priority is to extend the claim history past mid-2022 so the Polaris Forge and Delta Forge sites can be assessed on their own terms. Specific items to check are MISO or SPP queue entries and energization milestones for Ellendale and Harwood, parish and county permitting in Boyce and Brookwood, and whether any announced figure in those filings has a matching utility or regulator record.

The record supplied here stops in July 2022, mid-claim at 13856. None of the Polaris Forge or Delta Forge sites carries a dated claim. This read therefore covers the bitcoin-hosting era at Jamestown and the expansion rhetoric around it, not the current AI campus build-out. The 2006 and 2008 entries (21053, 21054, 21067-21069) appear to belong to a predecessor shell entity and should be set aside. The capacity arc is the most striking feature. In August 2021 the company described a single 100 MW energy services agreement (14511), with a 500 MW demand figure (14512) as aspiration. By November that had become two sites, 100 MW plus 85 MW (14373, 14375). By March 2022, before the first site was fully energized, the targets had jumped to 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), 5 GW over five years (14230) and a 1.5 GW Antpool joint venture (14235). That is roughly a fiftyfold increase in stated ambition within seven months, resting on a 200 MW Texas letter of intent (14236) rather than signed interconnection. Dates did not hold, and the slips were mostly carried forward rather than acknowledged. The Q4 2021 completion target for the first facility was restated without revision as late as November 30 (14364), and the site reached 55 MW only in February 2022 (14008). The remaining 45 MW was pinned to Q2 2022 in at least five separate filings between March and April (14227, 14196, 14069, 14009, 14030), and that target is also marked delayed. By July the conversation had moved to restoring power at Jamestown, temporarily by September 2022 and permanently by mid-2023 (13856, 13857). That marks a regression from expansion to recovery that the 800 MW and 1.8 GW targets never absorbed. On independent evidence, the record holds nothing: no permits, no interconnection queue positions and no site reporting for Garden City, the Antpool venture or any later site. Every project row, including the Jamestown site with 55 MW running, still shows announced status with blank capacity and cost, so the project table itself understates what was verifiably built. Over the next 90 days, the priorities are to backfill the post-2022 claim history for the Polaris Forge and Delta Forge campuses and to populate capacity on the project rows. It is also worth checking whether Harwood, Boyce and Brookwood appear in regional interconnection queues at the stated sizes. Finally, the company's next quarterly filing should be read for whether the Garden City and Antpool commitments were ever formally withdrawn or simply stopped being mentioned.

The record supplied here ends mid-entry in mid-2022 (claim 13856), so this read covers Applied Digital's crypto-hosting era and cannot assess the Polaris Forge and Delta Forge projects, which appear on the project list without any capacity, cost or timeline claims attached. The 2006 and 2008 entries (21053, 21054, 21067-21069) come from the predecessor shell and have no bearing on the facility trajectory.

Capacity ambition grew fast while deliveries lagged. In August 2021 the company described a 100 MW energy services agreement at Jamestown (14511, 14413) and 500 MW of expansion demand (14512). By November it had added an 85 MW second facility (14375, 14362). By March 2022 the framing had jumped to 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW over five years (14230). Alongside these sat a 1.5 GW Antpool joint venture (14235, 14203), a 200 MW Garden City wind letter of intent (14236, 14204) and a 200 MW Marathon hosting deal (13855). Through April 2022 the company kept repeating these targets without changes (14031, 14032, 14033), even though the near-term commitments beneath them were already slipping.

The Q4 2021 completion date for the first facility was repeated four times (14509, 14414, 14372, 14364) and missed. The company later reported 55 MW online as of February 2, 2022 (14008), which was disclosed as a fact rather than as a revision of the earlier date. The remaining 45 MW moved to Q2 2022 (14227, 14206, 14009) and is also flagged delayed, as is the 200 PH Bitmain equipment order (14504, 14416, 14378). By July 2022 the Jamestown narrative had shifted to restoring power, with a temporary fix promised by September 2022 and a permanent one by June 2023 (13856, 13857). Neither the 800 MW nor the 1.8 GW target is acknowledged as revised anywhere in the visible record.

No independent corroboration appears in this record: no permits, interconnection queue positions or site reporting. The announced pace therefore rests entirely on company statements, and the Garden City LOI never visibly converted into a contract.

Over the next 90 days, the first priority is completing the ingest so that claims about the HPC-era Ellendale, Harwood, Boyce and Brookwood projects can be read against this history. Beyond that, watch for North Dakota PSC siting filings, MISO and SPP queue entries for Ellendale and Harwood, and Louisiana and Alabama permit dockets. The key question is whether the stated energization dates for those sites are backed by utility commitments rather than letters of intent.

Applied Digital's capacity narrative escalated far faster than its delivered megawatts. The record opens in August 2021 with a single 100 MW energy services agreement at Jamestown (14511, 14413) and a loosely framed 500 MW of expansion demand (14512). A second facility at 85 MW appeared by November (14375, 14362). By March 2022, with roughly 55 MW energized, the company was citing 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), 5 GW over five years (14230), a 1.5 GW Antpool joint venture (14235, 14203), and a 200 MW Garden City wind letter of intent (14236). Headline capacity grew about fiftyfold in seven months while built capacity was still below the first site's contracted 100 MW.

Dates have not held. The year-end 2021 completion target for Jamestown (14414, 14372, 14364) slipped, though only modestly, with 55 MW reported online as of February 2, 2022 (14008). The remaining 45 MW was promised for Q2 2022 in at least five consecutive filings (14227, 14196, 14069, 14055, 14009) without any revision to that date, then missed. The 800 MW and 1.8 GW milestones (14228, 14031, 14229, 14032) are both marked delayed. The July 2022 claims about restoring Jamestown power (13856, 13857), with permanent restoration pushed to mid-2023, indicate that utility supply rather than construction was the binding constraint. The aggregate targets were never formally withdrawn. They stopped being repeated.

The supplied record contains no independent corroboration: no county permits, no MISO or SPP interconnection queue positions, and no site reporting for any project. The project register also shows capacity and cost blank for all eight sites. Ellendale and Polaris Forge 1 share an address, which suggests a rebrand rather than a second campus. Polaris Forge 2 and 3 and both Delta Forge sites have no dated claims at all in this record. Note also that the claim list provided ends mid-entry in July 2022, so this read cannot assess anything after that point.

Over the next 90 days, the most important test is the 5 GW target (14230), which falls due in March 2027. Management will either reaffirm it, rebase it to the Polaris and Delta Forge pipeline, or let it lapse. Beyond that, watch for interconnection queue filings or utility agreements at Harwood, Boyce, and Brookwood that would put physical substance behind those names, and for whether Garden City's 200 MW letter of intent (14214) ever becomes a dated, contracted commitment.

Applied Digital's capacity narrative escalated quickly and then outran its delivery. In August 2021 the anchor was a single 100 MW energy services agreement at Jamestown (14511, 14413), alongside a 500 MW expansion demand figure (14512). An 85 MW second site followed (14375, 14362). By March 2022 the ambition had multiplied: 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), 5 GW over five years (14230), plus up to 1.5 GW through the Antpool joint venture 1.21 Gigawatts LLC (14235, 14203). The company was repeating multi-gigawatt targets while its only operating asset was a partly energized 100 MW site. The 200 MW Garden City wind letter of intent (14236, 14204, 14074) stayed at LOI status through every restatement, with no dated milestone attached.

The dates held poorly. The first Jamestown co-hosting completion, promised for Q4 2021 (14509, 14414, 14372, 14364), is marked delayed. In April 2022 the company reported 55 MW online as of February 2 (14008, 14054), which suggests partial delivery rather than the full facility. The remaining 45 MW, promised for Q2 2022 (14227, 14196, 14055, 14030), slipped. By July 2022 Jamestown faced a power restoration problem, with temporary power targeted for September 2022 and permanent power for mid-2023 (13856, 13857). Neither the 800 MW nor the 1.8 GW milestone was met (14031, 14032). We find no acknowledgment of revised timing for these milestones in the record. The targets simply stopped recurring after spring 2022 rather than being formally reset.

The record contains no independent corroboration: no permits, interconnection queue entries, or site reporting for any project, including the Polaris Forge and Delta Forge sites listed as announced. The 2006 and 2008 claims (21053, 21054, 21067) concern unrelated predecessor-entity business and should not inform the datacenter read. The claim history provided also ends mid-2022, so Polaris Forge 1 through 3 and Delta Forge 1 and 2 carry no capacity or date claims here.

Over the next 90 days, the priority is backfilling the post-2022 claim history for the Polaris and Delta Forge sites. Their announced capacities should then be tested against utility interconnection filings, county permits for Harwood, Boyce and Brookwood, and evidence of energized load. Given the Jamestown precedent, any completion date should be read as provisional until power delivery is independently confirmed.

The capacity story in this record runs in two directions at once. The contracted base barely moved: a 100 MW energy services agreement at Jamestown anchors the earliest filings (14511, 14413, 14361) and is restated almost unchanged through April 2022 (14013, 14039), with an 85 MW second facility added in late 2021 (14375, 14362, 14040). The aspirational layer, by contrast, expanded sharply within seven months, from 500 MW of 'expansion demand' (14512) to 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), a 1.5 GW Antpool joint venture (14235) and a 5 GW five-year goal (14230). The gap between signed power and headline ambition widened to roughly an order of magnitude.

Dated promises have not held. Jamestown's first co-hosting completion was pinned to Q4 2021 in four successive disclosures (14509, 14414, 14372, 14364) and missed. By February 2022, 55 MW was reported online (14008), and the remaining 45 MW was promised for Q2 2022 across at least six claims (14227, 14196, 14206, 14069, 14055, 14030). That too slipped, and by July 2022 the site needed temporary power restoration targeted for September (13856), which was also marked delayed. The 800 MW and 1.8 GW targets were repeated verbatim in April 2022 (14031, 14032) with no re-dating after both passed, a pattern of silent lapse rather than acknowledged revision. The Garden City 200 MW wind LOI appears five times (14236, 14204, 14214, 14063, 14034) without advancing past letter-of-intent stage.

The record contains no independent corroboration, such as permits, interconnection queue positions or site reporting, for any announced pace. The 2006 and 2008 entries (21053, 21054, 21067 to 21069) belong to the predecessor shell and say nothing about datacenter delivery. The supplied claim list ends mid-2022 at 13857, so the Polaris Forge and Delta Forge projects, now the company's stated focus, carry no capacity, cost or date claims here. Ellendale and Polaris Forge 1 also share an address without reconciliation.

Over the next 90 days, watch whether the Polaris Forge 2 (Harwood), Delta Forge 1 (Boyce) and Delta Forge 2 (Brookwood) entries acquire dated capacity claims that can be checked against MISO or Southern Company queue filings and county permits. Watch whether the Ellendale and Polaris Forge 1 records are merged. Finally, watch whether the 5 GW target due March 2027 (14230) is reaffirmed, restated in AI-campus terms, or quietly dropped.

The record supplied for Applied Digital ends mid-2022, partway through claim 13857. The Polaris Forge and Delta Forge projects appear only as announced entries with no capacity, cost, or timeline claims attached. This read therefore covers the bitcoin-hosting era at Jamestown and Garden City, and its main lesson is about how the company frames ambition. The 2006 and 2008 items (21053, 21054, 21067-21069) belong to the predecessor shell and are not relevant to the facility story. Capacity language widened quickly while delivered megawatts moved slowly. The August 2021 filings anchored on a single 100 MW energy services agreement (14511, 14502) alongside a vague 500 MW expansion demand (14512). By March 2022 that had become 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), 5 GW over five years (14230), and an up-to-1.5 GW Antpool joint venture (14235, 14203). None of these carried a site, interconnection agreement, or cost figure. The Garden City 200 MW wind site was repeated across at least six filings (14236, 14204, 14214, 14074, 14063, 14034) but never progressed beyond a letter of intent. The dates tell a clearer story. The Jamestown completion target of year-end 2021 was restated four times (14509, 14414, 14372, 14364) and missed. The company then quietly reframed it as 55 MW online as of February 2022 (14008, 14054), with the remaining 45 MW moved to Q2 2022 (14227, 14196, 14030). That Q2 target also slipped, and by July 2022 the site was waiting on temporary power restoration (13856), with permanent restoration pushed to mid-2023 (13857). Each slip arrived as a new target rather than an explicit acknowledgement that the prior one failed. The supplied record includes no permits, queue positions, or site reporting, so none of the announced pace can be independently confirmed. The only corroborated figure is the utility contract at Jamestown (14233, 14013). Over the next 90 days, readers should look for Polaris Forge and Delta Forge claims that name energized megawatts rather than contracted or planned capacity. They should also watch for MISO and SPP interconnection filings for Harwood and Ellendale, and for any restatement of the 2023-2024 gigawatt targets that remain marked delayed (14031, 14032).

Applied Digital's capacity narrative grew quickly and without much hardware underneath it. In August 2021 the company described a 100 MW energy services agreement (14511, 14413) and a 500 MW expansion ambition (14512). By November it had added an 85 MW second site (14375). By March 2022 it had jumped to 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW over five years (14230). The Antpool joint venture at up to 1.5 GW (14235) and a 200 MW Garden City wind letter of intent (14236) sat on top of that. The Marathon 200 MW hosting agreement (13855) was a contracted-demand claim, not new capacity. Over roughly seven months, the stated ambition grew about fiftyfold while the only energized asset was Jamestown. The 2006 and 2008 entries (21053, 21054, 21067) appear to be the predecessor shell's filings and have no bearing on this trajectory. Jamestown's dates slipped repeatedly and quietly. The Q4 2021 completion target was restated unchanged as late as November 30, 2021 (14364), one month before it lapsed. Only 55 MW was online by February 2, 2022 (14008). The remaining 45 MW was reaffirmed for Q2 2022 across four filings (14227, 14196, 14069, 14009) without any change to the date, and it lapsed too. The July 2022 claims then recast the problem as power restoration, first by September 2022 (13856) and then permanently by mid-2023 (13857). None of those filings explicitly acknowledged the earlier Q2 date. The Bitmain 200 PH delivery went the same way: it was repeated four times (14504, 14416, 14378, 14360) and then marked delayed. The 800 MW and 1.8 GW targets are also marked delayed, with no revision on record. There is no independent evidence in this record: no permits, no interconnection queue positions, no site reporting. As supplied, the claims record stops mid-2022. Polaris Forge 1 through 3, Delta Forge 1 and 2, and the Ellendale site carry no capacity, cost or dated claims here, so the current AI-campus pace cannot be assessed against the company's own words. Over the next 90 days, watch four things. First, whether filings attach dated energization milestones and megawatt figures to the Polaris Forge phases. Second, whether MISO or SPP queue entries exist for Harwood and the Ellendale expansion. Third, whether local permits appear for Delta Forge in Rapides Parish and Tuscaloosa County. Fourth, whether the company formally retires the Garden City letter of intent and the 5 GW target due March 2027, or lets them lapse in silence as the Jamestown dates did.

The record supplied for Applied Digital ends in July 2022, so this read covers the crypto-hosting origin of the company and not the Ellendale, Polaris Forge, Delta Forge or Garden City build-outs listed as announced projects. No capacity, cost or timeline claims tie to those sites beyond the Texas wind letter of intent. The 2006 and 2008 entries (21053, 21054, 21067 through 21069) describe a predecessor shell and have no bearing on facility delivery.

Within that window the headline ambition grew very quickly while physical delivery stayed small. In August 2021 the company described 500 MW of expansion demand (14512) against a single 100 MW energy services agreement (14511). By March 2022 the targets were 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW over five years (14230), plus up to 1.5 GW through the Antpool joint venture (14235, 14203). The only energized capacity documented anywhere in the set is 55 MW at Jamestown as of February 2022 (14008, 14054). The 800 MW and 1.8 GW targets are both flagged delayed. The Garden City 200 MW wind project never moved past letter-of-intent language across six restatements (14236, 14204, 14074, 14063, 14034, 14015).

Dates slipped repeatedly and were usually carried forward without comment. First co-hosting completion was promised for Q4 2021 in four successive filings (14509, 14414, 14372, 14364), and the paired Bitmain equipment delivery shows the same pattern (14504, 14416, 14378, 14360). All of these were delayed. The remaining 45 MW at Jamestown was restated for Q2 2022 at least seven times (14227 through 14030) and is also marked delayed. Claims 13856 and 13857 then introduce a power-restoration problem at Jamestown with a permanent fix dated mid-2023. None of the preceding filings flagged that risk.

The supplied material contains no independent corroboration: no permits, no interconnection queue positions and no site reporting. The pace can therefore be assessed only against the company's own restatements, and those show ambition that was repeatedly deferred. That is a gap in this record, not evidence for or against the later campaigns.

Over the next 90 days, the priority is to extend the claim history past July 2022 so the Polaris Forge and Delta Forge commitments can be tested the same way. Beyond that, watch whether energization dates in filings match utility interconnection records for Ellendale and Harwood. Also watch whether any Delta Forge site in Louisiana or Alabama shows local permitting or substation activity. Finally, check whether capacity figures are restated as critical IT load or as gross utility power, since that framing has shifted before.

The record captured here stops in mid-2022, and it contains no capacity, timeline, or cost claims for the Polaris Forge or Delta Forge campuses. It therefore describes the company's crypto-hosting origins, not its current AI build-out. The 2006 and 2008 entries (21053, 21054, 21067-21069) belong to the predecessor shell and are not relevant to delivery. Ellendale and Polaris Forge 1 share an address and are probably the same site under two names.

Within that window, capacity ambition grew far faster than delivered megawatts. The 2021 filings anchor on a 100 MW energy services agreement at Jamestown (14413, 14361) and an 85 MW second facility (14375, 14362). By March 2022 the company was describing 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), and 5 GW within five years (14230). It also described a 1.5 GW Antpool joint venture (14235) and a 200 MW Garden City wind LOI (14236). None of those larger figures is followed in this record by a site-level claim, a permit, or an energization date. The Garden City and 1.21 Gigawatts items are restated almost word for word across at least five filings (14063, 14074, 14015, 14036) and never move past letter-of-intent language.

Dates did not hold. The Q4 2021 first-facility completion (14509, 14414, 14364) turned into 55 MW online as of February 2022 (14008). The remaining 45 MW was promised for Q2 2022 in seven separate claims (14227 through 14030) and then missed. The next disclosure is a Jamestown power loss with temporary restoration targeted for September 2022 (13856) and permanent restoration pushed into mid-2023 (13857). That slip was acknowledged only once an outage forced it. Earlier filings kept repeating the Q2 target without re-dating it.

This dossier includes no permits, interconnection queue entries, or on-the-ground reporting, so the announced pace cannot be independently verified here.

Over the next 90 days, the fiscal first-quarter 10-Q, due in October, is the natural test. The first question is whether it gives energized megawatts and tenant commencement dates for Polaris Forge 1 and 2 that can be cited. The next is whether Harwood, Boyce, and Brookwood appear in North Dakota PSC siting dockets, MISO or SPP queues, or Cleco, Entergy, or Alabama Power large-load filings. The last is whether the lapsing 5 GW horizon of March 2027 (14230) is formally retired or quietly dropped.

The record supplied here ends partway through claim 13857, in mid-2022, so this read covers Applied Digital's crypto-hosting phase. It cannot assess the Polaris Forge or Delta Forge campuses, which appear on the project list with no capacity, cost or dated claims attached. The 2006 and 2008 items (21053, 21054, 21067-21069) concern airport testing and corporate service providers. They have no bearing on facility delivery and are set aside.

The capacity arc is steep. In August 2021 the company described a 100 MW energy services agreement (14511, 14413) and a planned 500 MW of expansion demand (14512). By November it added an 85 MW second facility (14375, 14362). By March 2022 it had published 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032) and 5 GW within five years (14230, 14033). It also claimed a joint venture with Antpool for up to 1.5 GW (14235, 14036), a 200 MW wind-backed Texas letter of intent (14236, 14034) and a 200 MW Marathon hosting agreement (13855). None of the larger figures is tied to a named site, and both the 800 MW and 1.8 GW targets are now marked delayed.

The dates have not held. The Q4 2021 completion target for Jamestown was repeated unchanged in four consecutive filings (14509, 14414, 14372, 14364). It was then replaced without comment by a statement that 55 MW had been online since February 2022 (14008). The remaining 45 MW moved to Q2 2022 (14227, 14196, 14009) and slipped again. The July 2022 power loss produced a September 2022 temporary-restoration date (13856), also delayed, and a permanent fix pushed to mid-2023 (13857). The 200 PH Bitmain equipment order slipped in parallel (14504, 14360). The Garden City letter of intent never acquired a date in any restatement.

No independent evidence is attached to any of these claims: no permits, no interconnection-queue entries and no site reporting. The announced pace therefore rests entirely on company statements. Over the next 90 days, the priority is to extend the record past 2022 and attach dated capacity claims to Polaris Forge 1 through 3 and Delta Forge 1 and 2. The Ellendale and Polaris Forge 1 entries share an address and should be reconciled. MISO queue positions for the North Dakota sites and parish or county permits in Boyce and Brookwood would test whether the newer programme runs closer to its schedule than the 2022 targets did.

Applied Digital's capacity story in this record grew far faster than its delivered megawatts. In August 2021 the company described a 100 MW energy services agreement for its first site (14511, 14413) and 500 MW of expansion demand (14512). By November it had added an 85 MW second facility (14375, 14362). By March 2022 the targets had become 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW within five years (14230). On top of those sat a 1.5 GW joint venture with Antpool (14235, 14203) and a 200 MW wind-backed Texas letter of intent at Garden City (14236). Garden City appears repeatedly through April 2022 (14034, 14074) without ever moving past letter-of-intent status. The 2006 and 2008 entries (21053, 21068) belong to the predecessor shell and say nothing about this build-out.

The dated promises did not hold. The first co-hosting facility at Jamestown was promised for Q4 2021 (14509, 14414, 14364) and was delayed. The Bitmain 200 PH equipment order was repeated four times against the same deadline (14504, 14416, 14378, 14360) and slipped each time. By April 2022 the company reported 55 MW online as of February 2 (14008). The remaining 45 MW was then promised for Q2 2022 (14227, 14009), and that date also slipped. The July 2022 claims then turn to restoring Jamestown's power, with temporary restoration promised by September 2022 (13856) and permanent restoration by mid-2023 (13857). That implies an outage the earlier filings never anticipated. Meanwhile the 800 MW and 1.8 GW targets were restated unchanged in April 2022 (14031, 14032) even though the flagship site was running at roughly half its contracted load. We see that as revision by omission rather than acknowledgement.

The record contains no independent corroboration of the announced pace, whether permits, interconnection-queue positions or site reporting. The current project list is also thin. Polaris Forge 1 through 3, Delta Forge 1 and 2, and Ellendale all carry no stated capacity or cost, and the claims supplied here stop in mid-2022.

Over the next 90 days, watch for three things. First, whether the company attaches megawatt and cost figures to the Polaris Forge and Delta Forge sites. Second, whether interconnection filings appear for Harwood, Boyce and Brookwood. Third, whether the company formally addresses the five-year 5 GW target (14230), which comes due in March 2027.

Applied Digital's capacity story in this record has grown much faster than its delivered megawatts. The early commitment was modest and concrete: a 100 MW energy services agreement for Jamestown (14511, 14413, 14361), with 500 MW of vaguely described expansion demand (14512) and an 85 MW second site (14375, 14040). By March 2022 the ambition had expanded roughly tenfold. The company then promised 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), and 5 GW over five years (14230, 14033), alongside an Antpool joint venture sized at up to 1.5 GW (14235, 14203). The 800 MW and 1.8 GW milestones are both marked delayed, and no later claim in this record restates or retires them. The 200 MW Garden City wind site appears only as a letter of intent, repeated nearly word for word across five filings (14236, 14204, 14074, 14063, 14015), with no sign of conversion to a definitive agreement.

The timelines slipped quietly. The Q4 2021 completion target for Jamestown was still being reaffirmed on 2021-11-30 (14364). The next filings reframed it: 55 MW was online as of February 2022 (14008), and the remaining 45 MW became a Q2 2022 deliverable (14206, 14009). That date also slipped. By July 2022 the discussion had moved to restoring power to the site, with a temporary fix by September 2022 (13856, delayed) and permanent restoration by mid-2023 (13857). The company never described the phased reframing as a revision. The Bitmain 200 PH equipment order shows the same pattern of carried-forward delay (14504, 14416, 14378, 14360). The 2006 and 2008 entries (21053 through 21069) belong to the predecessor shell and say nothing about this trajectory.

The record contains no independent corroboration: no permits, no utility interconnection queue positions, and no on-the-ground reporting for any site. The Polaris Forge and Delta Forge projects are listed with no capacity, cost, or attached claims, and the claim history as supplied ends mid-2022. For the AI-era campuses, the current pace therefore cannot be assessed from this data.

Four things are worth watching over the next 90 days. The first is whether claims with stated megawatts and dates get attached to Polaris Forge 1 through 3 and Delta Forge 1 and 2, as this company's disclosures lean heavily on nameplate figures. The second is whether interconnection queue filings or county permits for Harwood, Boyce, and Brookwood match those figures. The third is whether the next quarterly filing addresses the lapsed 800 MW and 1.8 GW targets or leaves them unmentioned. The fourth is how the 5 GW target's March 2027 horizon (14230) is framed as it approaches.

The record for Applied Digital is really two records. The 2006 and 2008 items (claims 21053, 21054, 21067 through 21069), covering a Volpe Center contract, airport testing and auditor appointments, describe a predecessor shell and say nothing about datacenter delivery. The usable history begins in August 2021, and the capacity ambition expands quickly after that. It starts with a single 100 MW energy services agreement at Jamestown (14511, 14413) and a loosely stated 500 MW of demand (14512). By March 2022 it had become 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW within five years (14230), plus a 1.5 GW Antpool joint venture (14235). The 5 GW figure lost its date when it was restated in April (14033). The joint venture's recorded formation value of 8,000 USD (14234) is a reminder that the vehicle was nominal, not capitalised.

Delivery at the one site with concrete dates tells a more modest story. Full Jamestown completion by year-end 2021 was repeated across four filings (14509, 14414, 14372, 14364) and missed. The company then reported 55 MW online as of February 2, 2022 (14008) and reset the remaining 45 MW to Q2 2022. That target was restated at least seven times (14227, 14196, 14206, 14069, 14055, 14009, 14030) and also slipped. After the July 2022 power disruption, a September 2022 temporary-restoration date (13856) was missed as well, and a permanent fix was pushed toward mid-2023 (13857). The pattern is consistent: dates are restated verbatim until they pass, then quietly replaced. Neither the 800 MW nor the 1.8 GW milestone was acknowledged as missed. The Garden City 200 MW wind site never moved past a letter of intent (14236, 14063, 14015).

The record contains no independent corroboration of any of these claims. There are no permits, no interconnection-queue positions and no utility filings. The eight projects on file are all still marked announced with no capacity or cost figures, including Jamestown, which clearly operated. Ellendale and Polaris Forge 1 share an address and are likely duplicates. The file also ends in mid-2022, so the Polaris Forge and Delta Forge programmes have no claims on record at all.

Over the next 90 days, the priority is the fiscal first-quarter filing expected in October. That filing should be used to populate dated capacity and energisation claims for Polaris Forge 1 and 2 and the Delta Forge sites. Each should then be tested against North Dakota PSC siting dockets, MISO and SPP queue entries, and local permits in Boyce and Brookwood. Given the history above, any new date should be tracked from its first appearance.

The record supplied here ends in July 2022 and is cut off partway through claim 13857. The Polaris Forge, Delta Forge and Harwood projects listed on the company's project roster have no capacity or timeline claims in this excerpt. The read below therefore covers the Jamestown and Garden City era and should not be taken as a judgment on the current AI-campus buildout. Claims 21053 through 21069 predate the datacenter business and belong to a predecessor entity. They carry no weight for delivery analysis.

On capacity, the trajectory is one of rapid escalation on paper. In August 2021 the company framed a 100 MW energy services agreement for its first facility (14511) against 500 MW of expansion demand (14512), then added an 85 MW second facility (14375, 14362). By March 2022 the ambition had multiplied: 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), 5 GW within five years (14230), and an Antpool joint venture sized at up to 1.5 GW (14235). That venture's stated formation figure of 8,000 USD (14234) is difficult to square with its headline scale. The 200 MW Garden City wind site appears in at least six filings between March and April 2022 (14236, 14204, 14214, 14074, 14063, 14015) and never advances beyond letter-of-intent language in this record.

Dates have not held. The first Jamestown facility, promised for year-end 2021, was restated unchanged across four filings (14509, 14414, 14372, 14364) and then reported at 55 MW online as of February 2, 2022 (14008). That is a modest slip, and it was never acknowledged as one. The remaining 45 MW was pinned to Q2 2022 in seven separate claims (14227, 14196, 14206, 14069, 14055, 14009, 14030) and missed. By July 2022 the conversation had shifted to restoring temporary power at Jamestown by September (13856), itself delayed, and permanent power by mid-2023 (13857). Both the 800 MW and 1.8 GW milestones are marked delayed.

The record contains no independent corroboration: no permits, no interconnection queue entries, no on-the-ground reporting. Every capacity figure is self-reported.

Over the next 90 days, the priority is closing the data gap: ingest post-2022 claims for Ellendale, Harwood, Boyce and Brookwood, and check them against MISO and SPP queue positions and state utility commission dockets. The 5 GW target carries a March 2027 deadline (14230). Watch whether the company restates it, quietly retires it, or redefines it around the Polaris Forge campuses.

The record here stops mid-2022, at claim 13857, so this read covers Applied Digital's hosting-era build-out rather than its Polaris Forge or Delta Forge phase. None of those later projects carries a capacity or cost figure in the dataset. Claims 21053 through 21069 (2006–2008) describe an airport-testing contractor and routine auditor and transfer-agent engagements, and they look like a predecessor or misresolved entity. They should be reviewed before anyone leans on them.

On capacity, the arc is steep. In 2021 the company framed itself around a 100 MW Jamestown energy services agreement (14511, 14413, 14373) and an 85 MW second site (14375, 14362, 14040), with a loose 500 MW expansion-demand figure alongside them (14512). By March 2022 the ambition had multiplied. The company set 800 MW by May 2023 (14228) and 1.8 GW by May 2024 (14229), announced a 5 GW five-year target (14230, 14033), formed a 1.5 GW Antpool joint venture (14235, 14212), and signed a 200 MW Garden City letter of intent (14236). That escalation came while the only delivered asset in the record was 55 MW at Jamestown (14008, 14054).

The dates have not held. Every Q4 2021 completion promise for the first facility is flagged delayed (14509, 14414, 14372, 14364). The remaining 45 MW at Jamestown was restated for Q2 2022 in at least five filings between March and April (14227, 14196, 14030, 14009, 14055) and never revised before it lapsed. By July the company was talking about restoring temporary power at Jamestown by September 2022 (13856) and permanent power later (13857). That is a regression from expansion to recovery, and the record contains no acknowledgement that the Q2 target had failed. The 800 MW and 1.8 GW targets also lapsed with no restatement in the record. The Garden City LOI was repeated almost verbatim across filings (14204, 14015, 14034, 14063) without ever becoming a definitive agreement.

The record contains no independent corroboration: no permits, no interconnection queue positions, no on-the-ground reporting. Even the 55 MW online figure is company-sourced.

Over the next 90 days, three things matter. First, the dataset needs to ingest post-2022 claims so the Ellendale and Harwood announcements can be tested against the pattern above. Second, North Dakota PSC dockets and MISO or SPP queue entries for those sites would show whether the stated pace is backed by interconnection progress. Third, watch whether the 5 GW target dated March 2027 (14230) is reaffirmed, resized, or quietly dropped.

Applied Digital's claim record for this cluster ends in August 2022, while the project list now includes Polaris Forge 1 through 3 and Delta Forge 1 and 2 with no capacity, cost, or date attached. This read therefore covers the company's crypto-hosting origins, not its current AI campus pipeline. That gap is the first thing to note.

Capacity ambitions grew much faster than delivered capacity. The August 2021 baseline was a single 100 MW energy services agreement (14511) plus a loosely stated 500 MW of expansion demand (14512). By March 2022 the company was targeting 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), 5 GW over five years (14230), and up to 1.5 GW through the Antpool joint venture (14235, 14203). The second site moved from 85 MW (14375, 14040) to a 200 MW Texas wind-backed facility (14236). Ellendale entered at 180 MW (13779). Against this, the only capacity actually reported online was Jamestown's 55 MW (14008, 14054) and a later claim of full operation there (13806).

Dates have slipped consistently, and the slips were usually not acknowledged directly. The end-2021 Jamestown completion appears in four filings (14509, 14414, 14372, 14364). The first 55 MW was dated February 2022, and the target was never restated. The remaining 45 MW moved from a Q2 2022 promise (14227, 14009) to an August milestone (13806) without a revised date on record. The Bitmain 200 PH order is repeated unchanged across four filings (14504, 14416, 14378, 14360), all delayed. Power-restoration targets (13856, 13857) were issued weeks before the full-capacity claim, and the record alone does not reconcile that sequence. Garden City stayed at letter-of-intent stage through repeated restatements (14214, 14063, 14034) before being recast as operational by end-2022 (13765), which also slipped. Both the 800 MW and 1.8 GW milestones are marked delayed.

No permit, interconnection-queue, or on-the-ground evidence is attached to any of these claims. The announced pace is supported only by the company's own statements.

In the next 90 days, watch for three things. First, the 5 GW five-year target (14230) comes due in March 2027. Check whether it is restated, quietly retired, or re-scoped around the Polaris and Delta Forge sites. Second, the Forge projects need filed capacity and dates so their trajectory can be tested. Third, look for utility interconnection filings in North Dakota, Louisiana, and Alabama that would independently corroborate any new megawatt claims.

This record is truncated. The claims supplied end mid-entry in August 2022, and none of them attach capacity, dates or cost to the Polaris Forge or Delta Forge sites listed on the project roster. What follows is therefore a read of Applied Digital's crypto-hosting era, when its delivery habits first became visible. It is not a read of its current AI campus buildout.

Announced capacity grew much faster than delivered capacity. The anchor asset was a single 100 MW energy services agreement at Jamestown (14511, 14413, 14373), later joined by an 85 MW second site (14375, 14362, 14040). Within seven months the company was also talking about 500 MW of expansion demand (14512), 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032) and 5 GW over five years (14230, 14033). Layered on top were a 1.5 GW joint venture with Antpool (14235, 14036) and a 200 MW Garden City wind LOI (14236, 14204, 14034). The Garden City project never moves past letter-of-intent language anywhere in the record. Both dated gigawatt-scale targets are flagged as delayed.

Dates slipped repeatedly, and the targets were restated unchanged rather than revised. First co-hosting completion by Q4 2021 was reaffirmed in four separate disclosures through 30 November 2021 (14509, 14414, 14372, 14364) and still missed. The company reported 55 MW live as of 2 February 2022 (14008, 14029). The remaining 45 MW was then promised for Q2 2022 in at least six claims (14227, 14206, 14196, 14069, 14055, 14030), with no revised date recorded. By July 2022 the story had become a power loss at Jamestown, with temporary restoration promised by September 2022 (13856) and permanent restoration by June 2023 (13857). Both dates are marked delayed. A contemporaneous claim of a 100% operating-capacity milestone (13806) sits uneasily against that outage, and it needs reconciling with the source filings. The Bitmain 200 PH equipment order (14504, 14378) also missed its year-end 2021 date.

The record contains no permits, interconnection queue entries or site reporting, so none of the announced pace is independently corroborated here.

Over the next 90 days, the first task is to backfill the missing post-2022 claims. After that, watch four things. First, whether Polaris Forge 1, 2 and 3 and Delta Forge 1 and 2 acquire specific megawatt figures and energization dates. Second, whether the 5 GW target dated March 2027 (14230) is formally retired or restated. Third, whether utility queue or MISO/SPP and SERC filings for Harwood, Boyce and Brookwood match the company's cadence. Fourth, whether any new dated promise is ever explicitly re-dated rather than silently carried forward.

Applied Digital's capacity story in this record grew far faster than its delivered megawatts. The company opened in August 2021 with a single 100 MW energy services agreement at Jamestown (14511, 14413) and an 85 MW second facility (14375, 14362), framed against 500 MW of expansion demand (14512). By March 2022 the ambition had jumped to 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), 5 GW within five years (14230), a 1.5 GW joint venture with Antpool (14235), and a 200 MW wind-powered Texas site (14236). That is roughly a fiftyfold increase in headline ambition in seven months. At that point the only energized capacity on record was 55 MW at Jamestown as of February 2, 2022 (14008).

The dated promises did not hold. Jamestown's Q4 2021 completion was repeated across four filings (14509, 14414, 14372, 14364) and then became 55 MW online plus 45 MW by Q2 2022. That 45 MW date was restated at least five more times without change (14227, 14196, 14069, 14055, 14009) before the facility lost power in mid-2022. The company then set a temporary restoration date of September 2022 (13856) and a permanent one of June 2023 (13857), and both were missed. The August 2022 claim of 100% operating capacity (13806) carries no date, so it cannot be measured. The Bitmain 200 PH order kept its December 2021 date through four restatements after that date was already slipping (14504, 14416, 14378, 14360). Most notably, the 800 MW and 1.8 GW targets passed their deadlines, and nothing in the record shows them being re-dated or formally withdrawn.

No permits, interconnection queue positions, or on-the-ground reporting in the record corroborate any site beyond Jamestown. The Garden City project stays a letter of intent through every restatement (14204, 14034, 14074). The Antpool joint venture was reaffirmed as late as August 2022 (13782) without a named site.

The available record ends mid-entry at claim 13764 in August 2022. Polaris Forge 1 through 3, Delta Forge 1 and 2, and Ellendale are listed with no capacity, cost, or dated claims attached. Over the next 90 days, watch for three things. First, whether filings attach specific megawatts and energization dates to those sites. Second, whether matching entries appear in utility and ISO interconnection records. Third, whether the 5 GW by March 2027 goal (14230) is restated, resized, or quietly dropped as that deadline nears.

The record shows Applied Digital's stated ambition growing much faster than anything it reported as built. In August 2021 the plan was a single 100 MW energy services agreement (14511) and a 500 MW expansion figure (14512). By November that had become a first facility plus an 85 MW second site (14375). By March 2022 the company was naming 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), 5 GW over five years (14230), a 1.5 GW Antpool joint venture (14235), a 200 MW Texas letter of intent (14236), and later a 200 MW Marathon hosting agreement (13855). Throughout that stretch, the only energized capacity cited was the 55 MW at Jamestown (14008, 14054). Headline capacity rose roughly fiftyfold on paper while delivered capacity stayed flat.

Dates have mostly slipped, and the slips were not acknowledged. The Q4 2021 completion target for the first co-hosting facility was restated at least four times (14509, 14414, 14372, 14364) and then missed. The remaining 45 MW at Jamestown was promised for Q2 2022 across more than five filings through mid-April (14227, 14196, 14069, 14055, 14030) without any change in wording. The company did not claim 100 percent operating capacity until August (13806). That milestone sits awkwardly next to the July disclosure of a Jamestown power loss with restoration dates stretching into 2023 (13856, 13857). The Garden City LOI was repeated six times (14236 through 14034) with no progression, and its October 2022 opening target (13777) is marked delayed. The 800 MW and 1.8 GW targets (14031, 14032) are flagged delayed, and the supplied record contains no explicit restatement of either.

The record contains no independent corroboration: no permits, no interconnection queue positions, and no site reporting. Every capacity figure here is self-reported, so the pace should be treated as unverified. The supplied record also ends in August 2022, which means the Polaris Forge and Delta Forge projects appear with no capacity, cost or timeline claims at all. Ellendale and Polaris Forge 1 share an address and are likely one site.

Over the next 90 days, three things matter. First, watch whether the company attaches concrete megawatt figures and energization dates to Harwood, Boyce and Brookwood. Second, check whether matching large-load interconnection filings or utility agreements appear for those sites. Third, see whether any filing formally retires or restates the 2022 gigawatt targets rather than letting them lapse silently.

The record supplied here ends in August 2022, partway through claim 13770. Claims for Polaris Forge 1 through 3 and Delta Forge 1 and 2 are therefore absent, and this read covers only the Jamestown and Garden City era of Applied Digital's disclosures. Within that window, the ambition grew quickly while delivery stayed small. In 2021 the company described a 100 MW energy services agreement at its first site (14413, 14361) and an 85 MW second facility (14375, 14362). By March 2022 it was targeting 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW over five years (14230), alongside an Antpool joint venture for up to 1.5 GW (14235). Most of that new capacity rested on letters of intent and joint-venture framing rather than contracted power. The Texas 200 MW wind site, for example, was repeated as an LOI across at least six filings (14236, 14204, 14074, 14063, 14015, 14034) without advancing. The date discipline is weak. The first co-hosting facility was promised for year-end 2021 in four successive filings (14509, 14414, 14372, 14364), and all four are marked delayed. The 55 MW reported online as of February 2022 (14008) arrived after that deadline. The remaining 45 MW was pinned to Q2 2022 in six claims (14227 through 14009), and that slipped too. When the April 2022 restatement repeated the 5 GW goal (14033), it dropped the explicit deadline carried in 14230, and nothing in the record acknowledges the change. The Antpool LLC formation (13729) and the second-site ground lease (13763) are both recorded as delayed against dates that had already passed when those claims were logged. The August 2022 statement that Jamestown had reached 100 percent operating capacity (13806) sits uneasily beside July commitments to restore temporary power by September 2022 (13856) and permanent power by mid-2023 (13857), both of which slipped. The record contains no independent evidence: no permits, no interconnection queue positions, no utility filings and no site reporting. The announced pace is therefore unverified rather than contradicted. Over the next 90 days, watch whether the fiscal first-quarter 10-Q, typically filed in October, reconciles the 5 GW by March 2027 target against energized megawatts. Also watch for queue entries or county permits tied to Harwood, Boyce and Brookwood that would convert the Forge announcements into checkable milestones.

Applied Digital's delivery record, as captured here, is a story of ambition compounding faster than energized megawatts. The company began in August 2021 with a single 100 MW energy services agreement (14511) and a 500 MW demand signal (14512). Within seven months the headline had become a 5 GW, five-year target (14230, 14033), bridged by interim marks of 800 MW by May 2023 (14228, 14031) and 1.8 GW by May 2024 (14229, 14032). Both interim marks are flagged delayed. A 1.5 GW joint venture with Antpool (14235, 14036) was repeated heavily through April 2022 and then falls silent in the record, with no wind-down or restatement. The '85 MW second facility' (14375, 14362, 14040) is later displaced by a 200 MW Garden City 'second facility' (13765) with no reconciliation between the two.

On dates, the pattern is repetition rather than revision. The first Jamestown completion was promised for Q4 2021 in four successive filings (14509, 14414, 14372, 14364), alongside a Bitmain 200 PH delivery on the same deadline (14504, 14416, 14378, 14360), and all of it slipped. Jamestown reached 55 MW by February 2022 (14008). The remaining 45 MW was then promised for Q2 2022 at least six times (14227, 14196, 14206, 14069, 14055, 14030) and also slipped. By July 2022 the disclosures had shifted to restoring temporary and permanent power (13856, 13857), an operational setback that earlier claims did not foreshadow. The company claimed 100% operating capacity a month later (13806), but the permanent-power fix it had dated to mid-2023 is itself marked delayed. Garden City moved from a letter of intent (14236) to an October 2022 opening (13777) without evidence in the record of a site acquisition in between, and that date slipped too.

The record contains no permits, interconnection queue entries, or on-the-ground reporting, so none of the announced pace can be independently corroborated here. The SparkPool agreement continued to be cited into April 2022 (14018, 14019), after that pool was widely reported to have wound down operations in late 2021, which argues for reading partnership claims skeptically.

Over the next 90 days, the priority is to close the ledger gap. Claims stop in August 2022, while Polaris Forge 1 through 3 and Delta Forge 1 and 2 carry no capacity, cost, or date entries at all. Watch for MISO and SPP interconnection filings tied to the Ellendale 180 MW figure (13746) and Harwood, and check whether any new Polaris Forge targets carry dated, energized-MW milestones or only contracted-capacity language.

The most important finding is what this record lacks. Every claim that carries a number describes the 2021-2022 crypto-hosting business. The projects that now define Applied Digital's pitch, Polaris Forge 1 through 3 and Delta Forge 1 and 2, sit on the project list with no capacity, cost or date attached. The claim history supplied also cuts off in August 2022, at claim 13777. Any read of current delivery pace is therefore a read of how the company handled its earlier promises.

Those promises scaled fast and then quietly shrank. The starting point was a 100 MW Jamestown energy services agreement (14511, 14413) alongside a loose 500 MW of expansion demand (14512), and an 85 MW second site followed (14375). In March 2022 the ambition jumped to 800 MW by May 2023, 1.8 GW by May 2024 and 5 GW within five years (14228, 14229, 14230). A 1.5 GW Antpool joint venture (14212) and a 200 MW Garden City wind-site LOI (14236) sat on top of that. By August 2022 the combined target had been restated as 500 MW (13728). That is well under a third of the 1.8 GW goal, and nothing in the claims presents it as a revision. Garden City moved from LOI to fully contracted (13745) with no construction or energization date ever attached.

Timelines slipped in a consistent pattern. The first Jamestown facility was promised for Q4 2021 (14509, 14364), and the company reported 55 MW online in early February 2022 (14008), a modest miss. The remaining 45 MW was restated as a Q2 2022 target in five separate filings (14227, 14206, 14069, 14055, 14030) without delivery being confirmed. By July, a power disruption produced new temporary and permanent restoration dates (13856, 13857), and both are flagged as delayed. The August claim of 100% operating capacity (13806) gives neither a megawatt figure nor a date, so it cannot be reconciled with those restoration targets.

No independent evidence is attached to any of these claims. There are no permits, interconnection queue positions or on-the-ground reporting, so the announced pace rests entirely on company statements.

Over the next 90 days, the key question is whether the Polaris Forge and Delta Forge records acquire specific megawatt figures and energization dates. It also matters whether those figures can be checked against utility interconnection filings, county permits and tenant disclosures. The Ellendale entry needs attention too, since it shares an address with Polaris Forge 1 and may be a duplicate or a renaming. The earlier habit of shrinking targets without comment is the pattern to test against the new campus claims.

The record supplied for Applied Digital ends in late August 2022, so this read covers the crypto-hosting era anchored at Jamestown. It does not cover the Polaris Forge or Delta Forge AI campuses. Those sites appear in the project list with no capacity, cost or dated claims attached.

Capacity rhetoric expanded faster than delivered megawatts. In August 2021 the company described a 100 MW energy services agreement alongside 500 MW of planned expansion demand (14511, 14512). By March 2022 that had become 800 MW by May 2023, 1.8 GW by May 2024 and 5 GW over five years (14228, 14229, 14230). The 1.5 GW Antpool joint venture (14235, 14212) and a 200 MW Garden City wind LOI (14236, 13778) sat on top of those targets. Over the same period, the concrete figure in the record stayed at a 100 MW Jamestown agreement, of which 55 MW was reported online as of February 2, 2022 (14008, 14054). A second 85 MW site was repeatedly described as reserved but never located (14375, 14040).

Dates did not hold. The first co-hosting facility was promised by year-end 2021 across four successive filings (14509, 14414, 14372, 14364), and the 55 MW energization came roughly a month late. That lateness was not framed as a slip (14068). The remaining 45 MW was restated for Q2 2022 at least five times through April 12 (14227, 14206, 14069, 14030, 14009) and then overtaken by a power loss at Jamestown. The company set temporary restoration for September 12, 2022 and permanent restoration for mid-2023 (13856, 13857), and both are marked delayed. An August 2022 claim of a 100% operating-capacity milestone (13806) sits awkwardly beside the outage timeline and deserves scrutiny. The 800 MW and 1.8 GW targets were repeated verbatim in April 2022 (14031, 14032) and then disappear without any visible revision. That is a quiet retirement, not an acknowledged reset.

The supplied record contains no permits, interconnection-queue positions or site reporting, so the announced pace cannot be independently corroborated here. That gap is itself the finding.

Over the next 90 days, watch for four things. First, whether the 5 GW horizon (14230), which nominally expires in March 2027, is restated, rebased onto Polaris Forge, or dropped. Second, whether Polaris Forge 2 at Harwood and Delta Forge 1 and 2 acquire dated capacity claims. Third, whether those claims can be matched to North Dakota PSC siting dockets and MISO or SPP queue entries. Fourth, whether Garden City resurfaces or is formally abandoned.

Applied Digital's capacity narrative in this record inflated quickly and then stopped being updated. The 2021 filings anchor on a single 100 MW energy services agreement at Jamestown (14511, 14413, 14361) and a vaguer 500 MW expansion demand figure (14512). By March 2022 that had become an 800 MW target for May 2023 (14228), 1.8 GW for May 2024 (14229), a 5 GW five-year ambition (14230, 14033), and an Antpool joint venture sized at up to 1.5 GW (14235, 14212). That is roughly a fiftyfold increase in headline ambition within seven months. The only additions with a named site were a second 85 MW facility (14375, 14040) and a 200 MW Garden City letter of intent (14236, 14034). Neither the 800 MW nor the 1.8 GW milestone is ever revisited in later claims. Both dates have passed with no restatement on file.

Dates have not held. The Q4 2021 completion of the first co-hosting facility (14509, 14414, 14364) slipped. The company then reported 55 MW live at Jamestown as of February 2022 (14008) and promised the remaining 45 MW by Q2 2022 (14227, 14030). That promise was repeated almost verbatim across at least five filings (14206, 14069, 14055, 14009) and was not reframed. Instead, the July 2022 claims introduce temporary and permanent power restoration dates (13856, 13857), which implies an outage the earlier filings never anticipated. The 100 percent operating claim that follows (13806) carries no date. Garden City's opening moved from October 2022 (13777) to year-end 2022 (13765) within the same filing cycle, and neither date is marked as met.

The record contains no independent corroboration: no permits, no ERCOT or MISO interconnection queue entries, and no site reporting. Every capacity figure is self-reported. There is also a large structural gap. Polaris Forge 1 through 3, Delta Forge 1 and 2, and Ellendale are listed as announced, but none has a capacity, cost, or dated claim, and the claim history stops in August 2022.

Over the next 90 days, the priority is to backfill that gap. That means current capacity and energization claims for the Polaris Forge and Delta Forge sites, interconnection queue positions in North Dakota, Louisiana and Alabama, and any filing that quietly retires the 2023 and 2024 gigawatt targets. It also means checking whether the Garden City and 85 MW second-facility commitments were ever delivered, repurposed, or dropped.

Applied Digital's capacity story grew by an order of magnitude in seven months without a matching step up in delivered megawatts. The 2021 record centres on a single 100 MW energy services agreement at Jamestown (14511, 14413, 14373), with a loosely framed 500 MW of expansion demand (14512) and an unsited 85 MW second facility (14375, 14362, 14040). By March 2022 the ambition had jumped to 5 GW over five years (14230, 14033), with interim marks of 800 MW by May 2023 (14228, 14031) and 1.8 GW by May 2024 (14229, 14032), plus an Antpool joint venture for up to 1.5 GW (14235, 14203). The 85 MW second site then gives way to a 200 MW Garden City wind-powered facility (14236, 13765), and nothing in the record reconciles the two.

Dated promises have not held. First co-hosting completion by year-end 2021 was repeated in four filings (14509, 14414, 14372, 14364) and missed. What followed was 55 MW online by February 2022 (14008), and the remaining 45 MW was pushed to Q2 2022. That Q2 target was restated almost verbatim across March and April filings (14227, 14206, 14009, 14055), and it slipped too. The August 2022 claim of 100% operating capacity at Jamestown (13806) sits awkwardly beside July claims that temporary power would only be restored by September 2022 and permanent power by mid-2023 (13856, 13857), both of which were also delayed. Garden City's October 2022 opening (13777) and year-end 200 MW target (13765) slipped as well. The 800 MW and 1.8 GW interim targets lapsed without restatement in this record, which is revision by silence rather than acknowledgement.

Every claim here is company-sourced. The record has no permit filings, interconnection queue positions or third-party site reporting to corroborate any of the announced pace. The ledger shown also ends in August 2022. It carries no capacity, cost or date claims for Polaris Forge 1, 2 and 3 or Delta Forge 1 and 2, which are the projects that define the company's current positioning.

Over the next 90 days, the 5 GW-by-March-2027 target (14230) approaches its deadline, and any restatement or quiet retirement of it is worth logging. The larger gap is evidentiary. Before judging the Polaris Forge and Delta Forge buildout, the claim history from 2023 onward needs to be backfilled and matched against regional interconnection queue entries for the Ellendale and Harwood sites and parish-level permitting in Rapides Parish for Boyce.

Applied Digital's capacity narrative in this record inflates quickly from a single hosting site into a multi-gigawatt ambition before the first site was fully energized. The August 2021 framing paired a 100 MW energy services agreement with 500 MW of expansion demand (14511, 14512). By March 2022 that had become 800 MW by May 2023, 1.8 GW by May 2024 and 5 GW within five years (14228, 14229, 14230), alongside an up-to-1.5 GW Antpool joint venture (14212, 14235). Contracted, site-specific capacity grew far more modestly. The record shows Jamestown's 100 MW (14413), an 85 MW second facility with customers reserved but no named site (14375, 14040), and a 200 MW Garden City wind site that moved from letter of intent to fully contracted within five months (14236, 13745).

Dates have not held. The Q4 2021 Jamestown completion target was restated in near-identical language across four disclosures (14509, 14414, 14372, 14364) with no added caution. It was then quietly replaced by a split of 55 MW online and 45 MW by Q2 2022 (14068, 14206). That Q2 date slipped too. By July 2022 the company was describing temporary and permanent power restoration at Jamestown running out to mid-2023 (13856, 13857), even as 13806 in August cited a 100 percent operating milestone. The 800 MW and 1.8 GW dates are both marked delayed, and neither was explicitly withdrawn in this record.

Nothing in the supplied record counts as independent corroboration. The claims are self-reported filings and releases, with no permits, interconnection queue entries or on-the-ground reporting attached. The Antpool joint venture and Garden City figures in particular rest entirely on company statements.

The claim history supplied breaks off mid-entry on 2022-08-28. None of the Polaris Forge, Delta Forge or Ellendale projects listed on record carry capacity, cost or dated claims here, so this read cannot assess the AI and HPC pipeline that now defines the company. Over the next 90 days, the first priority is recovering the missing post-2022 claims. Next, check MISO queue positions for the Harwood and Boyce sites against announced energization dates. Finally, test whether Polaris Forge tenant lease commitments track the phased capacity the company has guided to. That last check matters because the pattern here is restated targets rather than acknowledged slips.

This record ends in late August 2022, cut off mid-entry at claim 13737. It therefore covers Applied Digital's hosting-era build-out and says nothing about how the Polaris Forge, Delta Forge or Ellendale projects have progressed. None of those projects has a capacity or cost attached. Ellendale and Polaris Forge 1 share the same address, which suggests a rename rather than two sites, and the two should be reconciled before either is scored.

The capacity story inflates quickly and then goes quiet. Through 2021 the company described a 100 MW energy services agreement at Jamestown (14511, 14413, 14373) and an 85 MW second facility (14375, 14362). By March 2022 the framing had jumped to 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW over five years (14230, 14033). A 1.5 GW Antpool joint venture sat alongside these (14235, 14036), as did a 200 MW Garden City letter of intent (14236) and a 200 MW Marathon hosting deal (13855). Both dated interim targets are marked delayed. The 5 GW figure is the only aggregate target still formally open, and its deadline is 2027-03-20.

Dates have slipped repeatedly, often with no reset. The Q4 2021 first-facility completion was restated unchanged on 12 August, 1 November, 21 November and 30 November 2021 (14509, 14414, 14372, 14364), with the last restatement coming a month before the deadline, which was then missed. Later disclosure put 55 MW online as of February 2022 (14008). The remaining 45 MW was promised for Q2 2022 in at least six filings (14227, 14206, 14196, 14069, 14055, 14009), and that date also slipped. Full Jamestown capacity is not claimed until August 2022 (13806). Even that milestone was followed by power-restoration dates for September 2022 and June 2023 (13856, 13857), and both were missed. At Garden City, the ground-lease and construction-start dates (13763, 13741) had already passed by the time they appeared in the August 2022 filing. The Bitmain 200 PH order was likewise repeated without revision until it lapsed (14504, 14416, 14378, 14360).

This record contains no independent evidence at all: no permits, no interconnection queue positions, no local reporting. Every figure above is company-sourced, so none of the announced pace can be treated as corroborated.

Over the next 90 days, the priority is to backfill claims after August 2022 and attach capacity figures to the Forge projects. Then check whether the 5 GW target (14230) is reaffirmed or quietly dropped. Finally, look for utility-commission filings and interconnection queue entries at Harwood and Boyce that independently confirm announced megawatts and energization dates.

Applied Digital's capacity story in this record grew quickly while delivery lagged it. The early claims were modest and tied to one site: a 100 MW energy services agreement at Jamestown (14511, 14413, 14373) and an 85 MW second facility with customers reserved (14375, 14040). By March 2022 the company was describing an 800 MW target by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW within five years (14230). It layered on a 1.5 GW Antpool joint venture (14203, 14036), a 200 MW Texas wind letter of intent (14236) and a 200 MW Marathon hosting deal (13855). Most of that added capacity rested on LOIs, joint-venture formations and partner agreements rather than energized megawatts.

The dates have not held, and some slipped without acknowledgement. The first co-hosting facility was promised for Q4 2021, and that promise was repeated as late as November 30 (14364, 14372). Its 55 MW came online in February 2022 (14008), and the company restated that milestone rather than labelling it a slip. The remaining 45 MW was then promised for Q2 2022 (14227, 14009) and was also delayed. A full-capacity milestone was claimed in August 2022 (13806), but it sits awkwardly beside claims from a month earlier that Jamestown had lost power and that permanent restoration was pushed to mid-2023 (13856, 13857). The Garden City construction start (13741) and ground lease (13763) first appear in an August 2022 filing carrying April 2022 dates that had already passed. Both the 800 MW and 1.8 GW targets are marked delayed. The SparkPool share forfeiture (13772) shows the original partner structure also unwinding.

The record offers no independent corroboration of the announced pace. There are no permits, no interconnection queue entries and no site reporting. The claim set supplied here also stops in August 2022. The later projects carry no capacity or cost fields at all: Polaris Forge 1 through 3, Delta Forge 1 and 2, and Garden City. That gap is itself a finding, and the current AI-campus narrative cannot be judged against this history until those claims are ingested.

Over the next 90 days, watch four things. First, whether any Polaris Forge or Delta Forge disclosure attaches dated energization milestones and megawatt figures to a specific building. Second, whether the 5 GW target due in March 2027 (14230) is reaffirmed, quietly retired or replaced by a new headline figure. Third, whether utility interconnection or state permitting records for Ellendale, Harwood and Boyce match the phasing the company describes. Fourth, whether past slips are acknowledged or simply restated as new baselines.

Note first that the claim record supplied here ends mid-list in August 2022, with claim 13737 cut off, so this read covers only Applied Digital's crypto-hosting era. The Polaris Forge and Delta Forge campuses carry no capacity or timeline claims in the extract. Any view of the current AI-campus trajectory needs the later claims reattached before publication.

Within that window, the capacity story inflated quickly and then detached from anything built. It began with a single 100 MW energy services agreement at Jamestown (14511, 14413) and a vaguer 500 MW demand figure (14512). By November 2021 it added an 85 MW second site with no named location (14375, 14362). In March 2022 it jumped to an 800 MW target by May 2023, 1.8 GW by May 2024 and 5 GW over five years (14228, 14229, 14230). Alongside those came a 1.5 GW Antpool joint venture (14235, 14212) and a 200 MW wind-sited Texas letter of intent (14236). None of the larger figures is tied to a site, a signed interconnection or capital. The 5 GW target was simply repeated a month later (14033) with no intermediate milestones.

Dates have not held, and slips were mostly absorbed rather than acknowledged. The first Jamestown facility was promised by year-end 2021 across four separate filings (14509, 14414, 14372, 14364) and reached 55 MW only in February 2022 (14008). The remaining 45 MW was then promised for Q2 2022 in at least six claims (14227, 14206, 14196, 14069, 14055, 14009). The next data point is a Jamestown power outage requiring temporary and permanent restoration (13856, 13857), with a 100% operating milestone asserted shortly after (13806). The Garden City ground lease and construction start appear only retroactively, with deadlines that had already passed when first cited (13763, 13741). The Bitmain 200 PH order (14504, 14416, 14378, 14360) slipped without restatement.

The record contains no permits, interconnection-queue positions or site reporting, so nothing independent supports the announced pace. The 800 MW and 1.8 GW deadlines lapsed while flagged delayed, with no revision on file.

Over the next 90 days, watch whether filings finally attach megawatts and energization dates to Polaris Forge 2, Polaris Forge 3 and Delta Forge 2. Check whether the Harwood and Boyce sites show up in utility or ISO interconnection queues and county permit records. Watch too whether management formally retires or restates the 5 GW horizon (14230) as it nears March 2027.

Applied Digital's capacity story in this record inflated fast. The 2021 filings describe a modest base: a 100 MW energy services agreement for the first facility at Jamestown (14413, 14361), 85 MW planned at a second site (14362), and a loosely framed 500 MW of expansion demand (14512). By March 2022 the ambition had grown by more than an order of magnitude, to 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), and 5 GW over five years (14230). On top of that sat a 1.5 GW joint venture with Antpool (14235) and a 200 MW wind-powered Garden City letter of intent (14236). Contracted power had not grown on anything like that scale. The only firm new commitment of comparable size is the 200 MW Marathon hosting agreement (13855).

The dates have not held. First-facility completion was promised for Q4 2021 in four successive filings (14509, 14414, 14372, 14364) and was only partly met: 55 MW was online as of February 2, 2022 (14008). The remaining 45 MW was pinned to Q2 2022 in at least five filings (14227, 14206, 14069, 14055, 14030), repeated word for word and then missed. The July 2022 disclosure of a power disruption at Jamestown set a September 2022 target for temporary restoration and a mid-2023 target for permanent restoration (13856, 13857), and both slipped. The 800 MW and 1.8 GW targets were restated in April 2022 (14031, 14032) without revision and later lapsed, and the record shows no acknowledgement of either. Garden City appears in six claims through April 2022 and then disappears.

Nothing independent in this record corroborates any of this. There are no permits, interconnection queue positions, or site reporting. Every capacity field on the project roster is blank. The Polaris Forge and Delta Forge projects are listed, but no claims tied to them appear in the supplied excerpt, which ends mid-entry at 13776. The current flagship program therefore cannot be assessed here, and this read covers only the 2021–2022 hosting era.

Over the next 90 days, the first thing to watch is whether the Polaris Forge and Delta Forge claims carry dated megawatt milestones that can be scored, rather than open-ended capacity language. The second is whether the company formally retires or restates the 5 GW target, whose five-year window closes in March 2027 (14230). The third is whether utility interconnection filings or local permitting for Ellendale, Harwood and Boyce appear, since they would be the first external check on the announced pace.

Read as a delivery record, Applied Digital's claims through August 2022 show headline capacity rising far faster than contracted power. The anchor is a single 100 MW energy services agreement at Jamestown, restated in nearly every filing (14511, 14413, 14361, 14039), plus an 85 MW second facility that never gets a location (14375, 14362, 14040). By March 2022 the ambition had become 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW within five years (14230). It was layered on a 1.5 GW Antpool joint venture (14235) and a 200 MW Garden City letter of intent (14236) that are never tied to a site, interconnection or cost. The 200 MW Marathon hosting agreement (13855) adds customer demand, not power. In seven months the stated target grew roughly fifty-fold while the energized base stayed at one 100 MW site.

Dates have not held, and slips are carried forward rather than acknowledged. The Q4 2021 Jamestown completion appears unchanged across four successive filings (14509, 14414, 14372, 14364), all now flagged delayed. The partial 55 MW energization dated February 2, 2022 (14008) was followed by the remaining 45 MW promised for Q2 2022 in six separate claims (14227, 14206, 14069, 14055, 14009, 14030), and that date also missed. By July, Jamestown needed temporary power restored by September 2022 (13856) and permanent power by mid-2023 (13857). That sits awkwardly against the August claim of a 100% operating-capacity milestone (13806), which the filings do not reconcile. The 800 MW and 1.8 GW targets both lapsed without a visible revision.

The supplied record contains no independent corroboration: no permits, no interconnection-queue positions, no on-the-ground reporting. The announced pace therefore rests entirely on company statements. The record also ends at 13776, so the Polaris Forge, Delta Forge, Ellendale and Harwood projects appear in the project list with no capacity, cost or claim history attached. Nothing here can speak to their trajectory, and they should not be judged on this record.

Over the next 90 days, the priority is to fill that gap. Pull the post-2022 claim history for the Polaris and Delta Forge sites. Check MISO queue and state utility-commission filings for the Ellendale, Harwood and Boyce locations. Look for whether the Garden City letter of intent or the Antpool venture was ever formally closed out. With the 5 GW target's March 2027 horizon (14230) now close, any restatement or quiet withdrawal of that figure is the clearest signal to watch.

The record supplied here ends in August 2022. None of the claims shown carry a Polaris Forge, Delta Forge or Ellendale tag, so this read covers the crypto-hosting era that preceded the AI campuses. The rest of the project list cannot be assessed from this input.

The capacity story is one of fast headline inflation built on a small operating base. In August 2021 the company described 100 MW of contracted power (14511) and 500 MW of expansion demand (14512). By March 2022 that had become 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW within five years (14230). The Antpool joint venture was billed as up to 1.5 GW (14235) against a stated $8,000 formation figure (14234). Beneath that, the concrete numbers moved the other way. The GMR arrangement first appeared as 100 MW (14502) and was later described as 50 MW (14377, 14359). The unnamed second facility stayed fixed at 85 MW (14375, 14362, 14040). The Garden City wind site was restated as a 200 MW letter of intent at least six times (14236, 14204, 14074, 14063, 14034, 14015) without advancing past that stage.

Dates have not held. The Q4 2021 Jamestown completion target (14364, 14372, 14414) slipped. The company reported 55 MW online as of February 2, 2022 (14008). The remaining 45 MW was then promised for Q2 2022 across seven separate claims (14227, 14206, 14196, 14069, 14055, 14009, 14030), and that target also lapsed. The July 2022 power disruption produced new temporary and permanent restoration dates (13856, 13857), both flagged delayed. An August 2022 statement of 100% operating capacity (13806) followed without a clear reconciliation to those restoration dates. The 800 MW and 1.8 GW targets are now past due and flagged delayed. We see no claim in this set that formally withdraws them.

This claim set contains no independent evidence, such as permits, interconnection queue positions or site reporting. The announced pace therefore rests on company statements alone. The repeated restating of unchanged letters of intent argues for caution rather than confidence.

Over the next 90 days, the useful checks are whether later filings tie the 2022 gigawatt targets to the Polaris Forge and Delta Forge programs or quietly retire them, and whether the 85 MW second facility maps to Ellendale. It is also worth confirming whether Garden City ever advanced past the letter-of-intent stage. Queue and permit records for Harwood and Boyce would show whether the newer campuses break from the earlier pattern of targets outrunning energized megawatts.

Applied Digital's capacity story runs in two registers that rarely meet. The operating register is modest and concrete. It begins with a 100 MW energy services agreement at Jamestown (14511, 14413, 14361), later amended and restated (14202, 14039). The aspirational register escalates quickly. A 500 MW expansion demand figure in August 2021 (14512) became, by March 2022, targets of 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032) and 5 GW over five years (14230, 14033). Alongside these sat an Antpool joint venture for up to 1.5 GW (14235, 14212, 14036) and a 200 MW Garden City wind LOI (14236, 14204, 14034). Not every figure grew. The GMR hosting relationship was framed at 100 MW in August 2021 (14502) and at 50 MW three months later (14377, 14359), with no explanation in the record.

On timing, the pattern is consistent slippage restated at face value. The Q4 2021 target for the first co-hosting facility and equipment delivery (14509, 14414, 14372, 14364, 14365) was missed. The record next shows 55 MW online as of February 2022 (14008, 14029). The remaining 45 MW carried a Q2 2022 date through at least seven filings (14227, 14206, 14196, 14069, 14055, 14009, 14030) without revision, and all are now flagged delayed. A July 2022 power disruption produced new temporary and permanent restoration dates (13856, 13857), and both also slipped. A 100% operating capacity milestone was claimed in August 2022 (13806), but it is hard to reconcile with a permanent-power date still ten months out. The 800 MW and 1.8 GW targets lapsed without any restatement in the claims supplied.

Independent corroboration is absent from this record. No permits, interconnection queue positions or on-the-ground reporting are attached to any claim, so the announced pace rests entirely on company statements. The more serious gap is structural. The supplied claims end in August 2022, and the entries for the current portfolio carry no capacity, cost or timeline claims at all. That covers Polaris Forge 1 through 3 at Ellendale and Harwood, and Delta Forge 1 and 2 in Boyce, Louisiana. The 2022 pattern of repeating a missed date unchanged across filings is the relevant prior for them, but any read of their delivery pace would be speculative until those claims are ingested.

Over the next 90 days, three things matter. First, whether the 5 GW horizon (14230) is formally restated, reduced or quietly dropped. Second, whether Polaris Forge and Delta Forge disclosures attach dated, megawatt-specific energization milestones that can be checked against MISO and SPP queue filings and county permits. Third, whether new dates carry forward unchanged across successive filings, as the Jamestown 45 MW target did.

The trajectory divides cleanly into two eras, and the seam between them is where the credibility question sits. The first era, roughly August 2021 through mid-2022, is a bitcoin co-hosting story anchored on a single 100 MW energy services agreement (14511, 14413, 14361) and an 85 MW second site (14375, 14362). Within seven months the public ambition moved from that base to 500 MW (14512, 13728), then 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), and 5 GW over five years (14230, 14033), with a further 1.5 GW attributed to a joint venture that existed on paper (14235, 14212, 14016). Contracted capacity grew by roughly a factor of two in that window; announced capacity grew by a factor of fifty.

The date record is where the pattern is clearest. Every commitment carrying a 2021-12-31 deadline — first facility completion (14509, 14414, 14372, 14364), the Bitmain 200 PH order (14504, 14416, 14378, 14360), equipment delivery (14510, 14365) — is marked delayed. By August 2022 the same milestone is restated retroactively as having occurred on 2022-02-02 (13722), and the JV formation as 2022-01-06 (13729), with no acknowledgement that either had been promised for the prior quarter. The 45 MW Jamestown expansion is the sharpest illustration: a single increment re-announced at least seven times across three weeks in spring 2022 (14227, 14206, 14196, 14069, 14055, 14030, 14009), all against a Q2 2022 date, all delayed. The Garden City letter of intent received the same treatment (14236, 14204, 14074, 14063, 14034) and the site still carries status=announced today. Note also the tension between the July 2022 power-loss commitments (13856, 13857, both delayed) and the 100% operating capacity claim three weeks later (13806).

Independent corroboration is the gap in this file: the entire current portfolio — Polaris Forge 1 through 3, Delta Forge 1 and 2, Ellendale — is recorded with null capacity, null cost, and no dated commitment, meaning nothing in the present-day story is falsifiable from company disclosure alone. Over the next ninety days, weight interconnection queue positions and county permit filings in Dickey and Cass Counties and Rapides Parish above press releases, watch whether Polaris Forge 3 and Delta Forge 2 acquire addresses, and treat any new gigawatt-scale figure as unproven until a queue entry or energization date attaches to it.

The striking feature of this record is not any single miss but the shape of the arc. Ambition inflates fastest when delivery is least verifiable. Within roughly seven months the stated ceiling moves from a single 100 MW energy services agreement (14511, 14413, 14361) to a 500 MW expansion appetite (14512), then to 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), and 5 GW over five years (14230, 14033), with a further 1.5 GW attributed to the Antpool joint venture (14235, 14212, 14016). By August 2022 the combined target is restated at 500 MW (13728). That is a substantial downward revision, and it arrives without any reconciliation to the gigawatt-scale figures it supersedes.

Dates behave similarly. A dense cluster of year-end 2021 commitments — first co-hosting facility completion, Bitmain equipment delivery, additional equipment (14509, 14510, 14414, 14372, 14364, 14360, 14416, 14378, 14365) — all carry delay markers. The 45 MW second phase at Jamestown is re-promised for Q2 2022 at least six times across March and April 2022 (14227, 14206, 14196, 14069, 14055, 14030, 14009) in near-identical language, which is repetition rather than progress. The most candid entries are the Jamestown power restoration claims (13856, 13857), where temporary power slips to September 2022 and permanent power to mid-2023; an operational interruption of that length sits awkwardly beside the 100 percent capacity milestone asserted weeks later (13806).

On corroboration, the record is thin by construction. All eight projects sit at status "announced" with capacity and cost fields empty, including Polaris Forge 1 through 3 and Delta Forge 1 and 2. Ellendale and Polaris Forge 1 share the same street address, which suggests renaming of one asset rather than two developments. No permit, interconnection, or site-visit evidence appears anywhere in this file, so the announced pace is currently self-attested.

Over the next ninety days, watch for three things: whether Polaris Forge 2, Polaris Forge 3 and Delta Forge 2 acquire named megawatt and dollar figures or remain placeholders; whether the Ellendale and Polaris Forge 1 identity is acknowledged rather than double-counted; and whether county permits in Dickey, Cass and Rapides parishes, plus interconnection queue positions, appear on a schedule consistent with the energization dates being implied.

The shape of this record is an escalation curve detached from an energization curve. Between August 2021 and March 2022 the announced ambition moved from a single 100 MW energy services agreement (14511, 14413) and an 85 MW second site (14375, 14362) to a 500 MW combined target (14512, 13728), then to 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), and 5 GW over five years (14230, 14033). That is roughly a fiftyfold expansion of the stated pipeline in seven months, during which the verifiable operating number moved from zero to 55 MW at Jamestown (14054, 14008, 14029). Ambition scaled; delivery did not.

Dates behaved in a specific and telling way: they were repeated rather than revised. The Q4 2021 co-hosting completion appears essentially unchanged across four separate disclosures (14509, 14414, 14372, 14364), and the 45 MW Jamestown phase carries a Q2 2022 date through at least seven restatements (14227, 14196, 14206, 14069, 14055, 14009, 14030). Restating a date that is already at risk is not the same as reaffirming it. The 800 MW and 1.8 GW targets subsequently lapsed with no retraction anywhere in this record. The one candid slip is Jamestown's power restoration (13856, 13857), and that was forced by an outage rather than volunteered.

On corroboration, the record is thin by construction. Garden City enters the capacity math at letter-of-intent stage (14236, 14204, 14063, 13762) and its opening then slips past October 2022 (13777); an LOI is not a signed interconnection position, and counting it toward a gigawatt total inflates the denominator of any later hit rate. No permit numbers, queue IDs, or site reporting appear against any claim here.

Note also that the claim series ends in 2022 while the project list runs through Polaris Forge 1 through 3 and Delta Forge 1 and 2, all carrying blank capacity and cost. Over the next ninety days the things worth checking are whether Ellendale and Polaris Forge 1, which share 9685 87th Ave SE, are one asset presented as two; whether Harwood and Boyce acquire actual queue entries and county permits; and whether the next disclosure separates energized megawatts from contracted and leased ones, since this history shows those three numbers being spoken as if interchangeable.

The arc here is a company that changed the unit of measure faster than it changed what it had built. Through late 2021 the announced scope was modest and specific: a 100 MW energy services agreement (14511, 14413, 14373, 14361) and an 85 MW second site (14375, 14362). By March 2022 the framing jumped two orders of magnitude to 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), and 5 GW by 2027 (14230, 14033), plus a 1.5 GW joint venture with Antpool (14234, 14203, 14212, 14036). The escalation was not accompanied by any corresponding growth in the delivered base, which was 55 MW at the time (14008, 14029, 14054).

On dates, the pattern is repetition rather than revision. The Q4 2021 co-hosting completion was restated essentially verbatim across four disclosures (14509, 14414, 14372, 14364), as was the Bitmain 200 PH order (14504, 14416, 14378, 14360). The 45 MW Jamestown phase promised for Q2 2022 was reiterated at least seven times in under a month (14227, 14196, 14206, 14069, 14055, 14009, 14030). None of these carry an acknowledged slip; the target simply reappears. The sharpest inconsistency is the July 2022 power loss at Jamestown, where temporary restoration was pledged for 12 September 2022 (13856) and permanent for mid-2023 (13857), while an August disclosure announced a 100 percent operating capacity milestone at the same site (13806). Separately, the Texas 200 MW wind letter of intent was carried as a capacity line for five months (14236, 14204, 14214, 14074, 14063, 14034, 14015, 13778) without converting into a signed agreement in this record.

On corroboration, this record contains no permit filings, interconnection queue entries, or on-the-ground reporting, so the announced pace is uncorroborated rather than contradicted. Note also that Ellendale and Polaris Forge 1 share one address, so the eight-project list overstates site count, and all eight carry no capacity or cost figures at all.

Over the next 90 days, watch whether Polaris Forge 2, Polaris Forge 3 and Delta Forge 2 acquire addresses and quantified capacity; whether Boyce moves from announcement to a filed utility agreement or MISO large-load queue position; and whether the 2027 5 GW target (14230) is reaffirmed or allowed to lapse unmentioned as its window narrows.

The arc here is a familiar one for the 2021-22 crypto-hosting cohort: capacity ambition expanded far faster than delivered megawatts, and the announced numbers grew by aggregation rather than by construction. The earliest framing is modest and concrete — a 100 MW energy services agreement and an 85 MW second site (14511, 14375, 14362). Within roughly four months that becomes a 1.5 GW joint-venture envelope with Antpool (14235, 14203), a 200 MW Texas letter of intent (14236, 14204), and staged corporate targets of 800 MW by May 2023 and 1.8 GW by May 2024 en route to 5 GW (14228, 14229, 14230). Note the quality of those increments: an LOI, a JV ceiling, and a five-year aspiration are stacked into the same capacity narrative as energized load. The only figure that ever hardens is 55 MW actually online at Jamestown (14008, 14054, 14029), against which the 45 MW balance was promised for Q2 2022 and re-promised at least five times in six weeks (14227, 14206, 14196, 14069, 14055, 14030, 14009) — repetition of a date is not confirmation of it.

Dates have not held. The Q4 2021 cluster — first co-hosting completion and Bitmain equipment delivery — slipped in place without a stated revision, restated verbatim on 2021-11-01, 11-21 and 11-30 (14509, 14414, 14372, 14364; 14504, 14416, 14378, 14360). The 2022-07-17 power-restoration items (13856, 13857) are the first admission that the energized base itself was impaired, which recontextualizes the 100 percent capacity claim a month later (13806) and the Texas October 2022 opening (13777).

On independent corroboration, this record contains none: every entry is company-sourced. Nothing here is matched to a utility interconnection position, a county building permit, or site reporting, and the supplied file terminates in August 2022, so the Polaris Forge, Delta Forge and Ellendale programs sit on the roster with capacity and cost blank and cannot be assessed.

For the next 90 days, watch whether energized megawatts are reported separately from contracted and LOI megawatts, whether the 45 MW balance and Texas are ever closed out explicitly, and whether any Polaris or Delta figure is tied to a named interconnection agreement rather than a ceiling.

The trajectory divides cleanly into a contracted phase and an aspirational one, and the gap between them widens fast. Through late 2021 every capacity figure traces to something signed: a 100 MW energy services agreement restated four times (14511, 14413, 14373, 14361) and an 85 MW second site (14375, 14362). By March 2022 the announced ceiling had moved to 800 MW within fourteen months (14228, 14031), 1.8 GW by May 2024 (14229, 14032), and 5 GW by 2027 (14230, 14033), with the incremental volume resting on a joint venture vehicle (14235, 14203, 14036) and a letter of intent (14236, 14204, 14074) rather than executed power contracts. That is roughly a twenty-seven-fold escalation in announced scale over seven months, funded by instruments that carry no delivery obligation.

On dates, the tell is repetition rather than revision. The Q4 2021 first-facility completion was re-asserted at the same deadline across four separate disclosures (14509, 14414, 14372, 14364) as that deadline approached and passed. The residual 45 MW then absorbed the same treatment at Q2 2022, restated at least six times (14227, 14196, 14206, 14069, 14055, 14009, 14030). Jamestown did reach 55 MW (14054, 14008, 14029) and was described at full capacity in August 2022 (13806), which sits uneasily beside the July 2022 power loss requiring temporary restoration (13856) and permanent restoration a year out (13857), both since slipped. Garden City's 200 MW by year-end 2022 (13765) slipped likewise, as did the Bitmain equipment order across all four of its restatements (14504, 14416, 14378, 14360).

On independent corroboration I can say only that the record supplied contains none. Every datapoint here is company-sourced and largely self-referential; the sole externally verifiable artifact is the amended utility services agreement (14202, 14013, 14039). Interconnection queue positions and county permits would settle the question and are not in evidence.

For the next ninety days: the project roster lists eight sites with capacity and cost blank, and Ellendale and Polaris Forge 1 share one address, so watch for rebranded capacity being counted twice. Watch for a first named in-service date at Ellendale, a named offtaker, and whether a missed date is again restated rather than acknowledged.

The arc here is one in which announced capacity outran energized capacity by a widening multiple. The company entered public view anchored to a single 100 MW energy services agreement and an 85 MW second site (14511, 14413, 14373, 14375, 14362). Within roughly seven months the headline figures scaled to 800 MW by May 2023, 1.8 GW by May 2024, 5 GW over five years, and a further 1.5 GW through the Antpool joint venture (14228, 14229, 14230, 14235, 14203). Over that same window the actually-operating number moved from zero to 55 MW (14008, 14054). The second facility was also quietly restated upward from 85 MW to 200 MW (13765) rather than reconciled against the earlier figure.

On dates, the pattern is re-announcement rather than acknowledged slippage. Q4 2021 first-facility completion was promised four separate times without the prior miss being addressed (14509, 14414, 14372, 14364). The residual 45 MW was then re-anchored to Q2 2022 and restated at least six times across three weeks of March and April 2022 filings (14227, 14196, 14206, 14069, 14055, 14009, 14030) — repetition that reads as progress but conveys no new increment. The Bitmain 200 PH order carried an unchanged 2021-12-31 date through four restatements (14504, 14416, 14378, 14360). Both aggregate targets, 800 MW and 1.8 GW, are now recorded as missed (14031, 14032).

Independent corroboration is thin because nearly all of this is self-reported. The most telling internal tension is the July 2022 disclosure that Jamestown needed temporary power by September 2022 and permanent power only by June 2023 (13856, 13857), followed three weeks later by a claim of 100 percent operating capacity (13806). The Garden City position remained a letter of intent through five restatements (14236, 14204, 14214, 14034, 14063); an LOI restated is not an executed interconnection.

For the next ninety days, the gap in the record is the signal. Polaris Forge 1 through 3 and Delta Forge 1 and 2 sit on the roster with no capacity or cost populated, and Ellendale and Polaris Forge 1 share one address, which invites double-counting of announced megawatts. Watch whether Polaris Forge 3 and Delta Forge 2 acquire addresses before they acquire numbers, whether North Dakota and Louisiana interconnection and permit dockets show filings matching the announced pace, and whether disclosure shifts to energized megawatts rather than contracted or leased capacity.

The record as supplied runs from August 2021 to August 2022 and stops there, which matters: the eight projects currently on the books, including Ellendale, the three Polaris Forge sites and the two Delta Forge sites, carry no capacity or cost figures in this dataset and are unaudited here. What can be assessed is the earlier cycle, and the pattern in it is instructive.

Announced capacity scaled roughly fiftyfold in seven months while delivered capacity barely moved. The 2021 disclosures are built around a single 100 MW energy services agreement (14511, 14413, 14361) and an 85 MW second site (14375, 14362). By March 2022 the same company was publishing an 800 MW target for May 2023 (14228), 1.8 GW for May 2024 (14229), 5 GW over five years (14230), and a further 1.5 GW through the Antpool joint venture (14235, 14203). Against that, the energized base was 55 MW (14054, 14008). The headline number grew; the gap to delivery grew faster.

Dates did not hold, and the manner of failure is the more telling issue. The Q4 2021 completion target was restated essentially verbatim across four successive disclosures (14509, 14414, 14372, 14364) rather than revised. Jamestown's remaining 45 MW was promised for Q2 2022 in at least seven places in March and April 2022 (14227, 14196, 14206, 14069, 14055, 14009, 14030), carried forward unchanged as the quarter ran out. The Bitmain 200 PH order behaves identically (14504, 14416, 14378, 14360). Repetition without acknowledgement is the house style, not an isolated miss.

Independent corroboration is thin by construction. The large figures are anchored to letters of intent, energy services agreements and JV formations. The Texas site enters as an LOI (14236, 14204) and is described with a firm end-2022 operational date months later (13765). None of these are permits or interconnection positions. The July 2022 outage disclosures (13856, 13857) and the 100 percent capacity claim three weeks afterward (13806) warrant reconciliation against utility filings rather than being read together.

Over the next ninety days, watch whether Polaris Forge and Delta Forge disclosures separate contracted from energized megawatts, whether dates move explicitly or silently between quarterlies, and whether queue entries and county permits appear on the timeline the announcements imply.

Read across the thirteen months this record covers, the defining pattern is that announced capacity scaled far faster than delivered capacity, and the two were rarely reconciled in the same breath. The starting commitment was a single 100 MW energy services agreement (14511, 14413) plus an 85 MW second site (14375, 14362). By late March 2022 the same company was describing a 200 MW Texas letter of intent (14236), a joint venture sized at up to 1.5 GW (14235, 14203), 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW inside five years (14230). That is roughly a fiftyfold expansion of the headline figure over about seven months, against an operating base that moved to 55 MW (14008, 14054) with the remaining 45 MW still pending (14009, 14030, 14069). Announcement velocity and construction velocity decoupled early.

On dates, the more instructive behaviour is the restatement without acknowledgement. The first co-hosting facility was promised for Q4 2021 in August, then again in November, then twice more in successive November disclosures (14509, 14414, 14372, 14364), each time as a forward commitment rather than a revision. The Bitmain 200 PH delivery followed the same cadence (14504, 14416, 14378, 14360). When the Q4 target lapsed, it was not retired; it was replaced by a 45 MW Q2 2022 phase (14196, 14206), which also slipped. The 800 MW and 1.8 GW marks are both flagged delayed, as is the 200 MW second facility due end-2022 (13765).

The corroboration question is where this record is weakest. Every entry here is issuer-sourced and self-characterised; none carries permit, interconnection or field verification. Two internal signals cut against the announced pace: Jamestown was described at full operating capacity in August 2022 (13806) while still carrying unresolved temporary and permanent power restoration items (13856, 13857), and Garden City remains logged as announced with no capacity or cost attached.

Over the next ninety days, watch whether Ellendale, the Polaris Forge sites and the Delta Forge sites acquire their first dated, megawatt-specific commitments rather than status labels; whether disclosure language distinguishes energised load from leased or contracted capacity; and whether any prior date is explicitly withdrawn rather than superseded.

The arc here is one of escalating headline numbers detached from the delivery cadence underneath them. The company entered the record in August 2021 anchored to a single 100 MW energy services agreement (14511, 14413) and an 85 MW second site (14375, 14362). Within seven months the framing had jumped two orders of magnitude: 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), 5 GW within five years (14230, 14033), plus a 1.5 GW joint venture with Antpool (14235, 14203). None of that growth was accompanied by a corresponding densification of near-term, dated commitments. The dated commitments that did exist are almost uniformly marked delayed, and they cluster into two cohorts: the Q4 2021 group (14509, 14510, 14504, 14414, 14372, 14365) and the Q2 2022 group, where the same residual 45 MW at Jamestown was re-promised at least six times across five weeks (14227, 14196, 14206, 14069, 14055, 14009, 14030) without any acknowledgement that the prior statement had been a restatement rather than progress. Repetition was doing the work that completion was not.

The Jamestown power loss is the clearest test case, and it failed openly: temporary restoration promised by September 2022 (13856) and permanent by mid-2023 (13857) both slipped, after the same site had been described at full operating capacity a month earlier (13806). The pivot to Ellendale at 180 MW (13746) and later to the Polaris Forge and Delta Forge families reads as a change of subject as much as a change of business.

On corroboration, the honest finding is negative: every claim in this record is company-sourced. There are no permit filings, interconnection queue positions, or independent site reports attached to any id, and two live projects, Polaris Forge 3 and Delta Forge 2, carry no address at all, which is the earliest reliable tell that a named project is a placeholder. Over the next ninety days, watch whether Polaris Forge 3 and Delta Forge 2 acquire parcels and utility study entries; whether Boyce moves from announcement to an energization date that survives a second quarter; and whether any Ellendale or Harwood capacity figure is restated downward quietly rather than withdrawn.

This read covers only part of Applied Digital's record. The claim list supplied here stops in late August 2022, mid-entry at claim 13725, while the project roster names Polaris Forge 1 through 3 and Delta Forge 1 and 2 with no attached claims. The conclusions below are provisional until the later record is attached.

Capacity language grew far faster than delivered megawatts. The August 2021 baseline was a 100 MW energy services agreement at Jamestown (14511) against 500 MW of stated demand (14512). By March 2022 that had become 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW over five years (14230), plus an Antpool joint venture sized at up to 1.5 GW (14235). The first two dated targets are now marked delayed. The second site grew from 85 MW (14375, 14040) to 180 MW once it was identified as Ellendale (13746). Commitments from counterparties moved the other way. GMR went from a 100 MW services agreement (14502) to a 50 MW hosting contract (14377), and the Garden City 200 MW wind letter of intent (14236) was repeated across four filings without ever becoming a definitive agreement.

Dates slipped repeatedly, and the slips were often softened in the retelling. First-facility completion, promised for Q4 2021 (14509, 14414, 14364), became 55 MW online as of February 2022 (14008), and the remaining 45 MW was then pushed into Q2 2022 (14227, 14009), which also slipped. After the July 2022 power event, temporary and permanent restoration dates (13856, 13857) were both missed. Yet claim 13806 in August 2022 described a 100% operating capacity milestone at Jamestown, which is hard to square with those missed dates. Claim 13736 restates the Bitmain partnership with a July 2021 deadline that predates the first claim on file, which suggests a backdated framing. The Bitmain 200 PH equipment order (14504, 14378) was never shown as fulfilled.

No independent evidence is on record here: no state siting filings, no utility or MISO interconnection queue entries, and no on-site reporting. The announced pace therefore rests entirely on company statements.

Over the next 90 days, three things matter. First, whether the Polaris Forge and Delta Forge claims show the same pattern of large headline figures followed by partial energization. Second, whether Ellendale and Polaris Forge 1, which share an address, are reported as one capacity pool rather than counted twice. Third, how the company frames the 5 GW target as its March 2027 horizon (14230) comes into view.

The visible record for Applied Digital runs only from August 2021 to August 2022 and cuts off mid-entry at claim 13725. It therefore shows the crypto-hosting origins of the Ellendale site but none of the Polaris Forge or Delta Forge claims. The two project rows that share the 9685 87th Ave SE address, Ellendale and Polaris Forge 1, appear to be a rename rather than two facilities, and should be merged before any cross-period comparison. Within the visible window, announced capacity scaled quickly on paper. A single 100 MW Jamestown energy services agreement (14511, 14413) grew to include an 85 MW second site (14375, 14040) and then a 180 MW figure for North Dakota tied to Ellendale (13746). Beside these sit portfolio targets of 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032) and 5 GW over five years (14230). The later restatement of the 5 GW goal drops its end date (14033). The Garden City 200 MW wind LOI was repeated at least seven times between March and April 2022 (14236 through 14034) with no change in language, and the record shows no step toward a definitive agreement. The same applies to the 1.5 GW Antpool joint venture (14235, 14212, 14016). Dates have repeatedly slipped without being revised. The first co-hosting facility was restated for Q4 2021 as late as 30 November 2021 (14364), weeks before it missed. The remaining 45 MW at Jamestown kept its Q2 2022 target in every filing from March to April 2022 (14227, 14069, 14009) and is marked delayed. After the July 2022 power disruption, temporary and permanent restoration dates (13856, 13857) both slipped. The August 2022 claim of 100 percent operating capacity (13806) sits uneasily next to them. Claim 13736 is also odd, because it attaches a 2021 due date to a partnership first cited in 2022. The record contains no independent corroboration of any kind: no permits, no utility or ISO queue positions, no site reporting. The delivery pace therefore rests entirely on company statements. Over the next 90 days, the priorities are to recover the truncated post-2022 claims, reconcile Polaris Forge 1 capacity and energization statements against utility interconnection filings and county permits in Dickey County, and check whether Harwood and Boyce show comparable third-party paper. Tenant-lease disclosures should also be compared with the megawatts the company reports as actually energized.

The record supplied here runs from August 2021 to late August 2022 and cuts off mid-claim at 13725. The Polaris Forge and Delta Forge projects appear only as project entries with no capacity, cost, or dated claims attached. This read therefore covers the Jamestown-era trajectory and the ambitions announced alongside it. It cannot assess the current AI-campus build.

Capacity language grew quickly while delivered capacity grew slowly. The company opened with a 100 MW energy services agreement (14511, 14413) and an 85 MW second facility (14375, 14040). Within roughly seven months that became 800 MW by May 2023 (14228, 14031), 1.8 GW by May 2024 (14229, 14032), and 5 GW over five years (14230). The 1.21 Gigawatts joint venture with Antpool added another 1.5 GW (14235, 14203). When the 5 GW target was restated (14033), it had lost its end date. Meanwhile, what was actually delivered was 55 MW at Jamestown as of February 2022 (14008, 14054). Garden City's 200 MW wind site stayed a letter of intent across at least six restatements (14236, 14214, 14063, 14034) and never became a definitive agreement. The joint venture likewise shows no follow-on capacity or site claims.

Dates have slipped repeatedly, and mostly without being acknowledged. The first co-hosting facility was promised for Q4 2021 in four successive filings (14509, 14414, 14372, 14364). The remaining 45 MW was restated for Q2 2022 at least six times (14227, 14196, 14069, 14055, 14009, 14030), all now marked delayed, with no revised date on record. The July 2022 power restoration claims (13856, 13857) are also delayed. Yet the 100% operating capacity milestone (13806) was asserted weeks later, and that tension goes unexplained. The Bitmain partnership resurfaces in 13736 with a July 2021 promised-by date, which predates the original December 2021 commitment (14504). That looks like a retroactive reframing rather than a delivery.

The record contains no independent corroboration of any claim: no permits, no interconnection queue positions, and no site reporting. Every project on file remains in announced status with blank capacity and cost fields.

Over the next 90 days, three things matter. First, whether Polaris Forge 1 and 2 claims arrive with energization dates and lease-commencement evidence rather than headline megawatts. Second, whether MISO queue entries or county and state permits exist for Harwood and Boyce. Third, whether the company ever formally retires the 800 MW, 1.8 GW, and Garden City commitments instead of letting them lapse silently.

The record supplied here ends mid-list at claim 13725 (August 2022). It carries no capacity or timeline claims for Polaris Forge 1–3 or Delta Forge 1–2, even though those projects appear on the roster. This read therefore covers the 2021–2022 crypto-hosting era, not the current AI-campus buildout.

The announced ambition grew very quickly. In August 2021 the anchor was a single 100 MW energy services agreement at Jamestown (14511, 14413), with 500 MW of expansion demand noted (14512) and an 85 MW second facility (14375, 14362). By March 2022 the company was describing 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229), 5 GW within five years (14230), a 1.5 GW Antpool joint venture (14235) and a 200 MW Garden City wind LOI (14236). By August 2022 it added Ellendale at 180 MW (13746). Over those months, headline scope rose by more than an order of magnitude, while delivered capacity stayed at or below the original 100 MW.

Dates generally slipped without being acknowledged. First co-hosting completion was guided to Q4 2021 in four successive documents (14509, 14414, 14372, 14364). The first reported delivery was 55 MW as of February 2, 2022 (14008). The remaining 45 MW was restated for Q2 2022 in at least six filings (14227, 14196, 14069, 14055, 14009, 14030) without revision. It is only partly resolved by the 100% operating milestone (13806), which sits beside a July 2022 power-restoration sequence whose temporary and permanent dates were both missed (13856, 13857). The 800 MW and 1.8 GW targets are flagged delayed, and the record never shows them being withdrawn. Claim 13736 back-dates a Bitmain milestone to July 2021, which conflicts with the December 2021 date in earlier claims (14504, 14378).

The supplied record contains no independent corroboration: no permits, interconnection queue entries or site reporting. The Garden City LOI and the 1.21 Gigawatts JV are repeated across filings with no sign of progress. They should be treated as unverified.

Over the next 90 days, the first priority is closing the data gap: sourced capacity and energization dates for Polaris Forge 1, 2 and 3 and for Delta Forge. It is also worth checking whether Ellendale's 180 MW and Polaris Forge 1 share one address and the same capacity lineage. Queue filings for Harwood and Boyce would provide independent evidence. Finally, watch whether Garden City, the Antpool JV or the 5 GW target resurface or quietly disappear.

Applied Digital's capacity narrative grew much faster than its delivered megawatts. It began in August 2021 as a single 100 MW energy services agreement at Jamestown (14511, 14413) with a loosely framed 500 MW of expansion demand (14512). By March 2022 it had become an 800 MW target by May 2023 (14228), 1.8 GW by May 2024 (14229), and 5 GW over five years (14230), along with a 1.5 GW Antpool joint venture (14235) and a 200 MW Texas wind LOI (14236). The second facility went from 85 MW (14375, 14362, 14040) to 180 MW at Ellendale (13746) with no stated bridge between the two figures. Headline ambition multiplied roughly fifty-fold in seven months. The only operating asset on record over that period is one 100 MW site.

The dates have mostly slipped, and the slips were often not acknowledged. The Q4 2021 target to complete the first co-hosting facility was repeated across four filings (14509, 14414, 14372, 14364) and missed. It later reappeared as 55 MW online as of February 2, 2022 (14008), with the remaining 45 MW moved to Q2 2022 (14227, 14196, 14055, 14009, 14030). That was a partial delivery reframed as a milestone rather than disclosed as a miss. The 45 MW was then overtaken by a Jamestown power disruption, with temporary restoration promised by September 2022 (13856) and permanent restoration by June 2023 (13857). Both are flagged as delayed, yet a 100 percent operating capacity milestone was also claimed in August 2022 (13806). Those two claims need reconciling. The Bitmain 200 PH order slipped past end-2021 in four iterations (14504, 14416, 14378, 14360), and a later filing attaches a backdated July 2021 date to the same relationship (13736). The 800 MW and 1.8 GW targets both lapsed.

The record contains no independent corroboration: no permits, interconnection queue positions, or site reporting. The Garden City LOI appears unchanged in at least seven filings (14236, 14214, 14204, 14074, 14063, 14015, 14034) and never moves past letter-of-intent status. The Antpool joint venture is restated repeatedly without a named site. Repetition is doing the work that milestones should.

The claim history supplied ends mid-2022, truncated at claim 13725. Polaris Forge 1 through 3 and Delta Forge 1 and 2 carry no capacity, cost, or timeline claims here, so the current AI-campus buildout cannot yet be assessed against its own promises. Over the next 90 days, three priorities stand out. First, backfill the 2023 to 2026 claims, especially any energization dates for Polaris Forge 1 at the Ellendale address, which is shared with the original Ellendale project. Second, pull utility interconnection filings and county permits for Harwood and Boyce. Third, check whether the next quarterly report ties announced megawatts to lease commencements rather than to letters of intent.

This extract ends mid-record at claim 13776 in August 2022, so this read covers only Applied Digital's crypto-hosting era. It contains no claims tied to Polaris Forge 1, 2 or 3, Delta Forge 1 or 2, or Ellendale, which leaves the company's current AI-campus trajectory unaddressed here. Within the window we have, the pattern is fast capacity inflation on top of slow physical delivery. In August 2021 the anchor was a 100 MW Jamestown energy agreement (14511) and a 500 MW demand pipeline (14512). By March 2022 that had become 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW over five years (14230), plus a 1.5 GW Antpool venture (14235) that never had a site attached. Some partner commitments moved the other way. The GMR hosting figure fell from 100 MW (14502) to 50 MW (14377, 14359) within three months, and that cut was never acknowledged as one. The second facility moved from an unnamed 85 MW site (14375, 14040) to a 200 MW Garden City commitment (13765) without the change being explained.

Dates did not hold. The first co-hosting facility and the equipment deliveries were promised for end-2021. That promise was repeated as late as November 30 (14364, 14365) and then missed. Only 55 MW was live by February 2022 (14008). The remaining 45 MW was pinned to Q2 2022 across five filings (14227, 14206, 14069, 14009, 14030), and each one slipped. The July 2022 power-restoration dates (13856, 13857) sit uneasily next to an August claim of 100 percent operating capacity (13806), and that tension deserves a direct reconciliation. Claims 13749 and 13729 carry promise dates that predate their own filing. They look like retrospective references miscoded as delays rather than true slips.

The record includes no independent corroboration: no permits, no MISO or ERCOT interconnection queue positions, and no on-the-ground reporting. The announced pace therefore rests entirely on company statements, and the multi-gigawatt targets were missed without any on-record revision.

Over the next 90 days, the first priority is to load the post-August-2022 claims and attach them to the Polaris Forge and Delta Forge projects. Ellendale and Polaris Forge 1 share an address and appear to be one site under two names, so they should be merged. For Harwood and Boyce, the MISO queue and county building permits would test whether energization dates track the announcements. The key question is whether the 2022 habit of repeating dates up to the deadline has carried into the lease-backed AI buildouts.

The record supplied here breaks off in August 2022, mid-entry at claim 13777. What follows is therefore a read of Applied Digital's crypto-hosting origin phase. The Polaris Forge, Harwood and Delta Forge campuses on the project roster have no dated capacity, cost or timeline claims in this input, so they cannot yet be judged. Within the window that is covered, announced capacity grew much faster than delivered capacity. The company opened with a 100 MW energy services agreement at Jamestown (14511, 14413) and a loosely framed 500 MW of expansion demand (14512), then added an 85 MW second site (14375, 14362). By March 2022 the ambition had jumped to 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW within five years (14230). Alongside those targets came a 1.5 GW joint venture with Antpool (14235, 14212) and a 200 MW Garden City wind LOI (14236). At that point the only energized capacity cited was 55 MW (14008). The Garden City LOI was restated at least five times (14214, 14063, 14074, 14015, 14034) without ever moving past letter-of-intent language. Dates slipped repeatedly, usually through restatement rather than acknowledged revision. The Q4 2021 Jamestown completion target (14509, 14414, 14372, 14364) became 55 MW online plus 45 MW by Q2 2022 (14196, 14069, 14009), and that date also missed. The Bitmain 200 PH delivery was repeated unchanged across four filings (14504, 14416, 14378, 14360) after it had plainly become unreachable. After the July 2022 outage the company promised temporary power by September 2022 (13856) and permanent restoration by June 2023 (13857), and both slipped. Yet an August 2022 claim of 100 percent operating capacity (13806) sits uneasily beside an unresolved permanent-power timeline. The 800 MW and 1.8 GW targets both lapsed without any recorded retraction. No independent corroboration appears in the record. It contains no permits, no MISO or ERCOT interconnection queue positions and no on-the-ground reporting, so every figure above is company-sourced. For the next 90 days, the first task is to ingest the post-2022 claims so the AI-campus pivot can be tested against this history. The Polaris Forge and Delta Forge energization dates should then be matched against MISO queue entries and county permitting, including Ellendale and Harwood in North Dakota and Rapides Parish in Louisiana. The 5 GW target is dated 2027-03-20 (14230), about six months out. The next filing should reveal whether it is quietly carried forward, reframed as contracted rather than energized capacity, or formally retired.

The record for Applied Digital is really two eras, and only the first is documented. Every dated claim ends in August 2022. The current project slate of Polaris Forge 1 through 3, Delta Forge 1 and 2, and Garden City carries no capacity, cost or timeline claims at all. Any read of today's delivery pace therefore rests on the company's crypto-hosting origins, and the gap itself is the main finding.

In that early period, the firm commitment held steady while the ambitions grew quickly. The utility energy services agreement stayed at 100 MW from the first filing through April 2022 (14511, 14413, 14373, 14062). The second facility grew from 85 MW (14375, 14040) to 200 MW (13765). GMR's hosting commitment moved the other way, halving from 100 MW (14502) to 50 MW (14377, 14359). The aspirational figures escalated in a single March 2022 release. A 500 MW demand figure (14512) became 800 MW by May 2023 (14228), 1.8 GW by May 2024 (14229) and 5 GW over five years (14230), plus a 1.5 GW Antpool joint venture (14235) that never acquired a date. Both dated targets are now marked delayed.

The dates show a pattern of repetition rather than revision. The Q4 2021 completion of the first co-hosting facility was restated four times without adjustment (14509, 14414, 14372, 14364). The same happened to the 200 PH Bitmain order (14504, 14416, 14378, 14360). The remaining 45 MW at Jamestown was promised for Q2 2022 in five separate claims (14227, 14206, 14055, 14030, 14009), all slipped. By July, the site needed temporary power restoration by September 2022 (13856) and permanent restoration by mid-2023 (13857). That sits awkwardly beside an August claim of 100 percent operating capacity (13806). The second facility's ground lease (13723), construction start (13741) and October 2022 operating date (13726) were all reported only after they had lapsed. The SparkPool partnership was carried as active (14358, 14019) until it ended in a share forfeiture (13772).

The record holds no independent corroboration: no permits, interconnection queue positions or site reporting. The Jamestown power interruption is the only external signal, and it points to utility-side constraints rather than faster delivery. Ellendale and Polaris Forge 1 share an address, which suggests a duplicate or renamed record.

Over the next 90 days, three things matter. The first is to backfill claims from fiscal 2023 onward so that Polaris Forge capacity statements can be compared with the 2022 targets. The second is to check the MISO and state queue entries for Harwood and Boyce. The third is to see whether the next quarterly report gives energized megawatts or only contracted megawatts.

This record stops partway through claim 13772, dated August 2022. The eight named projects, including the Polaris Forge and Delta Forge sites, have no capacity or cost figures and no claims attached. So this is a read of Applied Digital's early hosting phase, not its current AI campus pipeline. Ellendale and Polaris Forge 1 share a street address and are likely one site recorded twice.

The capacity story moves fast in both directions. The first-facility anchor held steady at 100 MW across 2021 and 2022 (14511, 14413, 14361, 14062). The GMR commitment fell from 100 MW (14502) to a 50 MW hosting contract within three months (14377, 14359), and the change was never explained. The second North Dakota facility grew from 85 MW (14375, 14040) to 180 MW (13779). In March 2022 the company jumped from single-site figures to 800 MW by May 2023, 1.8 GW by May 2024 and 5 GW over five years (14228, 14229, 14230). It repeated those targets in April (14031, 14032, 14033) without citing new power contracts beyond a 200 MW Texas letter of intent (14236) and the Antpool joint venture's 1.5 GW ceiling (14235).

Dates slipped quietly. The first facility was promised by year-end 2021 in four separate claims (14509, 14414, 14372, 14364). It was later reported online at 55 MW as of February 2, 2022 (14008). Claim 13722 then recast February 2 as the promised date, which erases the slip. The remaining 45 MW was pinned to Q2 2022 at least five times (14227, 14196, 14030, 14055, 14069). By July, the company was instead describing temporary power restoration at Jamestown by September 12 and permanent restoration by mid-2023 (13856, 13857). That points to an outage that pushed full capacity back by about a year. Claim 13806's undated 100% milestone does not settle whether Q2 was met. The Texas site went from an undated letter of intent to an October 1, 2022 opening (13777) within five months.

The record contains no permits, interconnection queue positions or site reporting, so none of these figures has independent support. Every claim is company-stated.

Over the next 90 days, the first priority is backfilling claims from late 2022 through 2026, which will show whether the 800 MW and 1.8 GW targets were quietly dropped. The next is attaching capacity figures and interconnection records to the Harwood and Boyce sites. The last is checking whether the company re-addresses the 5 GW target (14230) ahead of its March 2027 deadline.

The capacity story in this record contracts quietly. In April 2022 Applied Digital framed its ambition as 800 MW by May 2023 (14031), 1.8 GW by May 2024 (14032) and 5 GW over five years (14033), with a separate 1.5 GW joint venture with Antpool on top (14036). By the August 2022 annual filing the operative figure had become a 500 MW combined target (13728), built from a 100 MW first facility (13721), a 200 MW Texas site (13725) and a 180 MW third North Dakota site (13727). The filing does not acknowledge that this sits well below the 800 MW figure promised four months earlier. The September claim tagging the third facility at 500 MW (13683) looks more like a mislabeled portfolio total than a real upsizing, and should not be read as recovery. The Antpool venture is restated repeatedly (13730, 13782) but never attaches to a site, a megawatt energized, or a date. One extraction recording it as 8,000 USD (13747) needs source review.

On dates, Jamestown's remaining 45 MW was promised for Q2 2022 (14206, 14009). By July the disclosures had shifted to restoring power there, temporary by September 12 (13856) and permanent by mid-2023 (13857). That reads as an unscheduled setback rather than a phased build, and the undated 100% operating capacity claim (13806) does not establish when full output was reached. The Texas site did move from letter of intent (14214) to ground lease and construction start (13740, 13741) with capacity described as fully contracted (13745), consistent with the Marathon hosting deal (13855). Yet the same filing gives October 1 (13726, 13777), Q4 (13742) and year-end (13765) as the operational date, so slippage was already being absorbed into looser language. SparkPool's share forfeiture and exit (13772, 13681) removed one of the founding partners without any stated effect on build plans.

No independent evidence appears in this record. It contains no permits, no ERCOT or MISO/SPP queue positions, and no site reporting, so every pace claim rests on company statements. The record also ends in September 2022 and nothing ties it to Polaris Forge or Delta Forge. Nothing here speaks to the current AI/HPC program.

Over the next 90 days, check Cass and Dickey County filings for the Harwood and Ellendale campuses and Rapides Parish records for Boyce. Check whether interconnection queue entries match announced megawatts and energization dates. Also watch whether new disclosures restate or drop the 2022 gigawatt targets rather than silently superseding them.

The supplied record ends mid-entry at claim 12629 in May 2023, so this read covers Applied Digital's crypto-hosting era only. The Polaris Forge and Delta Forge projects appear on the project list but have no attributed claims, capacity or cost in this extract. That gap is the most important caveat here.

Capacity has been restated several times. In 2022 the company presented roughly 500 MW across three sites (13728, 13648): 100 MW at Jamestown (13776), 200 MW at Garden City (13744), and 180 MW for a future North Dakota site (13746). It also carried a 1.5 GW Antpool joint venture (13766, 13613) that was never tied to a location. By May 2023 the company quoted a North Dakota total of 280 MW (12631), which is Jamestown plus the 180 MW Ellendale site that energized in March 2023 (13183, 13200). It also added a 5 MW GPU build at Jamestown (12576), an early sign of the HPC pivot. The JV was tagged delayed without any new capacity figure (12632). Headline capacity did not shrink on paper. However, the gigawatt-scale figure never turned into a sited project.

Dates have not held. Garden City was due online October 1, 2022 (13777, 13726). The same August 2022 filing also gave a December 31, 2022 date (13765). A September 26 release still said October 1, five days before that date (13614). The date was pushed to year-end on October 10 (13559) and then to Q2 2023 (12573). That makes three schedules in nine months, and the first slip went unacknowledged in contemporaneous materials. The record contains no claim that Garden City ever reached operation. Jamestown also looks inconsistent. A July 2022 disclosure put temporary and permanent power restoration dates out to September 2022 and June 2023 (13856, 13857), yet an August claim reported 100% operating capacity (13806). No later claim confirms that permanent restoration happened.

There is no independent corroboration in the record: no permits, no interconnection queue positions, and no on-the-ground reporting. Even the fully contracted 200 MW at Garden City (13745) rests only on company statements.

Over the next 90 days, the priority is closing the data gap rather than reading market signals. Claims for Polaris Forge 1 through 3 and Delta Forge 1 and 2 should be ingested so the current pipeline can be tested against its own dates. The fiscal first-quarter 10-Q, expected around October, should be checked for energization dates at Ellendale and Harwood. Queue and utility filings for the Harwood and Boyce sites should be pulled. Finally, the file needs a dated resolution for Garden City and for permanent power at Jamestown.

Applied Digital's record here stops in mid-2023. It captures the crypto-hosting era and none of the Polaris Forge or Delta Forge claims implied by the current project list, so this read covers the foundation that later AI-campus promises rest on, not those promises themselves. Within that window, headline capacity grew mainly by adding new sites. The first North Dakota site held steady at 100 MW (13721, 13759), and the Garden City second site held at 200 MW throughout (13725, 13560). A 180 MW third site (13727, 13779) later resolved into Ellendale at the same figure (13200, 13183). Together these produced the 280 MW North Dakota total (12631) and a 500 MW combined target (13728, 13648). Claim 13683, which labels the third facility 500 MW, looks like an extraction conflation with that combined figure rather than a real upsizing. The 1.21 Gigawatts joint venture with Antpool was framed at up to 1.5 GW (13748, 13558) but was flagged delayed by May 2023 (12632), and nothing on record ties it to delivered megawatts. The SparkPool forfeiture and exit (13772) removed one of the original strategic partners. Timelines are where the record is weakest. Garden City was promised for 2022-10-01 as late as 13614 on 2022-09-26. Two weeks later it had moved to year-end 2022 (13559), and by May 2023 it had moved to Q2 2023 (12573). That is two slips inside eight months, and the first came with little lead time. Jamestown's power restoration dates (13856, 13857) have no recorded resolution, and the Ellendale land purchase also carries a delayed tag (12629). The corpus contains no independent evidence at all. Every entry is company-stated, and the project table lists no capacity or cost for any site, including Jamestown, which 13806 described as fully operating. Ellendale and Polaris Forge 1 share an address, which suggests a rebrand rather than two campuses. The dataset should merge them before anyone counts megawatts twice. Over the next 90 days, the priority is backfilling the 2023-2026 Polaris Forge and Delta Forge claims so the AI-era capacity figures can be tested against this slip history. Useful checks include MISO interconnection queue entries for Harwood and Boyce, North Dakota and Louisiana siting or building permits, and whether the next quarterly filing states energized megawatts or only contracted and planned ones. Given the Garden City pattern, pay particular attention to any in-service date that is quietly re-quoted without acknowledging the earlier target.

The record supplied here breaks off at claim 12616, in May 2023. That means the Polaris Forge, Delta Forge and later Ellendale material appears only as project names, with no capacity or cost attached. This read therefore covers the hosting-era trajectory and cannot yet judge the AI-campus build-out.

Through 2022 the capacity story was stable, and in places it was internally muddled. Filings consistently described a 100 MW first facility (13721, 13759), a 200 MW Garden City site (13725, 13778) and a 180 MW third North Dakota site (13727, 13779), summing to a 500 MW target (13728, 13648). One entry attaches the full 500 MW to the third facility alone (13683), which looks like an extraction or drafting error rather than an expansion. The 1.5 GW Antpool joint venture (13748, 13766) sat above all of this as a pipeline figure with no site or date. By May 2023 its formation claim was flagged as delayed (12632). The Ellendale figure also drifts: it is 180 MW at energization (13183, 13200) but 280 MW in 12620. That 280 MW matches the North Dakota total in 12631, so it reads as conflation, not new capacity. The 5 MW GPU build at Jamestown (12576) is small, but it marks the first signal of the pivot away from pure hosting.

Dates are where the record is weakest. Garden City was due to open on October 1, 2022 (13726, 13777, 13614). A same-day filing already carried a December 31 date (13765). That slipped to year-end 2022 (13559) and then to Q2 2023 (12573), with 12627 dropping any date at all. That is two revisions in eight months on a site where construction had started in April 2022 (13724, 13741), and the later filings do not explain the gap. Ellendale is the counterexample: the March 2023 energization (13183) is consistent with the 180 MW site described in 2022.

No independent evidence is attached. There are no ERCOT or MISO interconnection entries, no county permits and no local reporting. The Garden City slippage in particular cannot be attributed to power, equipment or partner issues from this record. The SparkPool exit (13772, 13681) is the only disclosed counterparty disruption.

The first task for the next 90 days is to backfill the post-May-2023 claims so the Polaris Forge and Delta Forge capacities can be traced against these earlier baselines. The fiscal Q1 10-Q, expected around mid-October, should show whether the Harwood and Boyce figures hold or are restated. It should also show whether Garden City's final status was ever reconciled explicitly with the Q2 2023 promise. Interconnection queue positions for the North Dakota and Louisiana sites are the obvious independent check.

The visible record ends mid-2023 at claim 12403, which is truncated. The Polaris Forge and Delta Forge projects on file carry no capacity or timeline claims in this extract. What follows therefore reads the 2022-2023 trajectory, and the later AI campus build should not be judged from it.

On capacity, the site-level numbers were stable while the framing drifted. Ellendale entered at 180 MW (13200, 13183) and resolved into a 280 MW North Dakota total with Jamestown (12575, 12631, 12598), which is arithmetically consistent. Garden City held at 200 MW throughout (13614, 13560, 12574). The 1.5 GW Antpool joint venture was repeated verbatim from 13644 through 12591 without ever acquiring a site, date or dollar figure. That makes it a headline number rather than a pipeline. By June 2023 the language pivoted to a 200 MW HPC target (12395, 12418) and 26,000 GPUs on order (12392). Meanwhile the Jamestown GPU build grew from 5 MW (12576, 12621) to 9 MW (12394), and its completion date moved out by a year.

Dates did not hold. Garden City was promised by October 2022 (13614), then year-end 2022 (13559), then Q2 2023 (12573, 12619, 12545). The Q2 target was still being restated on 2023-06-25 (12417), days before the quarter closed. The successor claims do not acknowledge the earlier misses. Ellendale's March 2023 energization (12630, 12620) was reported as achieved only in late June (12410, 12423), roughly a quarter late but eventually delivered. Several delayed flags carry early-2022 promised-by dates (12623, 12625, 12626, 12632). These look like historical event dates misread as forward promises, so they should not count as slips without review.

No independent corroboration appears anywhere in the record. There are no county permits, no ERCOT queue position for Garden City, no utility filings for Ellendale or Jamestown, and no on-the-ground reporting. Every energization milestone here is company-sourced, and the Garden City slippage is exactly the case where ERCOT interconnection data would settle whether the delay was grid-side or execution-side.

For the next 90 days, the priority is ingesting the post-2023 claims so the Polaris Forge 1, Polaris Forge 2 and Delta Forge 1 records have substance. Polaris Forge 1 and Ellendale also share an address and should be reconciled rather than tracked as separate projects. Beyond the backfill, the fiscal first-quarter results expected around October are the next checkpoint for restated energization dates. Harwood and Boyce permit or interconnection filings would be the first external test of the announced pace, and any quiet disappearance of Garden City's 200 MW from capacity tallies would deserve an explicit note.

The capacity record on file stops at the August 2023 filings, even though the project list shows the later Polaris Forge and Delta Forge sites. That gap is the first thing to note. The project list gives Ellendale and Polaris Forge 1 the same address, so the rebranding happened after the last claim we can see. Nothing in the record ties any of the newer campuses to capacity or dates.

Within the visible window, stated capacity mostly held but became less specific. Ellendale was described as 180 MW (13183, 13200). A later filing called the North Dakota March 2023 energization 280 MW (12620). That figure matches the combined Ellendale-plus-Jamestown total (12575, 12631), so it reads as a conflation rather than an expansion. Jamestown's GPU build grew from 5 MW (12576, 12635) to 9 MW (12394, 12415). The forward pipeline narrowed. In June, 246 MW of crypto hosting was 'coming online' (12396, 12404, 12416). By August, 200 MW was 'committed to be energized' (12335, 12347). The filings never explained the 46 MW difference. A 200 MW HPC expansion was repeated three times (12395, 12418, 12346) without being assigned to a site.

Dates held less well than capacity. Ellendale energization was promised by early March 2023 (12630), but the energized claims appear only in late June (12423, 12410). Garden City is the clearest case. It was guided to Q2 2023 operation (12619, 12573, 12594). On 25 June 2023, with five days left in the quarter, the company still said Q2 (12417). The August filings dropped the date entirely and did not state that it had slipped. Jamestown's end-2023 HPC online date (12394, 12402) was also restated in August without a date (12345). The pattern is dates that disappear rather than dates that get formally revised.

The record contains no independent corroboration: no permits, no interconnection queue positions, and no site reporting. The announced pace therefore rests entirely on company statements. The 1.5 GW Antpool joint venture (12616, 12591) was never tied to a location or a power agreement.

Over the next 90 days, the priority is closing the gap between August 2023 and today. That means ingesting filings that state Garden City's actual energization date and whether Jamestown's 9 MW HPC build came online. It also means capacity and date claims for each Polaris Forge and Delta Forge site. Those claims should be checked against MISO and ERCOT interconnection queues, North Dakota PSC siting dockets, and Louisiana parish permitting for Boyce. A further question is whether the 200 MW HPC plan was ever mapped onto Ellendale.

The record on file is narrow in time and wide in ambition. Every dated claim falls between February and August 2023, yet the project list now includes Polaris Forge 1 through 3 and Delta Forge 1 and 2, with no capacity, cost or timeline claims attached to any of them. Polaris Forge 1 sits at the same Ellendale address as the original 180 MW site (13200), so at least part of the newer portfolio is a rebranding of 2023 assets rather than new ground. The ledger cannot yet say how that site's scope has changed.

Within the 2023 window, capacity figures mostly held steady. North Dakota stayed at 280 MW (12575, 12598), which is Jamestown's 100 MW (12572) plus Ellendale's 180 MW. Garden City was repeated at 200 MW, fully contracted (12574, 12595). The one clear upsizing was the Jamestown GPU build, which grew from 5 MW (12576, 12635) to 9 MW with a year-end 2023 target (12402, 12415). Alongside it sat an undated 200 MW HPC expansion (12418, 12346) and a 1.5 GW joint venture with Antpool (12632, 12591) that never named a site. Those two headline numbers carry no location or date and should be treated as aspirational.

Timing is the weaker part of the record. Ellendale was described as energizing in March 2023 (13183, 12630), but "energized" claims only appear in late June (12410, 12423). That reads as a staged ramp that was never framed as a delay. Garden City is the sharper case. It was still slated for Q2 2023 on June 25 (12417), five days before the quarter ended. By August 1 the Q2 language had disappeared, replaced by an undated 200 MW of "committed additional" capacity (12313, 12347). The June figure of 246 MW coming online (12416) also quietly became 200 MW. Neither change came with an explanation, and nothing on record confirms Garden City was energized.

No independent evidence is attached. There are no permits, no ERCOT or MISO queue entries, and no on-the-ground reporting, so none of the announced pace can be corroborated from this file.

For the next 90 days, the priority is closing the three-year gap. That means backfilling 10-K and 10-Q facility disclosures for Polaris Forge and Delta Forge, and pulling North Dakota PSC and Cass County filings for Harwood. It also means Rapides Parish and Louisiana DEQ records for Boyce, plus interconnection queue positions for each site. Garden City's actual energization date should be reconciled against the Q2 2023 promise. The 9 MW Jamestown deadline should be checked against later filings.

The claim record for Applied Digital stops at August 2023, and that gap matters more than anything inside it. Polaris Forge 1, 2 and 3 and Delta Forge 1 and 2 are on file with no capacity, no cost and no attached claims. Any read of the company's current AI-campus trajectory therefore rests on evidence this record does not contain. What it does show is the crypto-hosting and early HPC phase, and the pattern there is instructive.

North Dakota capacity was stable and modestly accretive. The 280 MW combined figure (12575, 12631, 12549) became 285 MW described as currently energized (12334). The Ellendale energization, promised for March 2023 (12630, 12620), was reported as achieved by June (12423, 12410), so that date broadly held. Garden City is the weak point. It was repeatedly described as 200 MW, fully contracted and operating in Q2 2023 (12573, 12619, 12594, 12545), and that language was still present on 2023-06-25 (12417), days before the deadline. By August the company referred instead to 200 MW "committed to be energized" with no date (12327, 12335, 12347). That reads as a quiet slip rather than an acknowledged revision. The adjacent figure also shrank without explanation, from 246 MW of crypto capacity coming online (12416, 12396, 12404) to 200 MW.

The HPC ambition scaled quickly on paper. A 5 MW GPU facility (12576, 12550, 12635) became a 9 MW Jamestown build targeted for year-end 2023 (12415, 12394, 12402), and within weeks a 200 MW HPC hosting plan followed (12418, 12333). Those steps came from a base of single-digit megawatts and preceded any stated site, schedule or funding for the larger number. Several promised-by dates, such as 12623 and 12626, precede their filing dates. They are historical milestones picked up at extraction, not forward commitments, and should not be scored as held or missed.

No permits, interconnection queue entries or site reporting are attached to any project, so none of the announced pace is independently corroborated here. Over the next 90 days, the priority is backfilling filings after August 2023 so that Garden City's actual energization date and the Polaris Forge megawatt and tenant figures enter the record. Alongside that, look for independent anchors: MISO queue positions and Cass County permits for Harwood, ERCOT records for Garden City, and Rapides Parish filings for Boyce. Also check whether the next 10-Q restates campus capacity in terms consistent with earlier language or resets the baseline without reconciliation.

The capacity record for Applied Digital shows a pattern of upward ambition on high-performance computing (HPC) while the hosting capacity for crypto mining quietly shrinks. North Dakota hosting was 280 MW in May 2023 (12631, 12575, 12549) and 285 MW energized by August (12334), a marginal increase. Garden City was described as 200 MW and fully contracted (12574, 12628). However, the 246 MW said to be coming online in June (12416, 12396) became 200 MW committed to be energized by August 1 (12335, 12347). The filings give no reason for the missing 46 MW. The GPU footprint grew on paper from a 5 MW facility (12576) to 9 MW at Jamestown (12415, 12403), and then to a 200 MW HPC expansion plan (12395, 12418, 12333) with no site, date or power source attached. The 1.5 GW Antpool joint venture (12567, 12616, 12591) has never been tied to a location or megawatt figure in this record.

Dates have not held, and revisions have mostly gone unacknowledged. Ellendale energization was dated March 2023 (12630, 12620) but was reported as energized on June 22 (12423, 12410), roughly a quarter late, and the delay was not explained. Garden City's Q2 2023 target was repeated consistently (12619, 12573, 12594, 12545), including on June 25, five days before the deadline (12417). By August the date was gone, replaced with an undated commitment (12335). The Jamestown 9 MW HPC facility followed the same path. It had a year-end 2023 date in June (12394, 12402), and by August the timeline claim carried no date (12345). The 5 MW GPU construction milestone dated December 2022 (12635) was still described as under construction in May 2023 (12576).

None of the pace is independently supported. The record contains no permits, interconnection queue positions or on-the-ground reporting for any site. The record also stops in August 2023. Polaris Forge 1, 2 and 3, Delta Forge 1 and 2, and Harwood and Boyce carry no capacity, cost or date claims at all. Polaris Forge 1 also shares an address with Ellendale, which suggests a rebranded site rather than new capacity.

Over the next 90 days, the priority is to ingest filings after 2023 so the Polaris Forge and Delta Forge campuses have megawatt and date commitments to test. Separately, check the MISO and SPP queues for Harwood, and the MISO queue for Boyce, Louisiana. Confirm whether Garden City was ever reported energized, and at what capacity. Finally, reconcile the Ellendale and Polaris Forge 1 records so the same megawatts are not counted twice.

Applied Digital's record shows a change in business model more than a steady build-out. In May 2023 the company presented itself as a crypto-hosting operator: 100 MW at Jamestown (12572), 280 MW across North Dakota (12575, 12549), and 200 MW at Garden City described as fully contracted (12574). HPC was a 5 MW footnote (12576). By mid-2023 the language had shifted to plans for 200 MW of HPC hosting (12418, 12333). By April 2024 Ellendale was framed as a 600 MW campus (3236) anchored by a 400 MW hyperscaler letter of intent (3233), with 100 MW under construction (3238). Over the same period the portfolio shrank at the edges. Garden City's 200 MW left through a sale to Marathon (6130), and the North Dakota figures settled at 180 MW Ellendale and 106 MW Jamestown (6122, 6121). The headline potential grew several times over. The contracted base behind it went from a named 200 MW Texas site to an LOI.

The dates have not held well. Garden City's Q2 2023 energization was restated three times (12573, 12545, 12417). No claim on record confirms it happened. The next entries are a February 2024 outage (6259) and the sale, which itself missed its April 2024 target (6130). The Jamestown 9 MW HPC facility carried an end-2023 date in June and August 2023 (12415, 12394, 12325). Parallel August restatements (12345, 12332) dropped the date without comment, and the record has no completion claim. Ellendale's post-outage recovery target of May 2024 slipped (3235, 6275, 6276). Polaris Forge 1's January 2025 operational start is already tagged delayed (6125).

There is no independent corroboration in this record. Every entry is company-stated, and the 2024 financing claim (6109) is not tied to a lender or amount. The supplied history also ends mid-entry at claim 6083, dated May 2024, more than two years before today. Polaris Forge 2 and 3 and Delta Forge 1 and 2 appear as projects with no capacity, cost or date claims attached. Any read of the current trajectory is therefore provisional.

Over the next 90 days, the first priority is to ingest post-May-2024 filings, so that the record can show whether the 400 MW LOI became a definitive lease and whether Polaris Forge 1 energized. Independent checks should follow: North Dakota PSC siting dockets and Cass County permits for Harwood, Rapides Parish and Louisiana DEQ filings for Boyce, and utility or MISO/SPP interconnection entries matching the announced megawatts. Finally, watch whether the Jamestown 9 MW commitment is ever confirmed as delivered or quietly retired.

The claim record for Applied Digital ends in late August 2024, and every project entry still shows status=announced with no capacity or cost. That is two years stale against today, so this read covers the 2023–2024 trajectory only. Many claims are also repeated extractions from the same filings, so how often a figure appears says nothing about its reliability.

The headline pipeline grew fast while delivered capacity stood still. In August 2023 the company framed its HPC ambition as 200 MW (12333, 12326) alongside 200 MW of committed additional crypto capacity (12313, 12327). By April 2024 the pitch had become a 600 MW Ellendale campus potential (3236) anchored by a 400 MW hyperscaler LOI (3233). By August 2024 it was 300 MW of further buildings under design (3248) and 1.4 GW across three more campuses (3249). Operating capacity barely moved, from 285 MW (12334) to 286 MW (3244, 5881). The 200 MW of committed crypto capacity never reappears. Instead the 200 MW Garden City campus was sold to Marathon, and even that closing slipped past its 2024-04-01 target (6130, 6087).

Dates have not held, and the Jamestown slip was not acknowledged. The 9 MW Jamestown HPC facility was promised for H2 2023 (12311, 12325) and was still carried at 9 MW in mid-2024 (6085, 5979). In August 2024 it was described as nearing completion at 7.5 MW (5904). That is roughly eight months late and 17 percent smaller, with no stated revision. Ellendale's post-outage recovery moved from 2024-05-31 (3235) to 2024-06-28 (3245). The 400 MW LOI needed an exclusivity extension past 2024-07-26 (3247). The 100 MW Polaris Forge 1 building was already flagged against its 2025-01-01 operational date (6125, 6086). It was also called under construction in April (3238) but described as breaking ground in August (5905), which is an internal inconsistency worth resolving.

The record contains no independent corroboration. There are no permits, no MISO or SPP interconnection queue entries, and no site reporting. The amended electric service agreement (5868, 6092) is itself a company disclosure.

Over the next 90 days, the first priority is backfilling claims from September 2024 onward. Beyond that, watch three things. First, whether the 400 MW LOI became a definitive lease, which also gates the final $40M Yorkville tranche (6034). Second, whether Polaris Forge 2 in Harwood and Delta Forge 1 in Boyce gain capacity and date claims that can be checked against state siting dockets and queue filings. Third, whether Jamestown's 7.5 MW figure is formally restated.

Applied Digital's announced capacity has remained relatively static on the near term while growing increasingly abstract at scale. The company carried forward a consistent 286 megawatt baseline across Ellendale (180 MW) and Jamestown (106 MW) from May through September 2024 (claims 6122, 6121, 3244, 5881, 5894), though this masks internal composition shifts and some apparent double-counting between the stated 106 MW Jamestown hosting capacity and separately cited 7.5-9 MW GPU facilities. Beyond current production, forward targets escalated from a stated 600 MW Ellendale potential (claim 3236) to a more distributed pipeline: 100 MW in Polaris Forge 1 (claim 3238), followed by 300 MW under design (claim 3248), then three new campuses totaling 1.4 gigawatts (claim 3249). The reshuffling suggests a shift toward portfolio breadth over site concentration.

Timeline performance has been uniformly poor. The January 2025 operational start for Polaris Forge 1's 100 MW HPC facility (claims 6125, 6086) is marked delayed as of late June 2024, with no revised completion date visible in this record. A June 28, 2024 promise to restore Ellendale to full capacity (claim 3245) and a July 26, 2024 hyperscaler LOI exclusivity deadline (claims 3247, 5755) both missed. Most damaging is the Garden City divestiture to Marathon Digital, announced with an April 1, 2024 close date in four separate claims filed May through August 2024 (claims 6130, 6087, 5981, 5882), all remaining tagged delayed without revision. This repetition of the same missed milestone across months suggests either a genuinely stalled transaction or a claims process decoupled from execution tracking.

Independent verification is sparse in this dataset. The February 2024 Ellendale outage is documented, but recovery and capacity figures lack third-party production data or utility commission filing corroboration. No interconnection queue positions, construction permits, or land records appear in the claims to validate announced timelines.

Watch over the next 90 days for a revised Polaris Forge 1 completion date, closure of the Garden City transaction, and whether Jamestown—described as nearing completion in August 2024—delivers actual capacity output. Also monitor whether Polaris Forge 2/3 and Delta Forge campuses advance from announcement to timeline commitment, or remain indefinite future projects.

Applied Digital's announced capacity for its Northern US cluster has compressed while timelines have slipped substantially. In April 2024, the company disclosed a 600 MW long-term campus potential at Ellendale (claim 3236) alongside a 400 MW letter of intent with a hyperscaler partner (claim 3233). Yet by August and September 2024, operational capacity had settled at 286 MW (claims 3244, 5743)—unchanged across three months of quarterly disclosures—with an additional 300 MW noted only as under design, no construction timeline provided (claim 3248). The gap between announced ambition and deployed capacity reflects a material contraction in near-term delivery.

Timeline performance has deteriorated consistently. An Ellendale facility recovery target of May 31, 2024 (claim 3235) drifted to a June 28 deadline (claim 3245), both marked delayed as of late August. The 100 MW HPC Ellendale facility promised operational by January 1, 2025 (claim 6125) is now nearly two years overdue with no revised target disclosed. A February 2024 power outage at Ellendale (claim 6275) underscores operational challenges beyond schedule slip. A July 26, 2024 LOI exclusivity deadline (claim 3247) was similarly delayed. The Garden City campus sale to Marathon Digital Holdings, targeted for April 1, 2024 (claims 6130, 6087, 5981, 5882), remains unresolved nearly 30 months past the promised close date.

Independent verification of construction progress is absent from the claim set. Repetition of the same 286 MW figure across May, June, and August suggests no material capacity increments during that period despite active construction claims.

Watch for revised Ellendale HPC operational dates (the January 2025 target is no longer tenable), any regulatory filings for the 300 MW design-phase projects, and final disposition of the Garden City transaction. Silent milestone rescaling in future disclosures, or omission of prior commitments without explanation, would signal weakening execution transparency.

Applied Digital has held its core operating capacity at 286 MW across North Dakota sites since May 2024, combining 180 MW at Ellendale and 106 MW at Jamestown (claims 6122, 6121), with an additional 9 MW GPU facility at Jamestown (6123). This figure remained consistent through August and September. However, the company's announced expansion pipeline shows material compression and timeline strain. The marquee Polaris Forge 1 project—a 100 MW HPC facility promised operational by January 1, 2025 (claim 6125)—remains under construction as of late September, with no revised completion date disclosed. Similarly, the promised sale of Garden City to Marathon Digital, initially due April 1, 2024 (claim 6130), was still delayed as of August 29 (claim 5882), representing a five-month slip without public timeline revision. The 400 MW hyperscaler lease, central to Ellendale's expansion economics, remains in exclusivity negotiations. An LOI exclusivity extension promised by July 26, 2024 (claim 3247) was marked delayed in August reporting, and as of September the lease remained unsigned despite being repeatedly cited as material to the Polaris Forge 1 $40 million financing tranche (claim 6034). A discrepancy appears in the Jamestown buildout: while the site is marketed with 106 MW total capacity, the HPC facility specifically is described as nearing completion with only 7.5 MW (claims 5904, 5859), suggesting the stated capacity includes legacy systems not suited to current hyperscaler demand. In August, Applied Digital announced a 1.4 GW pipeline across three campuses (claim 3249), with 300 MW under design (claim 3248), but these remain conceptual. No permits, interconnection filings, or third-party validation for these projects appear in available records. The operational narrative is one of a company that stabilized its inherited base but has not yet executed on high-stakes capacity expansions or closed transformative deals. The Polaris Forge 1 facility will be the key tell: any further delay beyond January 2025, or downward capacity revision, would signal execution headwinds. Monitor whether the Marathon transaction closes, whether the hyperscaler lease signs before year-end, and whether any Polaris 2 or Delta Forge projects show tangible permitting or grid-queue activity.

Applied Digital's trajectory reveals a widening gap between announced ambitions and realized capacity. The company's core operating portfolio—Ellendale at 180 MW and Jamestown at 106 MW, totaling 286 MW (claims 6084, 6083, 5881)—has remained stable since Q2 2024, but this baseline masks significant execution delays in growth projects. Polaris Forge 1, the 100 MW HPC facility at Ellendale, was promised operational by January 2025 (claims 6125, 6086). As of September 2026, it remains under construction, representing a 21-month slip with no revised completion date provided. This is not a minor delay: the facility was described variously as under construction, nearing commencement, and awaiting hyperscaler commitment (claims 5774, 5873). Similarly, the 200 MW Garden City sale to Marathon Digital was promised by April 2024 (claim 6130), then repeatedly re-announced with no closure through August 2024 (claims 6087, 5981, 5882), suggesting either negotiation friction or deal structure changes unacknowledged in public statements.

Capacity announcements have scaled without delivery evidence. In August 2024, APLD claimed 300 MW under design for the next two Polaris Forge buildings and 1.4 GW across three additional campuses (claims 3248, 3249). No permits, grid interconnection queue filings, or ground-truth reporting supports active construction. The company's partnership stack—Dell, HPE, Super Micro agreements (claims 5865, 5876, 5877)—appears largely unmoored from actual facility ramp. Earlier disruptions, including an Ellendale outage in February 2024 with a recovery target of June 28 (claim 3245), suggest infrastructure fragility not addressed in growth narratives.

The hyperscaler LOI exclusivity deadline of July 26, 2024 (claims 3247, 5755) passed without public resolution, raising questions about the 400 MW lease commitment (claim 3233). Watch for: (1) actual operational start of Polaris Forge 1 with third-party power draw or interconnection queue advancement; (2) completion of the Marathon Digital sale; (3) any revised timelines or capacity restatements in near-term guidance; (4) actual revenue or utilization metrics from claimed partnerships. Silent missed dates and repeated announcements of unchanged capacity suggest internal constraints not reflected in public messaging.

Applied Digital's stated capacity has remained stable at 286 MW since May 2024 (claims 6121, 6122, 6083-6085, 5976-5978), holding across Jamestown (106 MW hosting plus 9 MW GPU facility) and Ellendale (180 MW). The company has not reduced announced current capacity but has expanded its development pipeline significantly—adding 300 MW in two planned buildings at Ellendale (claim 3248) and 1.4 GW across three additional campuses (claim 3249), announced in August 2024. This suggests internal confidence but creates a distinction between stated present and promised future that warrants scrutiny.

Timeline performance has been inconsistent. The Garden City facility sale to Marathon Digital, promised to close by April 1, 2024 (claims 6130, 6087, 5882), remains unreported as complete nearly eighteen months later—a material miss without revision. Ellendale's own recovery milestones slipped: the facility was to restore to full capacity by June 28, 2024 (claim 3245), which passed without acknowledgment. More critically, the 100 MW HPC facility at Ellendale was promised operational by January 1, 2025 (claims 6125, 6086), a date now imminent and already tagged as delayed before arrival. An interim milestone—LOI exclusivity extension promised by July 26, 2024 (claim 3247)—was also missed.

The Jamestown HPC facility presents a specific credibility test. Claims from August 2024 describe it as either "nearing completion" (claims 5904, 5859) or "under development" (claim 5888), suggesting internal uncertainty about status three months prior. No operational announcement has followed into September.

Independent verification of pace remains absent from the claim record. No permits, interconnection queue placements, or ground-truth reporting appears here to corroborate the construction narrative. The company anchors credibility to supply partnerships (Dell, HPE, Super Micro) and financing arrangements, but these are preparatory rather than delivery indicators.

Watch for: operational status confirmation of Jamestown HPC within thirty days; clarity on whether Ellendale's 100 MW comes online by end-2025 or receives a revised date; completion of Garden City divestiture; and any announcement of interconnection queue positions or permits for the announced 300 MW and 1.4 GW phases. Each represents a shift from planning to execution.

Applied Digital disclosed 286 MW of operating hosting capacity by mid-2024, comprising 180 MW in Ellendale and 106 MW in Jamestown (claims 5977, 5976, 5978). Yet the company's capacity narrative expanded significantly within months. By August, it announced a 400 MW letter of intent with an unidentified hyperscaler (claim 3246), committed to designing two additional 300 MW buildings (claim 3248), and outlined three additional campuses totaling 1.4 GW under development (claim 3249). The roadmap now implies a five-fold increase from stated current capacity to promised future supply within the announced portfolio.

Execution against stated timelines has lagged consistently. The Garden City facility sale to Marathon Digital, promised for April 1, 2024, was flagged as delayed by May (claim 6130) and remained unresolved into late August (claim 5882)—a persistent four-month slip. Ellendale's HPC facility (Polaris Forge 1), initially targeting operational status by January 1, 2025, was already marked as delayed by May 2024 (claim 6125). The facility's full restoration promise of June 28 (claim 3245) also missed. An extended LOI exclusivity period, promised by July 26 (claim 3247), slipped as well. Early-year operations faced unplanned events too: an Ellendale power outage in February (claim 6275) and a three-day Garden City outage (claim 6259).

The record contains no independent corroboration. No permits, interconnection queue filings, or third-party construction evidence appears. Jamestown's 7.5 MW HPC facility was described as nearing completion in August (claims 5859, 5904), yet no subsequent operational announcement has surfaced.

Over the next 90 days, several milestones will indicate whether the company's expanded roadmap reflects real execution momentum. The Marathon Digital transaction, now sixteen months overdue, must close to restore deal credibility. Jamestown's HPC facility should transition from nearing completion to operational declarations with actual workload data. Polaris Forge 1's 300 MW follow-on buildings should show design progression—permitting, equipment orders, or financing announcements. The stated partnerships with HPE, Dell, and Super Micro (claims 5884, 5865, 5883) should demonstrate actual capacity uptake if relationships remain active. Until Applied Digital resolves Ellendale's operational stability and closes Garden City, the 1.4 GW expansion narrative remains aspirational rather than grounded in demonstrated execution capability.

Applied Digital's disclosed capacity expansion has remained remarkably static despite eighteen months of progress claims. The company reported 286 MW of operating hosting capacity across North Dakota as of May 2024, with that figure repeated unchanged through September 2024. The stated pipeline evolved from a 600 MW campus potential announced in April 2024 to a broader portfolio claim of 1.4 GW across three additional campuses by August 2024, suggesting a shift toward greenfield development rather than near-term density gains on existing footprints.

The Ellendale HPC facility exhibits the most concerning pattern. A January 1, 2025 operational start date was established in May 2024 for the 100 MW installation. By August 2024, that facility was still described as under construction, and the operational timeline remained unstated. A separate recovery deadline for existing Ellendale capacity—June 28, 2024—also slipped without revision. The 400 MW hyperscaler lease letter of intent, cited as a funding condition and announced in multiple iterations between April and September 2024, shows no progression markers or exclusivity term updates despite repeated references to extended negotiation periods.

The Garden City divestiture to Marathon Digital presents the starkest execution gap. A April 1, 2024 closure date was announced and subsequently missed; by August 2024 the transaction remained unclosed and unrevised, suggesting either regulatory, financing, or deal-structure friction not disclosed in public filings. No public reporting, permit records, or interconnection queue documentation has emerged to validate the construction pace of the announced pipeline facilities.

Jamestown's smaller HPC build was characterized as nearing completion in August 2024 with 7.5 MW capacity, but no operational declaration or revenue metrics followed. The divergence between static reported capacity and dynamic pipeline announcements suggests APLD may be optimizing external messaging around promised headroom rather than actual delivery velocity.

Over the next ninety days, watch for Ellendale operational status announcements, any revised Polaris Forge timelines with specific capacity commitments, and the eventual resolution or withdrawal of the Marathon Digital transaction. If no Ellendale operational declaration appears by December 2024 and the January 2025 deadline passes unremarked, the pattern shifts from delayed to systematically deferred.

Applied Digital's capacity narrative contracted significantly between spring and fall 2024. In April, the company articulated 600 MW future potential at Ellendale (claim 3236) underpinned by a 400 MW hyperscaler LOI (claim 3233). By May, stated operating capacity was 286 MW—180 MW Ellendale, 106 MW Jamestown (claims 6121–6122). This figure remained static through September. The April announcements were later paired with 1.4 GW listed as "under development" (claim 3249), representing a shift from near-term delivery to extended pipelines.

Timeline adherence has been persistently weak. Ellendale capacity recovery was promised by May 31, 2024 (claim 3235) and full restoration by June 28 (claim 3245); both dates passed without documented revision. The 100 MW HPC facility, targeted for January 1, 2025 operationalization (claims 6125–6086), remained under construction through August 2024 (claim 5889), indicating the deadline will miss by months at minimum. The 400 MW LOI evolved from partnership statement in April to conditional exclusivity requiring extension past July 26, 2024 (claim 3247), suggesting no binding agreement emerged. Garden City's sale to Marathon Digital, promised for April 2024 close (claims 6130, 5882), remained unsigned in September—five months past deadline.

No independent permitting, interconnection queue data, or on-site reporting validate these timelines. The dataset comprises only company statements. Capacity announcements have systematically exceeded delivery, major partnerships requiring explicit milestones have slipped without acknowledged revision, and a landmark acquisition remains unsigned after contractual binding dates elapsed.

Through year-end 2026, monitor whether HPC Ellendale enters service, whether cumulative capacity crosses 400 MW, and whether the 400 MW LOI achieves binding status. The consistent pattern of optimistic announcement followed by unacknowledged delay suggests management's internal tracking diverges significantly from public claims.

Applied Digital's announced capacity roadmap has undergone substantial revision since early 2024, though the company has not explicitly recalibrated expectations publicly. In April 2024, management outlined a 600 MW future capacity potential for the Ellendale campus (claim 3236), with near-term commitments of 180 MW there and 106 MW at Jamestown solidifying into a stated 286 MW of operational hosting (claim 3244 by August). Alongside this, claims for a 100 MW HPC facility under construction at Ellendale via Polaris Forge 1 (claim 3238) and a 400 MW hyperscaler letter of intent (claim 3233) suggested a path to rapid scaling. By August 2024, the company had disclosed an additional 1.4 GW development pipeline across three unnamed campuses (claim 3249), a trajectory that implies the earlier 286 MW represents just the foundation.

However, timeline execution has diverged materially from stated commitments. Ellendale's power outage recovery, initially targeted for May 31, 2024 (claim 3235), slipped to June 28, 2024 (claim 3245) without public acknowledgement of the revision. The Garden City sale to Marathon Digital—promised for April 1, 2024 (claim 6130)—remains unresolved across multiple filings through August 2024. The Polaris Forge 1 facility was promised operational by January 1, 2025 (claim 6125), yet September 2024 announcements described it as merely under construction (claim 5774) with no revised timeline provided. The 400 MW hyperscaler letter of intent has shown similar slippage: an August announcement noted extended LOI exclusivity promised by July 26, 2024 (claim 3247), suggesting the original window had closed without deal closure. The company issued this revision after the promised date rather than before, a pattern repeated across Ellendale capacity milestones.

Independent verification of construction pace, interconnection queue status, or real-estate agreements has not been disclosed. Over 18 months, Applied Digital has maintained operational capacity at 286 MW while anchoring investor narrative to a 1.4 GW development pipeline—a 5x expansion—without disclosed permitting, interconnection entitlements, or major customer commitments beyond the stalled hyperscaler LOI. Watch for any announcement of Polaris Forge 1 operational status, confirmation or closure of the Garden City sale, and updates on the 1.4 GW development sites' real-estate or power agreements in the next 90 days.

Applied Digital's capacity roadmap has expanded markedly without resolution of prior timeline commitments. Operating capacity has remained flat at 286 megawatts across Ellendale and Jamestown since May 2024 (claims 5976–5977, 6083–6084), while the company announced up to 1.4 gigawatts under development across three additional campuses and 300 megawatts in next-phase construction under design (claims 3248–3249). The expansion narrative is substantial but contingent on execution against a visible track record of delay.

Ellendale's February 2024 power outage (claim 6275) triggered cascading missed recovery dates. The facility was promised restoration by May 31, 2024 (claim 3235), then revised to June 28 (claim 3245). Both deadlines have passed without stated resolution as of September 10. The 100-megawatt HPC facility at Polaris Forge 1, promised operational by January 1, 2025 (claims 6125, 6086), remains fourteen months ahead with no interim milestones disclosed. Jamestown's 7.5-megawatt HPC was described as nearing completion in late August (claims 5859, 5904) without subsequent progress updates.

The most glaring unresolved item is the Marathon Digital Holdings acquisition of the Garden City campus, originally promised by April 1, 2024 (claims 6130, 6087, 5981, 5882). This transaction sits five months overdue without public explanation across consecutive quarterly filings. Similarly, an LOI exclusivity extension for a 400-megawatt hyperscaler lease agreement missed its July 26, 2024 deadline (claim 3247) without revision or update.

No independent corroboration—permits, grid interconnection queue positions, local inspection records—appears in the record. Capacity and timeline claims rely solely on internal announcements and vendor partnership statements. Over the next ninety days, watch for January 1 operational confirmation at Polaris Forge 1, any disclosed Jamestown completion, closure of the overdue Marathon transaction, and formal execution or revised terms on the hyperscaler lease.

Applied Digital's published facility roadmap shows persistent distance between announced timelines and reported completion, with capacity claims that have held steady in near term but hinged on partnerships not yet realized at scale. The company's core North American footprint stabilized at 286 MW of operational capacity by mid-2024 (claims 3244, 5978): 180 MW at Ellendale, 106 MW at Jamestown, plus 9 MW of cloud services capacity (claims 3238, 6121, 6123). This baseline has remained consistent through September 2024 announcements. However, progression beyond this threshold has faced repeated delays. The HPC facility at Ellendale promised operational start by January 1, 2025 (claims 6125, 6086) remains listed as under construction as of late August 2024. Similarly, the Jamestown HPC facility announced as nearing completion with 7.5 MW capacity (claim 5904) lacks reported commissioning. The Garden City sale to Marathon Digital, initially promised by April 1, 2024 (claims 6087, 5882), has not been resolved as of September 2024 reporting despite multiple restatements of the 200 MW transaction. Ellendale's operational restoration, first promised by May 31, 2024 (claim 3235) then June 28, 2024 (claim 3245), was acknowledged as delayed in the company's August 28 filing without revision date. The company has maintained consistent messaging around partnership LOIs without clear evidence of signed capacity deployment. The 400 MW hyperscaler lease letter of intent, first announced in April 2024 (claim 3233), appeared in September 2024 documentation (claims 5752, 5775) as still pending, with exclusivity periods slipping (claim 3247 promised by July 26, 2024). Forward capacity claims introduce material scale: three additional campuses totaling 1.4 GW and a next-phase build of 300 MW (claims 3249, 3248), remaining categorized as under development without dated milestones. Partnership announcements with Dell, HPE, and Super Micro (claims 5770, 5771, 5863) recur across multiple quarters without disclosed deployment metrics. Monitor whether Polaris Forge 1 reaches operational status, whether the hyperscaler lease converts from LOI to executed agreement with capacity takedown schedule, and whether new timeline claims acknowledge prior delays or introduce revised dates. Independent verification via utility interconnection queue entries or regulatory filings remains absent from public record.

Applied Digital's stated capacity has remained flat through a year of operational claims. The company reported 286 megawatts of hosting capacity across Ellendale and Jamestown (claims 3244, 5881, 5894) as of August 2024, with no net growth announced since May 2024, despite repeated quarterly disclosures covering the period. The two current operational sites—Ellendale at 180 MW (claims 5858, 5896) and Jamestown at 106 MW (claims 5857, 5895)—showed no expansion through the data window ending September 2024.

Promised timelines have slipped significantly. The Ellendale HPC facility (Polaris Forge 1), initially promised operational by January 1, 2025 (claims 6125, 6086), now appears substantially behind as of this assessment. The Garden City divestment to Marathon Digital, promised for April 1, 2024 (claims 6130, 6087, 5981, 5882), was marked delayed repeatedly through late August 2024. The company tied $40 million in final funding to execution of a 400 MW hyperscaler lease agreement (claim 6034), yet that lease remained characterized only as a letter of intent through August 2024 despite multiple references suggesting imminent finalization.

Independent corroboration is sparse in the claim record. Permit filings or interconnection queue entries do not appear in the dataset. The facility outages documented (claims 6275, 6276 at Ellendale in February 2024; claim 6259 at Garden City) reflect reactive rather than proactive capacity operations. Supply partnerships with Super Micro, HPE, and Dell appear frequently, but these are procurement agreements rather than evidence of capacity delivery.

The gap between stated and achieved capacity is material: 1.4 gigawatts under development across three unnamed campuses and 300 megawatts under design at Polaris Forge (claims 3248, 3249) remain preliminary as of late summer 2024. No site address appears for Polaris Forge 2, 3, or the Delta Forge properties, and no capacity figures are tied to them.

Over the next 90 days, watch for formal commissioning of Polaris Forge 1 and any public revision of its timeline. Track whether the 400 MW lease converts from LOI to binding agreement and whether the three-campus pipeline moves beyond planning stages. Monitor whether the company revises capacity guidance or provides updated delivery schedules for future phases.

Applied Digital has positioned operating capacity at 286 megawatts across existing North Dakota facilities (Ellendale 180 MW, Jamestown 106 MW, and 9 MW GPU hosting) while announcing future development of approximately 1.4 gigawatts across three additional campuses. The company has restated the 286 MW figure consistently from May through September 2024 (claims 5976, 5977, 5978, 5881, 5894), indicating a stable baseload rather than growth in current operations. Future capacity announcements cluster around the Polaris Forge 1 project (100 MW HPC facility, claim 5871) and additional phases under design (claim 3248: 300 MW), suggesting management is positioning expansion capacity without near-term delivery.

Timeline execution shows material slippage. Ellendale facility recovery from a power outage, initially promised by May 31, 2024 (claim 3235), remained incomplete when reported in late August (claim 3245, marked delayed). The HPC Ellendale operational start is positioned for January 1, 2025 (claims 6125, 6086), now five months out with prior delays on record. More significant is the Garden City campus sale to Marathon Digital Holdings, promised by April 1, 2024 (claims 6130, 6087, 5981, 5882), which remains outstanding as of September. This 200-megawatt asset divestiture has slipped five months past its promised close date without public revision.

The 400 MW hyperscaler capacity lease is structured as a Letter of Intent with extended exclusivity (claims 5752, 5775, 5776). The exclusivity period extension, promised by July 26, 2024 (claim 3247, marked delayed), suggests negotiations remain unsettled. This lease is contingent on Ellendale HPC facility operationalization, creating a dependency chain where downstream commitments hinge on Ellendale delivery.

Within this dataset, independent verification markers are limited; however, the Garden City transaction delay is observable and material. A signed transaction now extends five months beyond close, raising questions about regulatory, technical, or counterparty readiness. The pattern of missed internal target dates across recovery, construction, and transactional fronts suggests either conservative initial guidance or execution headwinds in permitting and engineering.

Over the next 90 days, track Garden City transaction completion (overdue since April), any operational milestones or updated timeline for Ellendale HPC facility (January 1 target at risk), and formalization status of the 400 MW hyperscaler lease. Early Q4 guidance updates would indicate whether management is reconserving timelines.

Applied Digital's publicized facility roadmap reveals a pronounced pattern of timeline slippage coupled with remarkably static current capacity. The company reported 286 MW of operating hosting capacity across Ellendale and Jamestown as of May 2024 (claims 5978, 6084, 6083), a figure that remains unchanged through September 2024 (claims 5743, 5762). This consistency suggests either stable deployment or slower ramp than initially modeled.

Promised delivery dates have deteriorated substantially. The Garden City Campus sale to Marathon Digital Holdings was promised by April 1, 2024 (claims 6130, 6087), yet remains unresolved as of September 2026, nearly two and a half years past deadline. Ellendale's recovery to full operating capacity from an earlier outage was promised by June 28, 2024 (claim 3245), now almost two years overdue. Most significantly, the critical HPC Ellendale facility—positioned as a centerpiece 100 MW building—was promised operational by January 1, 2025 (claim 6125), a deadline that has slipped by approximately twenty months.

The 400 MW letter of intent with a hyperscaler (claim 3246) was accompanied by an extended exclusivity period promised by July 26, 2024 (claim 3247), creating ambiguity about deal status. No public evidence has materialized concerning equipment delivery, infrastructure completion, or customer ramp at these sites. The company's August 2024 filings (claims 3248, 3249) introduced two new 300 MW buildings under design and three additional campuses totaling 1.4 GW, significantly expanding announced capacity without publicly providing constructive timelines or financing specifics beyond initial concept.

Independent corroboration is limited: no recent permit updates, interconnection queue entries, or third-party site reporting have surfaced to validate the stated construction velocity. The eighteen- to thirty-month gaps between promises and delivery dates—evident across the Garden City transaction, Ellendale recovery, and HPC commissioning—suggest management timelines are substantially optimistic relative to execution.

Over the next ninety days, watch for official announcement of Garden City closing or termination, public disclosure of actual HPC Ellendale power-up or operational customer deployments, financing closure for the 300 MW expansion, and hyperscaler LOI conversion to definitive agreement. Until those events materialize, capacity growth remains more aspiration than realized.

Applied Digital's capacity and timeline narrative has expanded significantly while execution has consistently lagged. In April 2024, APLD announced 100 MW of HPC construction in Ellendale (claim 3238) alongside a 400 MW hyperscaler lease LOI (claim 3233), painting a 600 MW future potential for the campus (claim 3236). By August, this had scaled to 300 MW in the next two Polaris Forge buildings (claim 3248) and 1.4 GW across three additional campuses (claim 3249), all while maintaining the initial 286 MW of stated operating capacity (claim 3244). The company has restated this 286 MW figure repeatedly through June and August without material new operational capacity announced, suggesting these numbers represent the same assets described in earlier claims (claims 6121, 6083, 6122, 5976, 5977).

Promised delivery dates have experienced consistent erosion. The Ellendale HPC facility operational start was promised for January 1, 2025 (claims 6125, 6086), yet appears headed for delay with no revised date published. Ellendale's recovery to full operating capacity was promised by June 28, 2024 (claim 3245), but the company provided no update on whether this milestone was met. The Garden City sale to Marathon Digital, promised by April 1, 2024 (claims 6130, 6087, 5981, 5882, 5745), remains incomplete as of September, representing over five months of delay without public acknowledgment. An LOI exclusivity period was promised to extend through July 26, 2024 (claim 3247), also marked delayed.

Jamestown presents a credibility concern. Described as 'nearing completion' with 7.5 MW HPC capacity since late August 2024 (claim 5859), no operational announcement has followed, suggesting continued delays. An earlier outage at the Ellendale facility (claims 6275–6276) and a three-day outage at Garden City (claim 6259) indicate operational stress early in the year.

No independent verification appears in the record—no public permits, interconnection queue entries, or third-party on-the-ground reporting. All claims originate from APLD or its stated partnerships. Watch for actual January 2025 Polaris Forge 1 commissioning, completion of the Garden City transaction, and any published permits or utility interconnection notices that would evidence the 1.4 GW pipeline.

Applied Digital has repeatedly restructured its capacity narrative, beginning with ambitious projections before moderating expectations. In April 2024, the company claimed Ellendale's "future capacity potential" of 600 MW (claim 3236) alongside a 400 MW hyperscaler lease LOI (claim 3233). By May 2024, these figures collapsed into a narrower 286 MW operating base: 180 MW at Ellendale and 106 MW at Jamestown (claims 6122, 6121), with an additional 9 MW GPU facility at Jamestown (claim 6123). This capacity held steady through August 2024 (claim 3244). However, August announcements pivoted to future prospects: 300 MW in two buildings "under design" (claim 3248) and 1.4 GW across three unnamed campuses (claim 3249), moving claimed expansion into indefinite timelines.

Timeline discipline has fractured across multiple facilities. The Ellendale outage that began February 2024 (claim 6275) triggered promises of recovery by May 31, 2024 (claim 3235), later extended to June 28, 2024 (claim 3245). Both deadlines have slipped by over two years with no acknowledgement. The Garden City campus sale to Marathon Digital, promised April 1, 2024 (claims 6130, 6087, 5981, 5882), remains unrealized after 28 months. Most acutely, the Polaris Forge 1 HPC facility at Ellendale—a 100 MW marquee project—was promised operational by January 1, 2025 (claims 6125, 6086). It is now 20 months past its stated delivery date with no revised milestone disclosed. This pattern suggests management has shifted from committal to aspirational language; claims from August onward carry no delivery dates, instead describing facilities as "under construction" or "nearing completion" without quantified milestones (claims 5889, 5904).

Independent verification remains absent. No permit filings, interconnection queue entries, or third-party construction documentation substantiate the announced pace. Partnership announcements with HPE (claim 5770), Dell (claims 5865, 5902), and Super Micro (claims 5863, 5877) lack execution detail. Over the next 90 days, monitor: any revised timeline for Polaris Forge 1's completion; resolution of the 28-month-overdue Garden City transaction; and whether the 1.4 GW pipeline generates site-level specificity or remains strategic positioning.

Applied Digital's facility delivery trajectory shows widening gaps between announced capacity and on-time execution. The company promised Polaris Forge 1 operational status by January 2025 (claims 6125, 6086); as of September 2026, that deadline has passed unmet by over eighteen months. An Ellendale restoration target of June 28, 2024 (claim 3245) also slipped significantly, and an LOI exclusivity extension promised by July 26, 2024 (claims 3247, 5755) was delayed by August 2024.

Announced capacity has scaled aggressively without proportional evidence of delivery. Ellendale's capacity roadmap evolved from 100 MW under construction (claim 3238) to a 600 MW campus potential (claim 3236) to a stated 1.4 GW development across three additional campuses (claim 3249, announced August 2024). The company has maintained a stable 286 MW figure for total operating hosting capacity since June 2024 (claims 5978, 5881, 3244), raising questions about whether near-term builds are progressing or the baseline is optimistic.

The Garden City transaction to Marathon Digital illustrates non-transparent slippage. Originally promised by April 2024 (claim 6130), the deal was re-announced with the same April 2024 deadline in May (6087), June (5981), and August (5882)—each time marked delayed without revised timelines disclosed. Whether this represents deal failure, regulatory hold-up, or chronic non-disclosure remains unclear.

Hyperscaler commitments appear conditional on delivery. The 400 MW LOI (claims 3246, 5775) and the ELN-02 lease tied to a $40M tranche (claim 6034) suggest financing and capacity execution are tightly coupled; slippage in one likely affects the other. Without independent confirmation—interconnection queue status, recent construction permits, utility agreement updates—the gap between promised timelines and marked delays suggests execution risk has outpaced public disclosures. Watch for: whether Polaris Forge 1 achieves operations within ninety days; any updated Garden City close timeline; confirmation of the 400 MW lease signing; and operating capacity figures for the three campuses announced twenty-four months ago.

Applied Digital's performance against announced capacity and timeline targets reveals meaningful gaps between promise and delivery. The company's operating footprint stabilized at 286 megawatts distributed across Ellendale (180 MW) and Jamestown (106 MW) from May 2024 onward, with no subsequent expansion of existing sites announced [claims 5978, 3244]. Yet announced pipeline capacity accelerated: three additional campuses totaling 1.4 gigawatts entered the development category in August 2024, with immediate focus on two Polaris Forge buildings specified for 300 megawatts [claims 3249, 3248].

Timeline commitments have slipped consistently across facilities. An Ellendale power outage beginning February 2024 set recovery targets initially at May 31, revised to June 28, but remained incomplete when disclosed as delayed in late August [claims 3235, 3245]. The 100-megawatt HPC facility designated Polaris Forge 1 was promised operational by January 2025 but appeared throughout the 2024 claim record under unchanged construction status, despite May claims already flagged as delayed against that target [claims 6125, 6086, 5889]. A proposed 400-megawatt hyperscaler lease positioning remained at letter-of-intent through August 2024, with stated exclusivity terms already passed [claims 3246, 3247].

Garden City, Texas presents the starkest timeline divergence: APLD promised the divestiture to Marathon Digital by April 1, 2024, yet the transaction appeared marked delayed in five separate disclosures spanning May through August [claims 6087, 5981, 5882]. No announced revised closing date appears in the available record.

Independent verification of construction velocity remains constrained by available data. The Jamestown HPC facility (7.5 MW) is the sole project consistently described with forward progress language, approaching completion status by August [claims 5904, 5859]. Critical measures from that period forward included whether Polaris Forge 1 achieves the January 2025 operational target without further revision, whether Garden City closure materializes with disclosed timing, and whether the 1.4 gigawatt pipeline produces external evidence—permitting records, interconnection queue positions, signed power or supply agreements—indicating genuine site development momentum beyond announcement.

Applied Digital's public capacity claims have remained static for over two years despite aggressive construction announcements. The company has claimed 286 MW of operational hosting capacity since May 2024 (claims 6122, 6121, 6123, totaling 180 MW at Ellendale and 106 MW at Jamestown), repeating this figure consistently through September 2024 and beyond. No operational capacity additions have materialized since that announcement, suggesting execution shortfalls across multiple projects. The Polaris Forge 1 facility in Ellendale, promised as a 100 MW HPC build with operational start by January 2025 (claims 6125, 6086), has not come online as of current date, representing a miss of over 20 months. Similarly, the Ellendale main facility's capacity recovery was initially promised for May 31, 2024 (claim 3235) following a power outage in February 2024, then revised to June 28, 2024 (claim 3245), yet the facility appears to have continued operating below nameplate. The Garden City campus sale to Marathon Digital Holdings, announced at 200 MW and promised by April 1, 2024 (claims 6130, 6087, 5982), has not closed and remains unresolved nearly two and a half years later. Conversely, future-stage pipeline claims have grown substantially, with the company announcing 300 MW under design and 1.4 GW across three additional campuses as of August 2024 (claims 3248, 3249), though these remain at very early stages with no construction permits or interconnection queue entries reported. The extended LOI exclusivity for a 400 MW hyperscaler lease was promised by July 26, 2024 (claim 3247) with no public update on that deadline's outcome. The pattern reflects a widening gap between announced capacity expansion roadmaps and actual delivery, with all meaningful near-term builds now significantly delayed. Watch for any operational announcement of Polaris Forge 1 in the next quarter, closure of the Marathon Digital transaction, substantive updates on the 300 MW phase-two build, and whether any of the three new campus sites begin permitting processes.

Applied Digital's capacity announcements expanded substantially through mid-2024. An initial 100 MW HPC facility at Ellendale (claim 3238, April 2024) was reclassified to 180 MW total Ellendale capacity by May (claim 6122), alongside 106 MW at Jamestown (claim 6121). By August, the company claimed 286 MW across operating assets (claim 3244), then announced 300 MW under design (claim 3248), a 400 MW hyperscaler LOI (claim 3246), and 1.4 GW across three additional campuses (claim 3249). This trajectory shows consistent upward revision rather than capacity contraction, though the distinction between announced, under-construction, and operational capacity became increasingly muddled in successive statements.

Timeline adherence deteriorated significantly. Ellendale experienced an outage in February 2024 (claim 6275) with restoration promised within weeks. Recovery targets slipped from May 31 (claim 3235) to June 28 (claim 3245), both marked delayed in the record. More critically, Polaris Forge 1's HPC facility promised operational by January 1, 2025 (claims 6125, 6086), remained under construction through August 2024 (claim 5889). The Garden City asset sale to Marathon Digital, originally promised for April 1, 2024 (claim 6130), appeared unresolved as of August, with repeated re-announcements suggesting execution delay. A material financing condition—the 400 MW lease with an unnamed hyperscaler (claim 6034)—remained at LOI stage through September 2024 (claim 5752).

The dataset contains no independent verification layer: no permit filings, interconnection queue progression, or third-party site surveys. Claims distinguish vaguely between operational, under-construction, and announced capacity without precision. The 286 MW "operating" figure (claims 5894, 5881) appears to represent total Jamestown (106 MW) plus Ellendale (180 MW) portfolio capacity, yet the 100 MW Polaris Forge 1 HPC build at Ellendale remained incomplete, creating ambiguity about what proportion was genuinely energized.

Over the next 90 days, watch for: (1) clarification of Polaris Forge 1 operational status given the missed January 2025 target; (2) formal revision or confirmation of all infrastructure timelines; (3) execution of the hyperscaler lease or LOI expiration; (4) evidence of physical progress on the announced 300+ MW of unbuilt capacity; and (5) completion or dissolution of the Garden City transaction. Without independent corroboration, the delivered-versus-announced gap remains material.

Applied Digital has maintained stable claims of 286 megawatts in operating North Dakota capacity since May 2024, comprising 106 MW in Jamestown and 180 MW in Ellendale across six distinct announcements (claims 6121, 6122, 5894, 5762). The core capacity figures have not shrunk, but the growth trajectory has flatlined. An initial 100 MW HPC construction claim for Ellendale (claim 3238, April 2024) expanded to 180 MW by May without explicit explanation (claim 6122), suggesting design scope increased or capacity accounting methods changed mid-stream.

Timeline performance reveals systematic slippage. The Ellendale recovery target of May 31, 2024 (claim 3235) passed without published resolution. The January 1, 2025 operational start for Polaris Forge 1's 100 MW facility (claim 6125, promised in May) is flagged delayed in the record, and the promised June 28, 2024 return to full Ellendale capacity (claim 3245) was already unmet when reiterated August 28. The Garden City sale to Marathon Digital, promised by April 1 (claims 6087, 6130), carried a delayed marker by May and appears dormant in subsequent records; no final completion notice appears through September 10.

The 400 MW hyperscaler lease remains a letter of intent (claims 3233, 3246, 5752) with no advancement toward signature or capacity reservation. Future-phase claims lack deadlines: 300 MW under design for Polaris Forge builds (claim 3248) and 1.4 GW across three campuses (claim 3249) carry no promised completion windows. The Jamestown GPU facility, cited as nearing completion in August (claims 5859, 5904) at 7.5 MW, shows no operational confirmation by early September.

Independent evidence supporting this pace is absent from the record. No permit approvals, interconnection queue entries, or construction milestone documentation substantiates the claimed progression. The pattern—consistent operating capacity, delayed milestones without revision, and future claims stripped of timeline specificity—suggests either execution headwinds or de-risking through less specific public guidance.

Over the next ninety days, watch for: actual Jamestown GPU online status and power consumption; final resolution or re-announcement of the Garden City transaction; any third-party capacity reservation announcement supporting the 400 MW LOI; and new timelines attached to the 300 MW Polaris Forge design or the 1.4 GW campus pipeline.

Applied Digital entered 2024 projecting rapid expansion, but the company's timeline narrative has experienced persistent compression without public revision. The carrier has held its core operating capacity figure at 286 MW (claims 5978, 5894, 5881)—consisting of facilities in Ellendale (180 MW), Jamestown (106 MW), and legacy Garden City infrastructure—unchanged from May through September, suggesting either consolidation of earlier ambitions or a portfolio that has not grown despite nine months of activity.

The most consequential slip involves the Ellendale HPC restoration following an early-2024 outage. Recovery was initially targeted for May 31, 2024 (claim 3235), then extended to June 28 (claim 3245), both marked delayed by late August. Simultaneously, the company advanced a 100 MW Polaris Forge 1 facility on the same Ellendale campus with an original operational start of January 1, 2025 (claims 6125, 6086). That timeline, announced as delayed in May, was reiterated unchanged through August and September without updated guidance.

A critical proof point sits unresolved: the Garden City, Texas sale to Marathon Digital Holdings. This 200 MW transfer was promised by April 1, 2024 (claim 6087), with the deadline repeatedly reappeared through August (claims 5881, 5882, 5981) marked consistently delayed. No public resolution has been announced.

Forward claims lack operational specificity. Applied Digital claims a 400 MW letter of intent with an unnamed U.S. hyperscaler (claims 3233, 5752), stated since April with no exclusivity resolution or revision. By August, the company announced 1.4 GW under development across three additional campuses (claim 3249), but attached no timelines, funding commitments, or site identifiers. Multiple vendor partnerships—HPE, Dell, Super Micro—were announced without delivery schedules tied to facilities.

The underlying pattern suggests a company managing expectations around a more constrained asset footprint than earlier projections indicated. The absence of new capacity announcements, coupled with repeated timeline extensions and the unresolved Garden City transaction, indicates execution risk on major milestones.

In the next 90 days, watch for: Polaris Forge 1 construction progress and any revised operational targets; Garden City transaction closure or formal withdrawal; utility power agreement execution details; and whether the claimed 1.4 GW receives site-level disclosure or funding announcement.

Applied Digital's facility trajectory reveals a pattern of ambitious announcements followed by prolonged delays and stalled monetization. From April 2024, the company committed to a 100 MW HPC facility at Polaris Forge 1 (Ellendale) operationalizing in January 2025 (claims 6125, 6124). That timeline has now slipped substantially with no formal revision in public record. The February 2024 Ellendale outage (claim 6275) initially justified a May 31 restoration (claim 3235), subsequently moved to June 28 (claim 3245)—itself a delay unacknowledged in follow-on statements.

Operating capacity has remained flat at 286 MW across Ellendale and Jamestown since June 2024 (claims 5978, 3244), comprising 180 MW and 106 MW respectively. Yet forward-looking claims have expanded: August 2024 announcements added 300 MW under design and 1.4 GW across three additional campuses under development (claims 3248, 3249). The critical 400 MW hyperscaler lease—initially an April 2024 LOI (claim 3233)—remained in exclusivity through September 2024, with extension deadlines slipping past July 26 unresolved (claims 5755, 3247).

The Garden City divestiture to Marathon Digital provides clearest evidence of execution friction. Originally promised April 1, 2024 (claim 6130), this 200 MW asset remained unsold through September 2024, representing both delayed capital and portfolio drag. The repeated filing without closure updates suggests ongoing friction or deprioritization.

Independent verification is sparse. No recent interconnection queue filings, permitting activity, or on-the-ground reporting validate Polaris Forge 1 construction progress. The silence since September 2024 on the 2025-01-01 operational target is notable—there has been no partial ramp, delay acknowledgment, or revised milestone.

For the next 90 days, watch: (1) Polaris Forge 1 operational or revised timeline announcement; (2) hyperscaler lease closure or restructuring; (3) Garden City transaction completion or formal abandonment; (4) permit filings for the 300 MW and 1.4 GW pipeline projects.

Applied Digital's North Dakota footprint demonstrates a pattern of slipped timelines and shifting capacity frameworks. The company initially anchored commitments around Ellendale recovery from a February 2024 outage, promising restoration by May 31 (claim 3235) and again by June 28 (claim 3245). Both dates were missed without revision until August, when the status shifted to delayed acknowledgment. The January 1, 2025 operational start for Polaris Forge 1's 100 MW HPC facility (claims 6125, 6086) is now at risk, with zero construction progress signals beyond repeated "under construction" language across claims filed May through September.

Capacity accounting shows materialization of the portfolio: May's announced 286 MW across Jamestown (106 MW) and Ellendale (180 MW) has held flat through September despite claims of active construction. The April announcement of 600 MW future potential at Ellendale (claim 3236) disaggregated into operating capacity (286 MW), near-term builds (300 MW under design, claim 3248), and three additional campuses totaling 1.4 GW (claim 3249). This suggests the company downgraded near-term expansion expectations while papering the portfolio with longer-cycle projects.

Three independent indicators warrant skepticism. First, the Garden City sale to Marathon Digital—promised April 1, 2024 (claims 6130, 6087, 5981, 5882)—remained "delayed" as of late August, a five-month slip with no communication of revised close date. Second, the 400 MW hyperscaler LOI at Ellendale (claims 3233, 3246) mentions an exclusivity condition through July 26 (claim 3247, marked delayed in August), yet no signed agreement has materialized; the final funding tranche conditioned on this lease (claim 6034) remains outstanding. Third, Jamestown's "nearing completion" status (August 29, claims 5904, 5859) shifted to "completion status" (September 10, claim 5773) without delivery evidence, suggesting linguistic repositioning rather than progress.

Watch for three signals in Q4: (1) a revised Polaris Forge 1 operational date—the January 2025 target is now implausible; (2) closure of the Garden City sale or public acknowledgment of termination; (3) filing of permits or interconnection queue entries for the 300 MW pipeline at Ellendale. Absent these, the 286 MW figure may remain the company's practical ceiling through 2025.

Applied Digital's capacity narrative has contracted significantly while delivery timelines have slipped materially. The company announced 600 MW of future potential at Ellendale (claim 3236) and a 400 MW hyperscaler lease LOI (claim 3233) in April 2024, but reset to 286 MW of operating or near-term capacity by May—180 MW at Ellendale and 106 MW at Jamestown (claims 6122, 6121). The 400 MW LOI persists in statements through August 2024 (claim 3246) but shows no execution milestones or binding language; no customer has been named. A 300 MW expansion is cited as "under design" (claim 3248) and three additional campuses totaling 1.4 GW are announced as "under development" (claim 3249), but these remain conceptual without facilities, addresses, or timeline commitments beyond Jamestown and Ellendale. The near-term footprint is thus roughly one-third the initially telegraphed range.

Timeline execution is the sharper concern. Polaris Forge 1 (HPC Ellendale) was promised operational by January 2025 (claims 6125, 6086). As of today, that deadline has slipped 20 months with no revised completion date visible; as of August 2024 the facility remained "under construction" (claim 5889). The Garden City, Texas facility sale to Marathon Digital was promised for April 2024 (claims 6130, 6087, 5981, 5882) and appears unresolved as of late August—mentioned repeatedly without either completion or formal timeline revision, a pattern suggesting internal stalling. The extended LOI exclusivity period for hyperscaler demand was promised for July 2024 (claim 3247) and appears missed.

No independent evidence enters the record—no interconnection queue positions, electrical interconnection agreements, power purchase confirmations, or construction permits appear. Partnerships cited (HPE, Super Micro, Dell) are structural arrangements, not proof of active build or customer pull.

In the next 90 days: watch for actual operational status at Polaris Forge 1 or a revised timeline; either completion of the Garden City divestiture or formal withdrawal; permit filings or construction notices for the 300 MW design-phase buildings; and evidence that the LOI translated to a binding customer commitment. Silent timeline extension without public revision is the highest-risk signal.

Applied Digital's capacity announcements reveal significant expansion rhetoric paired with persistent execution delays. From April through August 2024, operating capacity held steady at 286 MW across Ellendale and Jamestown facilities (claims 5978, 5894), with a 100 MW HPC facility under construction in Polaris Forge 1. However, future capacity claims escalated dramatically: by August 2024, APLD announced 1.4 GW across three additional campuses under development (claim 3249) and 300 MW in next-phase buildings under design (claim 3248), suggesting ambitious pipeline expansion without corresponding near-term delivery. The timeline trajectory shows systematic slippage. Polaris Forge 1's operational start was promised for January 2025 (claims 6125, 6086), yet records from May 2024 already flagged this date as delayed. Ellendale's restoration to full operating capacity, promised by June 2024 (claim 3245), was unmet by late August. Most acutely, the Garden City campus sale to Marathon Digital—promised by April 2024 (claims 6130, 6087, 5981)—remained unsigned four months past deadline as of August 2024 (claim 5882). A hyperscaler LOI worth 400 MW (claims 3246, 5775) required exclusivity extensions repeatedly pushed from July 26, 2024 onward (claims 3247, 5755), suggesting difficulty converting intent to commitment. Supporting evidence remains thin. No permits or interconnection queue filings appear in the record. Jamestown's GPU facility was claimed as nearing completion at 7.5 MW (claims 5859, 5904) in late August 2024, yet no operational announcements followed. The steady repetition of partnership claims across dozens of May-August releases—HPE, Super Micro, Dell agreements (claims 5876–5902)—without execution detail suggests heavy reliance on LOI or preliminary arrangements. Near-term observation should track: confirmation of Polaris Forge 1's actual operational status relative to the missed January 2025 date; Garden City sale closing or formal termination; hyperscaler lease signature; and Jamestown GPU facility energization. Any announcement of revised Polaris Forge 1 timelines without explicit acknowledgment of prior delays would signal continued execution friction.

Applied Digital has rewritten its Ellendale expansion narrative twice in nine months. The company first announced a 100 MW HPC facility under construction at Ellendale (claim 3238, 6124), then shifted to 180 MW of datacenter hosting capacity across the existing North Dakota footprint (claims 6122, 6084). That holding has remained stable—286 MW total across Ellendale and Jamestown—since at least June 2024 (claims 5978, 3244).

The company faces a clear pattern of timeline misses. Ellendale's capacity recovery target slipped from May 31, 2024 (claim 3235) to June 28, 2024 (claim 3245), both dates now passed. The 100 MW HPC facility promised operational by January 1, 2025 (claims 6125, 6086) is the next major deadline; current language describes it only as "under construction" (claim 5860) or "expected to become operational" (claim 5873), without a revised date attached.

The Garden City sale to Marathon Digital offers the clearest evidence of execution friction. Originally promised by April 1, 2024 (claims 6130, 6087), it remained unresolved five months later (claim 5882 on August 29). No revised timeline or explanation has been published. This 200 MW transaction's non-closure would erase a significant portion of announced capacity.

Applied Digital has anchored ambitions well beyond the current 286 MW. The company claims 1.4 GW across three additional campuses under development (claim 3249) and 300 MW in next two buildings under design at Ellendale (claim 3248). A 400 MW hyperscaler exclusivity LOI at Polaris Forge 1 (claims 3233, 3246) may be aspirational layering rather than committed near-term capacity.

No independent verification—permitting records, interconnection queue position, third-party facility reporting—is available in the claim set. Jamestown's HPC facility is described as "nearing completion" (claim 5904) but lacks an operational date. The company's recovery from a February 2024 power outage remains formally unexplained in public claims.

Watch whether Polaris Forge 1 clears the January 2025 operational bar or joins Ellendale's recovery and Garden City's sale in the slip column. Any resolution on Garden City or public revision to HPC timelines will signal either conviction or caution in the 1.4 GW pipeline.

Applied Digital's capacity claims stabilized at 286 MW across North Dakota while recovery timelines contracted repeatedly through mid-2024. From May onward, the company restated existing holdings—180 MW Ellendale, 106 MW Jamestown—with mechanical consistency across claims 6122, 5978, 3244, and recurring August-September iterations. Yet promised operational dates slipped without explicit revision. Ellendale's power recovery was initially targeted for May 31, 2024 (claim 3235), then pushed to June 28, 2024 (claim 3245), both dates passing without acknowledgment in subsequent disclosures. The flagship Polaris Forge 1 project, a 100 MW HPC facility at Ellendale, carries a January 1, 2025 operational deadline (claims 6125, 6086) marked as delayed across all eight months of available claims through September.

Capacity expansion narratives diverged sharply in late August 2024. While existing facilities remained pegged at 286 MW, Applied Digital announced 1.4 GW under development across three additional campuses (claim 3249) and 300 MW under design for next-phase buildings (claim 3248). Critically, these expansive figures float unmoored to delivery timelines. The sale of Garden City, Texas to Marathon Digital, promised by April 1, 2024 (claim 6130), persisted unresolved across at least four claim iterations (claims 6087, 5981, 5882), each tagged delayed. The Garden City asset nominally represents 200 MW of disposed capacity, yet its exit from the portfolio remains incomplete.

Jamestown's smaller 7.5 MW HPC facility was repeatedly characterized as nearing completion with no definitive operational target, suggesting protracted development. No independent evidence—regulatory filings, interconnection queue positions, or verified construction progress—substantiates the stated trajectories in the available record. Claims demonstrate pattern repetition rather than forward motion: nine separate disclosures restate the identical 286 MW North Dakota figure with no material updates.

Watch whether Polaris Forge 1 reaches operational status before the January 2025 deadline shifts further, whether the 1.4 GW announced in August enters permitting workflows, whether Garden City finally closes, and whether Jamestown's facilities achieve completion or remain indeterminate. The next 90 days will clarify whether new capacity claims represent executable roadmaps or portfolio inflation.

Applied Digital's datacenter trajectory reveals a pattern of ambitious capacity announcements paired with consistent timeline slippage and a widening gap between promised and operational capacity. As of August 2026, the company claims 1.4 gigawatts of contracted capacity (claim 487) and $36 billion in lease contracts (claim 485), yet only 100 megawatts remains operational (claim 884). This 14-fold gap between contracted and live capacity suggests execution challenges across the portfolio. Initial facilities have experienced material delays: Polaris Forge 1's first 100 megawatt data center missed its October 2025 target (claim 64), while Polaris Forge 2's groundbreaking slipped from August 2025 (claim 89) to at least December 2026 (claim 67). Delta Forge 1's groundbreaking, promised for January 2026 (claim 69), remained incomplete as of late August. Within Polaris Forge 1 specifically, the second building's 75 megawatt capacity addition missed a June 2026 delivery date and appeared only partially operational by August (claims 863, 869). The company withdrew its promised 300 megawatt Polaris Forge 3 capacity target (claim 338, stated May 2026) without public explanation, though it maintained delivery commitments for the facility through 2028 (claim 72). Recent months show inconsistency in timeline communication: Polaris Forge 2 capacity delivery was stated as March 2027 (claim 91) in late July but groundbreaking was separately promised by December 2026 (claim 67), creating ambiguity about the actual construction sequence. The ChronoScale separation, promised for May 2026 (claim 267), was still delayed in July announcements (claims 80, 100, 130). Over 90 days, watch for: whether Building 2 at Polaris Forge 1 achieves promised capacity; any permits or interconnection queue filings validating Polaris Forge 2 or Delta Forge timelines; third-party confirmation of the $36 billion lease pipeline's actual execution status; and whether announced capacity figures shift further without explanation.