jain.com — data centers, connectivity and security, analyzed daily

Category: AI Infrastructure

  • Applied Digital’s 1 GW Finland Deal Buys an AI Power Option, Not Yet a Campus

    Applied Digital’s 1 GW Finland Deal Buys an AI Power Option, Not Yet a Campus

    TL;DR · 30-second read

    The Short Version

    Applied Digital, a Dallas company that builds the huge computer warehouses used to run artificial intelligence, has signed a deal for access to a large supply of electricity in Finland. It is the company’s first project outside the United States.

    At full size, the site could use about as much power as one large nuclear reactor produces. But the first electricity is not expected until 2028, and no customer has signed yet.

    The lesson is that in this business, the electricity gets locked up first. The buildings and the paying tenants come later, if they come at all.

    Applied Digital (Nasdaq: APLD) announced on October 6, 2026 that it has entered into an agreement providing access to up to 1 gigawatt (GW) of potential power capacity in Finland, its first development opportunity outside the United States. The company plans a large-scale “AI Factory” campus that could scale to as much as 1 GW, with initial power availability anticipated beginning in 2028.

    Applied Digital said it has begun marketing the site and is in discussions with leading global hyperscale customers, the largest cloud and AI platform operators. It did not name the site’s location, the counterparty to the agreement, its financial terms or any prospective tenant.

    Executive Summary

    The Finland agreement gives Applied Digital a claim on a very large block of future electricity in a cool-climate market, the input that most constrains how fast AI data centers can be built. The company framed the deal as a measured first step abroad, structured to limit its initial exposure, with further investment contingent on customer demand, commercial milestones and regulatory approvals.

    That framing matters as much as the headline number. “Up to 1 GW of potential” capacity, power from 2028 and an unsigned customer base describe an option rather than a project under construction. The announcement establishes where Applied Digital wants to grow next and how it intends to manage the risk; it does not yet establish how much capacity will actually be built, when, or for whom.

    For industry readers, the deal is a test of whether a US developer’s site-selection model, built around abundant power and cold weather in North Dakota, transfers to the Nordics. For investors, it adds long-dated growth optionality while the company’s own guidance keeps near-term attention on its US pipeline.

    What “Up to 1 GW of Potential” Actually Buys

    Read closely, the release qualifies its headline figure at every step. The agreement provides access to “up to” 1 GW of “potential” power capacity; the campus is one “capable of scaling to as much as” 1 GW; and only “initial” power is anticipated, beginning in 2028. The company says future investment and development decisions will be tied to customer demand, commercial milestones and regulatory and development requirements, and that the deal was structured to limit initial exposure. Each of those statements points the same way: Applied Digital has secured the right to pursue a campus, not a commitment to build one at full scale.

    That is a rational way to buy the scarcest input in AI infrastructure. A gigawatt is roughly the output of one large nuclear reactor, and AI campuses of this size are gated less by land or buildings than by how quickly a large, reliable grid connection can be delivered. Securing power years before a tenant signs lets a developer offer hyperscalers a dated path to capacity, which is the thing those customers are shopping for. The trade-off is that the value of the option depends on terms the company has not disclosed, including what it pays to hold the capacity and what happens if no tenant materialises.

    The people most affected are prospective tenants, who gain another large-scale European option with a stated 2028 start, and shareholders, who should treat the 1 GW as a ceiling on what could be built rather than a measure of what has been contracted.

    Why Finland Fits the North Dakota Playbook

    Chief executive Wes Cummins tied the choice directly to the company’s US model: Finland “stood out because many of the characteristics that have supported our success in North Dakota are also present in the Nordics — a favorable climate, an attractive energy ecosystem, strong connectivity and the ability to develop at meaningful scale.” Each of those factors has an operational meaning. Cold ambient air reduces the energy spent removing heat from dense AI servers; a strong energy ecosystem determines whether large power blocks can be delivered; and fibre connectivity determines whether the site can serve customers beyond its immediate region.

    The release presents this as continuity, not reinvention: securing scalable power and land where AI capacity can be developed “efficiently, responsibly and repeatedly.” That is a reasonable thesis, but it is a claim to be proven. Cummins said as much, describing Finland as the chance “to begin proving that platform can travel.” The company’s own risk disclosures in the release flag what could differ abroad, including sourcing materials and labour internationally and complying with regulations outside the US.

    The 2028 Clock and the Hyperscaler Question

    With initial power not expected until 2028, the commercial work happens now. Applied Digital says it has begun marketing the site and is in discussions with leading global hyperscale customers. Large AI campuses are typically financed around long-term leases with creditworthy tenants, so the timing and size of any signed lease will largely determine whether the Finland project moves from option to construction, and how quickly it scales beyond its first phase.

    The release does not indicate how much of the 1 GW becomes available in 2028 versus later years. That phasing matters: a hyperscaler weighing the site will compare not just the headline ceiling but the megawatts deliverable in each year against competing sites. Until a tenant and a phasing schedule are disclosed, the near-term financial contribution of the Finland site should be treated as zero, consistent with the company’s own emphasis on its US projects.

    Discipline as the Message

    The release spends notable space reassuring investors. Cummins called the deal “a deliberate first step, not a change in our strategy,” and said scale “is not about planting flags in as many markets as possible.” That emphasis is consistent with the risk factors the company lists, which include its ability to raise additional capital for data center construction and to obtain financing for leases. Framing Finland as low-exposure optionality signals that the company does not intend to divert capital from US execution to fund it.

    Whether that framing holds will depend on disclosures still to come: the cost of holding the capacity, any deposits or milestone payments, and how a Finnish build would be financed if a tenant signs.

    Background

    Applied Digital, founded in 2021 and headquartered in Dallas, designs, builds and operates large data centers for high-performance computing, including artificial intelligence, cloud, networking and blockchain workloads. The company markets closed-loop cooling, a system that recirculates coolant rather than continuously consuming water, along with rapid deployment, and describes its build model as an “AI Factory” franchise. Its US development has centred on North Dakota, which the company credits for traits it now seeks in Finland: cold weather, available energy and room to scale. Datacloud named it Best Data Center in the Americas for 2025.

    Across the AI infrastructure market, developers increasingly compete to secure large blocks of electricity years ahead of construction, because access to reliable power at scale has become a principal constraint on how fast AI computing capacity can be added. The Nordics, which the company describes as an emerging European AI market, have drawn growing investment from global technology firms on the strength of their climate, energy and digital networks.

    Sources

    Source: Applied Digital Secures Up to 1 GW of Potential Power Capacity in Finland, Marking First International Development in the Nordics: Applied Digital’s October 6, 2026 announcement of its first development opportunity outside the United States.

  • AOI’s $588M Third Stock Sale of 2026 Shows Its AI Optics Ramp Runs on Equity

    AOI’s $588M Third Stock Sale of 2026 Shows Its AI Optics Ramp Runs on Equity

    TL;DR · 30-second read

    The Short Version

    Applied Optoelectronics, a Texas company, makes parts that carry data as pulses of light through glass fiber cables. The huge computer warehouses that power artificial intelligence need more of these parts every year.

    The company just finished selling about $600 million of brand-new stock. This is its third stock sale this year. Most of the money will buy factory machines so it can make more of these parts.

    The catch is that every new share makes each existing owner’s slice of the company a little smaller. Shareholders are paying for this expansion before it has earned anything.

    Applied Optoelectronics, Inc. (NASDAQ: AAOI), known as AOI, said in a release distributed via PR Newswire on October 5, 2026 that it has completed its $600 million “at-the-market” equity offering program. An at-the-market program lets a company sell newly issued shares gradually into the open market at prevailing prices. AOI sold 5,694,845 shares at an average price of $105.36, for proceeds of $587,999,985.89 after commissions and fees.

    The Sugar Land, Texas-based maker of optical and HFC (hybrid fiber-coaxial, the network design cable operators use) networking products first disclosed the program on August 21, 2026, when it filed a prospectus supplement with the U.S. Securities and Exchange Commission. Chief executive Dr. Thompson Lin described it as the company’s third at-the-market program this year.

    Executive Summary

    AOI has raised roughly $588 million in net proceeds by selling stock in about six and a half weeks. It intends to spend the money primarily on expanding manufacturing of optical devices for data centers. Those devices convert electrical signals into light and back, so data can travel over fiber between the servers and switches in AI computing clusters. The company says the money will go to “new equipment and machinery for production and research and development use” as it scales its manufacturing capacity. Formally, the proceeds are designated for general corporate purposes that may also include debt repayment and working capital.

    This matters because it shows how one AI-exposed component supplier is paying for growth. The company is not using contracted customer funding or debt that it has disclosed. It is selling new shares while its stock commands a high price, and then using that cash to buy production equipment. Doing this three times in one year is a pattern rather than an opportunistic one-off. AOI has chosen to build capacity ahead of revenue and to let shareholders carry the up-front cost through dilution.

    Shareholders Are Fronting the Factory Build

    The headline mechanism is straightforward. Building optical components at scale requires buying production and test equipment before the products it makes can be sold. AOI is paying for that gap with equity. Net proceeds of about $588 million are earmarked “primarily in support of its manufacturing expansion efforts for data center optical devices.” Lin framed the raise as supporting the company’s “ability to address growing demand for high-speed optical connectivity in AI data centers.” That makes the link between the capital and the AI buildout explicit. This was the third at-the-market program of 2026, so equity issuance has become a recurring funding channel for the expansion rather than a single balance-sheet top-up.

    The trade-off is dilution. AOI issued 5,694,845 new shares, and every one of them reduces existing holders’ percentage ownership of future profits. The bet is that the new capacity will earn back more per share than the dilution costs. That depends on utilization: whether the added lines run full on orders at acceptable prices. Equity carries no interest or repayment schedule, which makes it a forgiving way to fund capacity whose payback timing is uncertain. In return, the company gives up a share of the upside permanently instead of temporarily.

    Several groups are affected. Existing AOI shareholders absorb the dilution. Hyperscale and cloud buyers, the very large operators of AI data centers, stand to gain another expanding supply source for optical links if the capacity comes online. Competing optics suppliers face a better-capitalized rival that is adding production.

    What the Offering Mechanics Reveal

    The numbers say something about market appetite. At the $105.36 average price, 5,694,845 shares come to roughly $600 million gross, essentially the full program cap. Net proceeds of about $588 million imply commissions and fees of around $12 million, or roughly 2%. Selling the full $600 million between the August 21 disclosure and the October 5 completion shows the market absorbed a large volume of new stock over about 45 days without the program stalling short of its cap.

    An at-the-market program differs from a traditional follow-on offering. In a follow-on, a company prices a large block at a discount in one night. An ATM drips shares into normal trading, which usually costs less and lets the company sell more when the price is strong. The flip side is that investors face continuous supply of new stock while a program is open. With three programs in one year, that overhang has been a near-constant feature of the stock in 2026.

    Capacity Is the Bet, Customer Concentration Is the Risk

    Capacity only pays if it is used. AOI’s own risk disclosures in the announcement are frank about what could undermine utilization. They list reduced customer orders, “the company’s reliance on a small number of customers for a substantial portion of its revenues,” potential pricing pressure, and changes in the rate of design wins. A design win is when a customer selects a supplier’s part for a product. When a few buyers dominate revenue, a shift in one buyer’s deployment plans can leave expensive new equipment underused.

    Geography adds another variable. AOI manufactures at its Sugar Land headquarters and in Taipei, Taiwan, and Ningbo, China, and its risk factors cite tariffs, international trade regulation, and economic conditions in the United States and China. Where the new machinery is installed will shape its cost base and its exposure to trade policy. The equity raise has secured the funding. Whether that funding becomes profitable output depends on order flow and on where the equipment is placed.

    Background

    Applied Optoelectronics, Inc. is headquartered in Sugar Land, Texas, and trades on Nasdaq under the ticker AAOI. It develops and manufactures optical and hybrid fiber-coaxial networking products and describes them as building blocks for AI data centers, cable TV and broadband fiber access networks. It sells to what it calls tier-one customers in cloud computing, CATV broadband, telecom, and fiber-to-the-home markets. Its R&D is in Atlanta, and it manufactures in Sugar Land, Taipei, and Ningbo.

    The company has said it relies on a small number of customers for a substantial share of its revenue, which is common among suppliers to large cloud operators. As AI computing clusters grow, demand for high-speed optical connectivity inside data centers has become a central part of AOI’s pitch to investors. In 2026 the company has repeatedly turned to at-the-market stock sales to fund manufacturing expansion aimed at that demand.

    Sources

    Source: Applied Optoelectronics Announces Completion of Previously Announced At the Market Offering: AOI’s announcement of the completed $600 million at-the-market equity program, with share count, average price, net proceeds, and intended use of funds.

  • In Cisco’s Sovereign AI Pact With Bell, Consumption Pricing Tests a 35.9% Margin

    In Cisco’s Sovereign AI Pact With Bell, Consumption Pricing Tests a 35.9% Margin

    TL;DR · 30-second read

    The Short Version

    Cisco, a giant maker of the equipment that moves internet traffic, and Bell, one of Canada’s biggest phone and internet companies, have agreed in principle to offer artificial intelligence computing that stays entirely inside Canada.

    The target customers are government agencies and heavily regulated businesses, such as banks, that often cannot let sensitive data leave the country.

    The twist is how customers would pay. They would pay for what they use, like a utility bill, instead of buying costly equipment upfront. Someone still has to pay for that equipment. Neither company has said who, or put a price on the deal.

    Simply Wall St reported that on 29 September 2026, Bell Canada and Cisco Systems announced a memorandum of understanding to jointly develop sovereign AI infrastructure offerings for Canadian organizations. A memorandum of understanding is a preliminary agreement that sets out an intent to work together. The offerings would run in Bell’s domestic facilities and use Cisco’s AI, security and observability technologies. They emphasize modular infrastructure and flexible consumption models, and they are aimed at the public sector and regulated industries.

    No contract value, capacity figure, named customer or launch date was disclosed. Cisco shares traded at US$106.44 on 1 October 2026.

    Executive Summary

    On its face this is a modest step: a memorandum of understanding, not a purchase order. It does show how Cisco intends to sell AI infrastructure outside the hyperscale market, the handful of giant cloud operators behind its US$9.3 billion in hyperscale AI orders. The route is through a domestic carrier. Bell supplies the in-country location and Cisco supplies the networking, security and monitoring layer.

    The commercial detail worth watching is the consumption model. Paying for capacity as it is used fits public-sector budgets. AI deployments, however, are hardware-heavy, and hardware mix and memory costs already pressured Cisco’s gross margin in fiscal 2026. This partnership tests, in a small and visible setting, whether bundled software and security subscriptions can offset that pressure.

    For buyers in Canada’s regulated sectors, the deal points to a domestic option built on one integrated stack. For investors, it is a template rather than a revenue event, because nothing disclosed allows the deal to be sized.

    Sovereign AI Splits the Job Between Carrier and Vendor

    Sovereign AI means AI computing that is hosted in-country and run under domestic control, so sensitive data and models stay within a nation’s legal jurisdiction. The Bell–Cisco arrangement divides the work clearly. Bell brings domestic facilities. Cisco brings AI, security and observability technology, which is the tooling that monitors how systems and applications behave so operators can spot faults and attacks.

    That structure matters for a networking vendor. In the hyperscale market, Cisco sells to a small number of very large cloud operators that own their own buildings and buy at enormous volume. In sovereign deployments, the buyer’s first requirement is where the data physically sits and who operates the environment. That requirement favors a partner that already owns facilities inside the country, so a vendor that wants into the segment needs a local host.

    The partnership lets Cisco meet that qualifying condition in Canada, with the public sector and regulated markets as the stated targets. Data residency decides who gets considered. The MOU does not show whether residency outweighs price once vendors are compared. The first signed customers will answer that question.

    Consumption Pricing Is Where the 35.9% Margin Gets Tested

    Cisco’s non-GAAP operating margin is 35.9%. Non-GAAP means operating profit adjusted to exclude certain non-cash and one-time items. The main risk to that figure is mix. Hardware-heavy AI deployments earn lower margins than software and subscriptions, and hardware mix and memory costs already weighed on gross margin in fiscal 2026. The biggest risk to the investment case is sustained gross margin pressure if hardware grows faster than higher-margin software.

    Sovereign AI contains both sides of that equation. The physical build is hardware-heavy, since it requires networking, compute and security equipment placed in domestic facilities. The offering also bundles the software Cisco wants customers to consolidate on: Hypershield, AI Defense, and Splunk with Cisco Cloud Control. Under flexible consumption pricing, customers pay for capacity as they use it rather than buying equipment outright. That suits budget-constrained public bodies, but revenue arrives over time while someone pays for the hardware upfront. Two outcomes are possible:

    • If the recurring software and security layer is large relative to the hardware, consumption-priced sovereign deals could improve Cisco’s mix.
    • If hardware dominates and Cisco carries its cost, the deals deepen the pressure that margin forecasts already account for.

    An MOU of undisclosed size will not move a company-wide margin on its own. What it tests is the template. If Cisco repeats the carrier-hosted, consumption-priced model in other countries, the mix of each deal becomes a material input to whether the forecast expansion in profit margin, from 21.0% today to 24.1% by around 2029, actually happens.

    The Valuation Already Assumes the Mix Works

    Consensus models project revenue growing 9.6% a year to US$83.4 billion by 2029. They project earnings rising from US$13.3 billion to US$20.1 billion, with a bullish camp at US$23.1 billion. The most cautious analysts expect US$78.0 billion of revenue and US$19.3 billion of earnings, because they worry more about AI demand fading than about new deals like Bell’s. Reaching consensus price targets near US$137.25 would require the stock to trade at 34.5 times 2029 earnings. Today it trades at 31.6 times earnings, and the US communications sector trades at 35.4 times. A modest buyback, reducing share count by 0.22% a year, supports per-share earnings.

    Price targets range from US$115 to US$170, against a US$106.44 share price. Much of that spread reflects disagreement about how resilient margins will be as hardware-heavy AI and sovereign builds grow. Sovereign partnerships therefore cut both ways in the model. They open a growth avenue beyond the US$9.3 billion of hyperscale AI orders, and they add exposure to the mix risk analysts are already debating. Security-sensitive customers also raise the stakes of failure: a breach or outage in a sovereign deployment would land directly on the integrated stack Cisco is promoting.

    Background

    Cisco Systems designs and sells networking, security and observability technology used to run and protect the internet and corporate networks across the Americas, Europe, the Middle East, Africa and Asia-Pacific, and it is an established dividend payer. In 2024 it completed its acquisition of Splunk, the data-analytics and security software company. Splunk now sits at the center of Cisco’s effort to combine networking, security and monitoring into a single stack.

    Bell Canada, owned by BCE Inc., is one of Canada’s largest telecommunications companies. Sovereign AI has become a distinct market as governments and regulated industries look for AI computing that stays under domestic jurisdiction. That gives national carriers with in-country facilities a natural role as hosts for vendors that want to sell into it.

    Sources

    Source: Is Sovereign AI Partnership Altering The Investment Case For Cisco (CSCO)? (Simply Wall St). Analysis of the Bell Canada–Cisco sovereign AI memorandum of understanding and Cisco’s consensus analyst outlook.

  • Miners’ $1.5B Hashrate Loss Shows the AI Pivot Keeps the Power, Not the Machines

    Miners’ $1.5B Hashrate Loss Shows the AI Pivot Keeps the Power, Not the Machines

    TL;DR · 30-second read

    The Short Version

    Companies that mine Bitcoin have spent years building huge, power-hungry computer warehouses. Many now want to rent that electricity and space to artificial intelligence companies instead.

    The catch is that Bitcoin mining machines can only mine Bitcoin. In the first half of 2026, publicly listed miners lost mining machines worth about $1.5 billion, as Bitcoin prices fell and some sites switched to AI-style computing.

    The takeaway: in these businesses, the valuable part is the power hookup and the land. The machines are what gets thrown away.

    Tokenpost reported, in an item carried by Pluang on October 4, 2026, that publicly traded Bitcoin miners lost about 75 exahashes per second (EH/s) of mining capacity in the first half of 2026, equipment valued at roughly $1.5 billion. Exahashes per second measure how many guesses a mining fleet makes at the Bitcoin network’s puzzle each second; 75 EH/s is a large slice of the public miners’ combined fleets.

    The report ties the decline partly to miners such as Cipher Digital moving away from mining and toward high-performance computing (HPC), the dense, GPU-based computing used for AI training and inference, which has produced significant impairment charges. Separately, Core Scientific reported $266.5 million in equipment write-downs, which it attributed to lower Bitcoin prices and reduced mining revenue.

    Executive Summary

    The headline figure is a price tag on retirement. About 75 EH/s of mining capacity, worth roughly $1.5 billion in equipment, left public miners’ fleets in six months. Some of it was retired because Bitcoin’s economics weakened, and some because operators decided their power and sites would earn more hosting AI hardware than running mining rigs.

    That matters to anyone building or financing AI infrastructure. Bitcoin miners control a meaningful pool of already-energized industrial sites, the scarcest input in today’s AI data center buildout. The write-downs show what those sites cost to free up. The specialized mining hardware on them has little use outside Bitcoin, so it is written down, sold or scrapped, while the grid connection, land and electrical infrastructure carry forward into the new business.

    The numbers also carry a caution. Not all of the $1.5 billion is an AI cost. Core Scientific’s $266.5 million charge, the largest named figure, was attributed to falling Bitcoin prices, not to conversion. Readers should treat the total as a mix of cyclical mining losses and deliberate strategic exits.

    Why the Rigs Go and the Power Stays

    A Bitcoin mining rig is built around an ASIC, an application-specific integrated circuit. It is a chip designed to do one calculation, Bitcoin’s SHA-256 hash, extremely efficiently, and it can do nothing else. It cannot train a language model or serve AI queries. When a miner decides a site is worth more as an AI facility, the rigs have no role in the new business. Their book value has to be written down to what they will realistically fetch, which is the impairment charge the report describes at Cipher Digital.

    What carries over is everything around the machines: the utility interconnection, the substation, the land, the permits and the high-voltage electrical gear. For AI tenants those are the hard parts. New large grid connections in many US markets can take years to secure. A miner that already holds hundreds of megawatts of energized capacity has something an AI cloud provider cannot quickly build. The conversion trade is therefore asymmetric: the asset with a long lead time is kept, and the asset with a short life and a single purpose is discarded.

    The arithmetic makes the trade concrete. Roughly $1.5 billion across 75 EH/s implies about $20 million per exahash, or around $20 per terahash, of equipment value leaving the sector. For a converting miner, that is part of the entry cost of the AI business. It sits alongside the far larger spending needed to rebuild halls for liquid cooling and the much higher rack power densities that GPU clusters demand.

    Two Write-Downs, Two Different Causes

    The report groups two different phenomena under one number, and they should be separated. Cipher Digital’s charges are described as linked to the move toward HPC, a strategic decision to redeploy a site. Core Scientific’s $266.5 million in equipment write-downs is attributed to lower Bitcoin prices and reduced mining revenue, a cyclical decline in what its mining fleet can earn, regardless of any AI plans.

    That distinction matters for interpretation. If most of the $1.5 billion reflects Bitcoin’s weaker economics, the figure says more about mining profitability than about the cost of the AI pivot. The two forces also reinforce each other: when mining margins thin, the case for converting a site strengthens, and when a site is earmarked for conversion, the rigs on it lose their future cash flows. The aggregate figure does not show how the losses split between those causes.

    Impairments are also non-cash. They reduce reported earnings and the value of assets on the balance sheet, but they do not drain the bank account in the quarter they are booked. The cash was spent when the rigs were bought. The charge acknowledges that the spending will not be recovered at the original rate. The decisions that matter for cash come next: the capital spent to convert sites, and the lease revenue that conversion brings in.

    What It Means for AI Builders, Miners and Investors

    For AI developers and cloud providers hunting for power, retiring mining capacity is a supply signal. Every exahash taken offline for conversion frees electrical load that could, after rebuilding, host GPU clusters. Miners with large, well-located interconnections become potential landlords, and their negotiating position rests on the scarcity of that power rather than on their mining history.

    For miners that keep mining, retirements by rivals can help at the margin. Bitcoin’s protocol adjusts the difficulty of its puzzle roughly every two weeks to keep block times steady. When aggregate capacity falls, each remaining machine earns a larger share of rewards. That cushion depends on what happens network-wide, not only among public companies, and it does not offset a falling Bitcoin price.

    For investors, the report’s own framing is apt: the financial outcomes of conversion are uncertain. A write-down records the cost of leaving mining. It says nothing about whether a site will land a creditworthy tenant, on what terms, or when revenue starts. The pivot’s value is realized only when signed, energized leases replace the mining revenue the impaired rigs once produced.

    Background

    Bitcoin mining is the process of running specialized computers to secure the Bitcoin network in exchange for newly issued bitcoin and transaction fees. Large US public miners built fleets of hundreds of thousands of ASIC rigs at industrial sites chosen for cheap, abundant electricity. Their revenue rises and falls with Bitcoin’s price and with the network’s total computing power. The network also cuts the new-bitcoin reward in half roughly every four years, most recently in April 2024.

    Since the AI boom created a scramble for powered data center capacity, several miners have repositioned as infrastructure landlords. Core Scientific, one of the largest US public miners, has signed hosting agreements with AI cloud provider CoreWeave. Cipher Digital, which reports to the SEC as a public company and filed a Form 8-K current report on September 25, 2026, is among the miners shifting toward high-performance computing.

    Sources

    Source: Bitcoin miners lost $1.5B in mining capacity in H1 2026 amid shift to high-performance computing (Pluang, via Tokenpost): public miners shed about 75 EH/s of capacity as some, including Cipher Digital, move toward HPC, and Core Scientific reported $266.5 million in equipment write-downs.

    Primary sources: Cipher Digital, Form 8-K filed September 25, 2026 (SEC EDGAR).

  • Toshiba’s 30% HDD Share Goal Leaves AI Storage Pricing Riding on Supply Discipline

    Toshiba’s 30% HDD Share Goal Leaves AI Storage Pricing Riding on Supply Discipline

    TL;DR · 30-second read

    The Short Version

    Toshiba, one of only a handful of companies that still make the hard drives holding much of the world’s data, says it will build twice as many drives by its 2027 business year. It also wants to sell nearly a third of all of them.

    Investors took that as a threat to rival Seagate. Its stock fell about 10% in a single day, after its value more than tripled in a year.

    The reason: the giant computer buildings behind artificial intelligence need enormous storage, and tight supply let drive makers charge more. More drives could mean cheaper storage, and thinner profits for the makers.

    Toshiba plans to double its hard disk drive (HDD) output by fiscal 2027 and is targeting a 30% share of the market, Simply Wall St reported on October 3, 2026. Seagate Technology Holdings (Nasdaq: STX) shares fell about 10% in a day and about 7% over the week, to a last close of $848.99.

    Western Digital (Nasdaq: WDC) shares also moved lower on Toshiba’s expansion plans, Yahoo Finance reported.

    Executive Summary

    Toshiba’s plan has two parts that matter: a volume goal, doubling hard drive output by fiscal 2027, and a market-share goal of 30%. Together, they signal that a third manufacturer intends to grow its supply faster than the market grows. That is the scenario most likely to weaken the pricing power Seagate and Western Digital have enjoyed as AI and cloud data centers absorbed high-capacity drives.

    The market’s reaction was immediate. Seagate’s one-day drop of roughly 10% came after a one-year total shareholder return above 200%, and while the stock trades at 60.6 times earnings. Peers trade at 26 times earnings. A plan with a 2027 horizon could move the stock that much because Seagate’s valuation assumes today’s margins will last.

    The analytical takeaway is that the AI storage boom has not ended. Nothing in the announcement points to weaker demand. What the move shows is that hard drive pricing depends heavily on how much new capacity the few manufacturers choose to add. Seagate’s main counterargument is a shift toward its higher-capacity HAMR drives, which would let it grow margins without relying on price.

    Why a Share Target Hits Scarcity Pricing Hardest

    The hard drives that fill AI and cloud data centers are high-capacity “nearline” drives. They store data that must stay accessible but is not read constantly, such as training datasets, archives and backups. These drives come from a very small group of manufacturers, with Seagate, Western Digital and Toshiba at the center. In a market that concentrated, prices depend less on how much customers want and more on whether any maker adds supply faster than demand grows. The most popular bullish thesis on Seagate describes it as a position “built on physical scarcity” and names its own key risk plainly: that a supplier “loosens capacity discipline.”

    Toshiba’s plan turns that risk into a stated strategy. Doubling output while targeting 30% of the market is a volume objective, and a share target can only be met by taking sales from someone else. That explains why the reaction landed on Seagate and Western Digital and not on indicators of AI demand. The announcement suggests no slowdown in data center storage needs. What changed is the assumption that supply would stay tight long enough for prices to keep rising.

    Two groups feel this most directly. Cloud providers and data center operators buy drives in bulk and have been negotiating in a seller’s market. They now have a named third source to cite well before any additional Toshiba units ship. On the other side are Seagate and Western Digital, whose valuations reflect scarcity lasting for some time. The plan does not end scarcity on its own. It is a target, and its completion date is in fiscal 2027. But in a concentrated market, an expected change in supply can shift bargaining power before the actual supply arrives.

    A 10% Day After a 200% Year

    Seagate’s one-year total shareholder return is above 200%, and its five-year return is more than tenfold. Its price-to-earnings ratio, meaning what investors pay for each dollar of annual profit, is 60.6x. That compares with a peer average of 26x and a global tech average of 19.5x. It is also above a “fair ratio” estimate of 57.5x.

    A multiple like that assumes elevated margins will persist. When a stock is priced that way, a plan with no disclosed spending figure or plant-level detail can still move it 10% in a day. Investors are not asking whether Toshiba will actually ship twice as many drives by 2027. They are asking whether the possibility shortens how long scarcity pricing lasts. The size of the drop says more about how much durability investors had priced in than about how much new supply is actually coming.

    Views on what the pullback means still diverge. The leading bullish narrative puts Seagate’s fair value at $1,158.73, about 27% above its last close of $848.99. The earnings multiple tells a more cautious story. Both readings can hold at once: the stock has upside if margins stay where they are, and it is exposed if they do not.

    HAMR Is the Counterargument, and It Runs on a Different Clock

    HAMR, or heat-assisted magnetic recording, uses a tiny laser to briefly heat the spot on a disk being written. This lets data be packed more densely, so each drive holds more. The bull case argues that Seagate’s margins do not depend only on price. One thesis models HAMR’s share of Seagate’s mix rising from 40% to 70%. It argues that the shift “happens on a qualification schedule, not a price schedule.” Qualification is the lengthy testing process large buyers run before deploying a drive model at scale.

    This is the strongest response to the supply-discipline argument. If more supply pressures prices per drive, more capacity per drive can offset some of that pressure. But the same thesis flags the counter-risk: Western Digital could close the HAMR gap faster than expected. Toshiba has framed its plan in terms of output and share. It has not said which capacity tiers or recording technologies its added units will use. That detail decides whether new Toshiba supply competes directly with the highest-capacity drives AI buyers prioritize, or mostly with lower tiers.

    For storage buyers, the practical reading is to watch qualification timelines more closely than headline output figures. Extra drives only change the pricing picture once customers have tested and approved them for large deployments. The timing of those approvals, for both Toshiba’s added volume and Seagate’s HAMR ramp, will show whether this was an early signal or an overreaction.

    Background

    Seagate Technology Holdings provides data storage solutions in Singapore, the United States, the Netherlands and internationally. Alongside Western Digital and Toshiba, it is one of the few companies still manufacturing hard disk drives at scale. Hard drives have stayed central to data centers because they store large volumes of data cheaply, even as faster flash storage handles more performance-sensitive work.

    The AI buildout has turned high-capacity hard drives into a constrained product. Tight supply has supported pricing and margins for the incumbents, and Seagate’s share price rose more than tenfold over five years. Seagate’s push into HAMR drives, which store more data per unit, has become a central part of investor debate about whether those margins can outlast the current supply squeeze.

    Sources

    Source: Is Seagate Technology Holdings (STX) A Bargain After Toshiba’s HDD Expansion Plan? (Simply Wall St): Seagate’s share reaction and valuation after Toshiba’s plan to double HDD output. Also: Why Toshiba’s Expansion Plans Just Sent Western Digital Stock Lower (Yahoo Finance): Western Digital’s share reaction to the same news.

  • Musk’s TSMC Talks Show a 1.4nm Captive AI Fab Still Rents Its Process Edge

    Musk’s TSMC Talks Show a 1.4nm Captive AI Fab Still Rents Its Process Edge

    TL;DR · 30-second read

    The Short Version

    Elon Musk is planning a huge chip factory, called Terafab, that would make chips only for his own companies: the carmaker Tesla, the rocket maker SpaceX and his artificial intelligence company xAI.

    On Saturday he confirmed early talks with Taiwan Semiconductor Manufacturing Company, the world’s biggest maker of chips for other firms. Intel is already lined up to supply its most advanced manufacturing recipe.

    The surprise is that owning the factory is not enough. Money pays for the building, but the know-how to make the most advanced chips still has to come from one of a handful of outside companies.

    Elon Musk confirmed on Saturday, October 3, that Terafab, the chip-manufacturing initiative meant to supply only Tesla, SpaceX and xAI, is in discussions with TSMC, the world’s largest contract chipmaker. Tim Culpan’s Culpium newsletter reported the talks on Friday. Musk wrote on X, “Just discussions, but something may come of it.” TSMC has not commented.

    Intel is the only other named Terafab partner. It is slated to supply its 14A process, a 1.4-nanometer-class manufacturing technology. Culpium described two possible structures. In the first, TSMC would own and operate the fab while SpaceX/Terafab invests, commits to buy a guaranteed volume of chips, or does both. In the second, SpaceX would hold a majority stake while TSMC operates the plant and supplies process technology. Culpium also linked TSMC’s rumored Texas fab complex to an effort to support Terafab.

    Executive Summary

    Terafab is a captive fab, meaning a chip factory that serves its owner rather than taking orders from the open market. Musk’s confirmation that it is talking to TSMC, while Intel is already named as its process supplier, shows how such a fab gets access to leading-edge manufacturing. In every arrangement described so far, the process technology comes from an established foundry. A process is the full recipe of hundreds of manufacturing steps that turns a chip design into working silicon.

    That matters for AI infrastructure because the custom chips behind Tesla’s vehicles, SpaceX’s systems and xAI’s models are only as available as the factories able to make them. Musk’s companies can supply capital and guaranteed demand. The talks suggest they cannot supply the process expertise themselves, at least not yet, and must secure it from Intel, TSMC or both.

    Nothing is agreed. Musk called the talks “just discussions,” TSMC is silent, and the ownership structure, the process choice and Intel’s future role are all open.

    The Process Is the Part Musk Cannot Build Himself

    Look at the arrangements that have been named so far. The constant across all of them is where the manufacturing know-how comes from. Today, Terafab’s named process is Intel’s 14A, a 1.4-nanometer-class technology. Nanometer figures are now generational labels rather than literal feature sizes, and smaller numbers denote newer, denser chips. In Culpium’s “Plan A,” TSMC would own and run the fab and bring its own process. In “Plan B,” SpaceX would hold the majority stake, yet TSMC would still provide “operational expertise” and its process technologies. Ownership changes from one scenario to the next. The source of the process does not.

    The economics explain why. A foundry, a company that manufactures chips designed by others, develops each process and refines its yield over years. Yield is the share of chips on each wafer that work. The foundry recovers that cost and learning across an open order book of many customers. A captive fab serving three affiliated companies brings concentrated demand but not that breadth. What Terafab can offer is the same thing TSMC’s joint-venture partners offer in Japan and Germany: capital and committed volume. That is why the talks look less like a factory purchase and more like an effort to license or borrow a capability.

    Intellectual property adds a second layer. Tom’s Hardware points out that Intel holds a broad cross-patent license with TSMC, meaning each can use the other’s patents, while Terafab has no such license. On that reading, running an Intel-developed process at a non-Intel fab could expose Terafab to TSMC patent claims. That gives even an Intel-based Terafab a reason to keep TSMC close. TSMC is also a leading supplier of advanced packaging, the step that assembles multiple chips into a single high-performance module, and packaging is a common choke point for AI accelerators.

    Plan B Would Ask TSMC to Do Something New

    Plan A follows a known template. TSMC’s JASM venture in Japan and ESMC venture in Germany pair local partners, who supply money and demand, with TSMC, which supplies process technology and operates the fabs. Plan B would be different: TSMC would run a fab that it does not control, for a closed set of customers. For TSMC, the trade-off would be guaranteed volume on one side and, on the other, putting its process know-how and engineers into a facility owned by someone else.

    TSMC’s capacity is not open-ended. It is allocated by its board in large monthly tranches. In a 6-K filed September 24, TSMC reported that its board approved, for August 2026, $16,035 million for machinery for advanced-technology capacity, $4,791 million for advanced packaging, mature and specialty capacity, and $8,616 million for real estate and capitalized leased assets, about $29.4 billion in total. The filing does not break those figures out by site or mention Terafab. Any TSMC role in Terafab would compete with that existing program for tools, engineers and management attention.

    Intel’s 14A Role Is the Variable to Watch

    For Intel Foundry, Terafab is a high-profile external customer for 14A. If TSMC joins and brings its own node, 14A could drop out of the project. Tom’s Hardware also describes a middle path, in which Terafab produces wafers on Intel’s 14A and TSMC packages them in Texas. That split would keep both suppliers involved.

    Intel has said it expects 14A to perform within 5% of TSMC’s A14. If that holds, raw transistor performance is not the strongest argument for bringing TSMC in. The stronger arguments are TSMC’s record of operating fabs at volume, its packaging capacity and its patent position. Those are the parts of a foundry’s value that a new captive fab finds hardest to replicate.

    Financing Looks Like the Easier Constraint

    Capital is the input Musk’s companies are best placed to provide. In an 8-K filed September 29, Tesla disclosed $30 billion of new unsecured credit facilities. They are a $20 billion three-year delayed-draw term loan, which can be drawn in stages over 18 months, an $8 billion five-year revolving facility and a $2 billion 364-day revolver. The revolvers can be expanded by up to $4 billion. Tesla said the proceeds are for general corporate purposes, and nothing had been drawn at signing. The filing does not connect the facilities to Terafab, but it shows the group can raise large sums of money on short notice.

    Demand also exists. Tesla told investors on October 2 that it delivered more than 486,000 vehicles in the third quarter and deployed 13.7 gigawatt-hours of energy storage. As Tom’s Hardware frames it, Musk wants an uninterrupted, surplus supply of custom silicon, not a foundry that competes with Intel, Samsung or TSMC. Capital and demand are what that goal brings to the table. Leading-edge process access is what it still has to obtain from an outside foundry.

    Background

    TSMC is the world’s largest contract chipmaker. It manufactures chips that other companies design, including many of the processors used in AI data centers. It has expanded outside Taiwan through joint ventures in Japan (JASM) and Germany (ESMC) and through large fab investments in Arizona. Intel is building its own foundry business around new processes such as 14A, and Samsung Foundry is the other major leading-edge contract manufacturer.

    Musk’s companies already design custom chips. Tesla’s AI5, a 2-nanometer-class chip, taped out at both TSMC and Samsung Foundry. Terafab is intended to take that a step further: a dedicated manufacturing operation whose entire output serves Tesla, SpaceX and xAI, with the goal of an uninterrupted supply of chips for AI, vehicles and spacecraft.

    Sources

    Source: Elon Musk confirms discussions with TSMC about Terafab chipmaking collaboration — Intel is the only other named partner, Terafab to exclusively supply Tesla, SpaceX, and xAI (Tom’s Hardware). Musk confirms TSMC talks on Terafab, a captive fab for his companies, with Intel’s 14A as the named process.

    Primary sources: TSMC Form 6-K filed September 24, 2026 (August 2026 capital appropriations); Tesla Form 8-K filed October 2, 2026; Tesla Exhibit 99.1: Third Quarter 2026 Production, Deliveries & Deployments; Tesla Form 8-K filed September 29, 2026 ($30 billion credit facilities).

  • Applied Digital’s Energized 150 MW Puts Its AI Lease Backlog to a Revenue Test

    Applied Digital’s Energized 150 MW Puts Its AI Lease Backlog to a Revenue Test

    TL;DR · 30-second read

    The Short Version

    • Applied Digital just finished switching on a huge computing building for artificial intelligence. Its electricity capacity is comparable to a small city’s demand.
    • Its main customer, CoreWeave, rents out artificial intelligence computing power. CoreWeave is reported to have signed 15-year rental deals worth about $7 billion in total.
    • Signing a deal is not the same as getting paid. Rent usually starts once the space is powered and ready, and this building now is.
    • Investors remain cautious. The shares are down nearly a quarter in three months, and the company carries heavy debt.

    Applied Digital (Nasdaq: APLD) has switched on the second phase of Building 2 at Polaris Forge 1, its campus purpose-built for artificial intelligence and high-performance computing (HPC), Simply Wall St reported on October 2, 2026. The phase adds 75 megawatts (MW) of IT capacity. IT capacity is the power available to run servers and GPUs, as distinct from cooling and other building loads. The phase completes the facility’s 150 MW buildout.

    The shares rose 5.05% on the day to US$25.38. They remain down 23.23% over three months, though the three-year total shareholder return is still roughly fourfold.

    Executive Summary

    The milestone itself is simple. A second 75 MW block is now live at Building 2, and the facility’s planned 150 MW is complete. Its significance lies in what it does to the company’s revenue story. Polaris Forge 1 is tied to long-term leases with CoreWeave, a cloud provider that rents out AI computing power. Those leases are described as 15-year agreements worth $7 billion in total contracted revenue. A contract of that kind is a promise to pay. Energized, delivered capacity is generally what turns the promise into rent that appears on an income statement.

    That conversion is where the valuation debate now sits. On a price-to-sales basis, the stock trades at about 12.1 times revenue, against roughly 1.9 times for the broader US IT sector. Bullish valuation models put fair value well above the current price, but they depend on revenue that has not yet been reported. With the 150 MW buildout finished, the coming quarters should begin to show how much of the backlog is converting into revenue, and on what timetable.

    The risks are the ones an energization milestone does not remove. The company depends on a small number of hyperscale tenants, meaning very large cloud and AI compute buyers, and it carries a sizeable debt load. If contracts slip or financing tightens, the upside case is exposed quickly.

    Contracts Are Promises. Energized Megawatts Are Invoices.

    In wholesale data center leasing, rent typically starts when a tenant’s capacity is delivered and accepted, not when the contract is signed. A signed lease establishes backlog, the revenue a company is entitled to over time. Building, powering and handing over the space is what starts the billing clock. Switching on the second 75 MW phase of Building 2 completes the 150 MW facility at Polaris Forge 1. The question for Applied Digital is no longer whether it can finish this building. It is how fast the capacity turns into recognized revenue.

    The scale of that backlog gives the milestone its weight. The most-followed valuation narrative on Applied Digital describes the CoreWeave agreements as 15-year leases totalling $7 billion in contracted revenue. Spread evenly, that works out to roughly $467 million a year. That figure is simple arithmetic, not a disclosed payment schedule, and real lease revenue usually ramps as phases come online. It still shows the size of the revenue stream that delivered capacity is meant to unlock. Until the building is live, that stream exists only on paper.

    Three groups are most directly affected. Shareholders are paying today for revenue that energization begins to release. Lenders’ repayment depends on that rent arriving on schedule. CoreWeave depends on Applied Digital to house the computing capacity it in turn rents to its own customers.

    Why a 12.1x Sales Multiple Rides on Delivery

    Simply Wall St’s own price-to-sales comparison puts the stock at about 12.1 times sales. That compares with roughly 1.9 times for the wider US IT group and a 5.5 times multiple the publication treats as a fair ratio. A high price-to-sales ratio means investors are paying for revenue they expect rather than revenue already booked. As the publication notes, a gap like this can close through stronger fundamentals, a lower share price, or both. Delivered megawatts are the most direct route to stronger fundamentals. Each phase that starts paying rent raises the sales line without the share price having to fall.

    The bullish narrative assumes much more than one campus. Its US$68.35 fair value, about 63% above the US$25.38 close, rests on an 11.07% discount rate. It also assumes Applied Digital reaches roughly US$3.0 billion in revenue and US$295.0 million in earnings by around 2029, which still implies a forward price-to-earnings multiple near 104.7 times. Against that US$3.0 billion target, the CoreWeave leases at their simple annual average would cover only about a sixth. The remainder of the bull case therefore depends on capacity beyond this 150 MW facility, along with the financing and tenants needed to fill it.

    A fair reading cuts both ways. Completing the facility is real execution, and it narrows the gap between contracted and earned revenue. It does not, by itself, validate a valuation built on several times more revenue than this campus is reported to be contracted for.

    The Risks Energization Does Not Retire

    Bringing capacity online moves risk without eliminating it. Construction and delivery risk on this building falls away. Counterparty risk, the chance a tenant pays late, renegotiates or fails to take full occupancy, becomes more prominent as revenue concentrates on the leases now running. The valuation commentary flags dependence on a few hyperscaler tenants as a central risk. With CoreWeave named as the anchor customer at Polaris Forge 1, much of the campus’s near-term revenue rests on one counterparty.

    The balance sheet is the other pressure point. Data center campuses are capital-intensive, and the sizeable debt load cited in the valuation commentary must be serviced whether or not rent ramps on schedule. The recent three-month decline of 23.23%, after a roughly fourfold three-year return, suggests the market is giving more weight to these execution and funding questions than it did earlier in the cycle. A completed building is a necessary condition for the bull case, but not a sufficient one.

    Background

    Applied Digital designs, develops and operates digital infrastructure for high-performance computing and artificial intelligence customers in North America, and trades on the Nasdaq Global Select Market under the ticker APLD. Its fiscal year runs from June 1 to May 31, according to the definitive proxy statement it filed with the SEC on September 25, 2026.

    The company has positioned Polaris Forge 1 as a purpose-built AI campus anchored by long-term leases with CoreWeave. CoreWeave is one of the specialist cloud providers, often grouped with hyperscalers, the largest buyers of data center capacity, that rent GPU computing to AI developers. Investors have rewarded that strategy with a roughly fourfold three-year shareholder return, though the shares have cooled sharply in recent months as attention has shifted to execution and funding.

    Sources

    Source: Applied Digital (APLD) Completes Polaris Forge 1 Expansion, Is The Valuation Gap Too Wide? (Simply Wall St). Report on the second 75 MW phase of Building 2 completing Polaris Forge 1’s 150 MW buildout, with valuation commentary.

    Primary sources: Applied Digital, Form DEF 14A (definitive proxy statement), filed with the SEC on September 25, 2026.

  • For Anthropic’s 401 MW AI Lease, TeraWulf’s Real Asset Is a Smelter’s Grid Connection

    For Anthropic’s 401 MW AI Lease, TeraWulf’s Real Asset Is a Smelter’s Grid Connection

    TL;DR · 30-second read

    The Short Version

    TeraWulf, a company that used to mine bitcoin, has lined up the artificial intelligence developer Anthropic as the tenant for a giant computer campus on the site of an old aluminum plant in Kentucky.

    The lease runs 20 years and is estimated to be worth about $19 billion. Its computers will use about as much electricity as a small city. Averaged out, the yearly rent would be more than five times what TeraWulf now takes in from sales in a year.

    The takeaway: for artificial intelligence, the scarce ingredient is electricity, and an old factory already wired for enormous power loads was the prize.

    StocksToTrade reported on October 2, 2026 that TeraWulf Inc. (Nasdaq: WULF) has secured Kentucky Public Service Commission approval for a Retail Electric Service Agreement providing up to 482 megawatts (MW) of power to its Justified Data Campus, a repurposed former aluminum smelter, and that the campus is anchored by a 20-year lease with Anthropic covering roughly 401 MW of critical IT load. The lease carries options to extend by up to 10 more years and an estimated $19 billion of contracted revenue, backed by expected investment-grade credit. Initial investment in the project is estimated at $4.0–4.5 billion.

    WULF shares rose 4.15% on the day, closing near $15.49 after a pullback from the $17–18 range in early September. The same period brought a run of analyst initiations: Wells Fargo, UBS, Freedom Capital and Jones started the stock at Buy or Overweight with targets from $19 to $30, while Rothschild & Co Redburn started at Neutral with a $15 target.

    Executive Summary

    TeraWulf, a U.S. company that built its business mining bitcoin, is turning a former Kentucky aluminum smelter into an AI data center campus with Anthropic as its anchor tenant. The two numbers at the center of the deal — 482 MW of approved power and a 401 MW lease — sit close enough together that the lease reads as sized to fit inside the power the state regulator signed off on.

    That is the significance for the wider market. AI developers need enormous, reliable electricity, and power access is widely described as the main bottleneck in building data centers for them. A site that already carries heavy industrial power infrastructure, paired with a regulatory approval, gives a landlord something a tenant will commit to for 20 years before the buildings exist.

    For TeraWulf, the lease implies revenue on a scale far beyond anything its mining business has produced. It also means the story now turns on execution and financing: roughly $4 billion or more must be spent before that revenue arrives.

    The Lease Fits Inside the Power Approval

    Two numbers explain this deal. The Kentucky Public Service Commission, the state’s utility regulator, approved a Retail Electric Service Agreement — in effect a large-customer supply contract with the local utility — for up to 482 MW at the Justified Data Campus. Anthropic’s lease covers roughly 401 MW of critical IT load, meaning the power reserved for the servers themselves, before cooling, power conversion and building systems. The difference is about 81 MW, so the approved supply is roughly 1.2 times the IT load. That ratio is consistent with the overhead a modern, efficient facility needs for cooling and electrical losses, which suggests the tenant commitment was sized to the power envelope rather than the other way round.

    That ordering matters for anyone valuing a converted miner. The expensive part of the campus is still ahead: the $4.0–4.5 billion of initial investment is planned spending, not completed construction. What TeraWulf could offer a 20-year tenant today is a site with heavy power infrastructure already in place and a regulatory path to nearly half a gigawatt of supply. In that sense the grid connection — not the buildings, and not bitcoin mining hardware — is the asset Anthropic is committing to.

    The consequence reaches several groups. Utilities and state regulators become gatekeepers of AI leasing, because their approvals set how much capacity a landlord can sell. Owners of retired heavy-industry sites hold something newly valuable. And tenants sign multi-decade commitments whose value depends on construction that has not yet happened.

    Why an Old Smelter Beat a Greenfield Site

    Aluminum smelting is among the most electricity-hungry industrial processes, so smelter sites are typically built with high-voltage connections and substations sized for very large, continuous loads. Repurposing one is a “brownfield” project — building on previously developed industrial land — as opposed to a “greenfield” build on empty land that must wait for new transmission and utility upgrades. Wells Fargo, launching coverage with an Overweight rating and a $30 target, highlighted exactly this point: existing infrastructure cuts both time and capital spending compared with starting from scratch.

    Notably, the flagship is not a converted bitcoin mine. What TeraWulf’s mining history appears to contribute is experience in finding, contracting and energizing industrial-scale power, rather than the mining sites themselves. That distinction matters for investors who value miners on the megawatts they already run: this deal suggests the premium goes to power that can be delivered at data-center quality and scale, wherever it sits, and to the ability to get it approved.

    It also shows who is signing. Anthropic, an AI developer, is named as the direct tenant of a former miner. One deal does not make an industry trend, but Wells Fargo’s reference to an 839 MW contracted capacity pipeline across three tenants indicates TeraWulf is building a portfolio around this model rather than a one-off project.

    $19 Billion of Revenue, $4 Billion-Plus of Spending First

    Simple division puts the estimated $19 billion of contracted revenue at an average of about $950 million a year over the 20-year base term — a rough figure that ignores rent escalators and the construction ramp. That is more than five times TeraWulf’s trailing revenue of about $168.5 million. Against it sits an estimated $4.0–4.5 billion of initial investment, or roughly $10–11 million per megawatt of IT load.

    The spending comes first. TeraWulf reported free cash flow of about −$992 million in its latest quarter, raised more than $1.1 billion of equity and put over $1.0 billion into property and equipment. Debt is modest relative to equity, at about 0.16, but working capital is negative and the current ratio — short-term assets divided by short-term liabilities — is below 1. In plain terms, the company is spending well ahead of the cash the lease will eventually produce.

    This is where “expected investment-grade credit” carries weight. A long lease backed by investment-grade credit — a rating signaling low default risk — is the kind of contract lenders will finance against. If that support materializes as described, much of the build could be funded with debt rather than further share sales; if it does not, the burden shifts back toward equity and dilution of existing shareholders.

    Where the Bulls and the Skeptics Actually Disagree

    Analysts largely agree on the asset and differ on the backdrop. Wells Fargo ($30, Overweight), UBS ($24, Buy, arguing a fully leased development portfolio could be worth up to $30 a share), Freedom Capital ($19, Buy, with a “floor” near $14) and Jones ($30, Buy) frame TeraWulf as a favored AI infrastructure name, and FactSet data put the consensus at Buy with a mean target in the mid-$30s.

    Rothschild & Co Redburn’s Neutral rating and $15 target — close to where the stock closed — rests on a different question: not whether the Kentucky campus is valuable, but whether tightening credit markets and the balance sheets of hyperscalers, the largest cloud providers, will keep funding AI buildout at the current pace. That is a fair challenge to a business built on long-dated tenant commitments and borrowing against them. Equally, the bull case rests on a named tenant and a regulatory approval, which are concrete rather than speculative.

    Concentration is the risk both camps must weigh. One tenant anchors the flagship, and a 20-year lease is only as strong as that tenant’s ability to pay through cycles in AI spending. Delivering 401 MW of capacity on schedule is the near-term test that will settle which view was closer.

    Background

    TeraWulf Inc. (Nasdaq: WULF) built its business as a U.S. bitcoin miner, running large fleets of specialized computers that draw electricity around the clock. That model depends on securing large amounts of reliable, low-cost power — the same input that AI data centers need. As demand for AI computing has surged, TeraWulf and other public miners have begun converting capacity into data centers that host AI workloads, shifting from mining revenue toward long-term leases.

    Anthropic is an AI developer whose models require very large amounts of computing capacity. In data center leasing, “critical IT load” measures the power available to the servers, and state utility regulators such as the Kentucky Public Service Commission approve the supply agreements that make large new loads possible — which is why power approvals have become central milestones in AI infrastructure projects.

    Sources

    Source: WULF Stock Rallies As Massive Anthropic AI Deal Takes Shape — StocksToTrade on TeraWulf’s Kentucky power approval, Anthropic lease and analyst coverage.

  • Against a $95B AI Backlog, Dell’s Japan Deal Is About Power Access, Not Scale

    Against a $95B AI Backlog, Dell’s Japan Deal Is About Power Access, Not Scale

    TL;DR · 30-second read

    The Short Version

    Dell, the American company known for laptops and for the powerful computers that run artificial intelligence, has joined a project to build an artificial intelligence computing center in Japan. Its partners include JERA, Japan’s biggest electricity producer, and a company called RHAELM.

    The project is designed around one problem: securing a steady supply of electricity, which these computers use in huge amounts. It is due to open in 2028.

    One oddity: the price tag was given as both 15 billion yen and about 15 billion dollars. In fact, 15 billion yen is less than 1% of 15 billion dollars.

    Simply Wall St reported on October 2, 2026 that Dell Technologies (NYSE: DELL) has joined an AI data center project in Japan involving JERA, the country’s largest power generator, and a company named RHAELM. The project is designed around securing reliable electricity for power-intensive AI computing. The site is planned to open in 2028 and reach full operation in 2029.

    The project’s size was given as ¥15 billion and also described as roughly $15 billion. Those two figures are far apart: ¥15 billion converts to less than one percent of $15 billion.

    Executive Summary

    Dell has become a named participant in a Japanese AI data center project whose defining feature is its power partner. JERA generates more of Japan’s electricity than any other company. The project is explicitly built around solving the problem that limits AI buildouts in many markets: getting enough dependable power to the building.

    For Dell, the deal is significant less for its dollar value, which has been stated two incompatible ways, than for what it signals about strategy. The analysis accompanying the announcement cites more than US$130 billion in AI server orders and a US$95 billion backlog. A backlog is orders booked but not yet delivered. That backlog turns into revenue only when there are powered, cooled buildings ready to receive the servers. A partnership anchored by a utility-scale generator is one way to secure those destinations.

    What the announcement does not establish is equally important. It does not define Dell’s role, the facility’s power capacity, its customers or how it is financed. Without those details, the deal is best read as a strategic marker rather than a quantified commitment.

    Why Power Access Outweighs the Price Tag

    The useful way to read this deal starts with Dell’s order book, not the project’s headline value. According to the analysis published alongside the news, Dell’s AI growth case depends on converting more than US$130 billion of AI server orders and a US$95 billion backlog into delivered systems. AI servers are racks of graphics processors (GPUs) and supporting hardware. They are installed in data centers that must supply far more electricity per square meter than conventional facilities. Orders cannot ship into a building that does not yet have power, so every delay in grid connections or generation pushes revenue recognition further out.

    Measured against a US$95 billion backlog, a single project worth either ¥15 billion or $15 billion is not what moves Dell’s results. The mechanism that matters is the partner list. Joining a project co-led by Japan’s largest generator gives Dell a line of sight to a site whose electricity supply is part of the design from the start, rather than something to be negotiated after construction. That reduces one source of what the analysis calls execution friction: the gap between building hardware and having somewhere to switch it on.

    The parties affected are concrete. Dell gains a reference project where compute and power are planned together. JERA gains a large, steady electricity customer of the kind utilities increasingly court. Any eventual tenants gain a facility where power availability is less of an open question. None of those benefits depend on whether the project is worth $100 million or $15 billion.

    A Size Figure That Needs Reconciling

    The project was described as ¥15 billion and as roughly $15 billion. At any exchange rate the yen has traded at in decades, those figures differ by more than a factor of one hundred. That is the difference between a modest specialized facility and a campus large enough to rank among the bigger AI builds anywhere.

    This matters for anyone trying to size the opportunity. A $15 billion project would imply a very large server and infrastructure order flowing through partners like Dell over the 2028–2029 ramp. A ¥15 billion project would be a meaningful but contained deployment. Until the participants state the figure in a consistent currency, capacity and revenue estimates built on either number rest on an unresolved premise.

    The 2028–2029 Clock and Dell’s Margin Question

    The site is planned to open in 2028 and reach full operation in 2029. For AI infrastructure, that is a long horizon: server generations turn over roughly yearly, so the hardware eventually installed may differ substantially from what is shipping today. A long lead time cuts both ways for Dell. It provides visibility into future demand, but it also means the project contributes little near-term revenue against the current backlog.

    The accompanying analysis also notes a structural risk the deal does not change: AI GPU servers carry mid-single-digit operating margins for Dell, well below what its traditional storage and services businesses earn. The suggestion that Dell could bundle compute, storage and power reliability into a single offering for hyperscale clients (the largest cloud operators) is one route to better economics. It remains an interpretation, however; neither a bundled product nor a named hyperscale customer has been disclosed. Whether this project improves Dell’s margin mix depends on how much higher-margin storage, networking and services it ultimately includes.

    Does Japan’s Timeline Now Run Through Dell?

    It is tempting to conclude that the project’s schedule now depends on Dell’s rack deliveries. The facts do not yet support that. The announcement does not say whether Dell is the primary server supplier, one of several vendors, an equity participant or a systems integrator. In practice, the critical path for a power-first data center usually runs through generation, grid connection and construction well before servers arrive. The partner most likely to set the pace here is the one supplying electricity. If Dell’s role is later confirmed as lead hardware supplier, its delivery performance would become a factor in the 2029 ramp. For now, that link is a question, not a finding.

    Background

    Dell Technologies, based in Round Rock, Texas, sells PCs, servers, storage and IT services worldwide. Over the past few years its server business has been reshaped by demand for AI systems built around high-end GPUs. That demand has produced a large order backlog but at thinner margins than Dell’s traditional enterprise hardware.

    JERA was created by two of Japan’s major utilities, TEPCO and Chubu Electric, to combine their fuel procurement and thermal power generation, and it is now the country’s largest power producer. Across many markets, AI data center developers increasingly treat electricity supply as the first problem to solve rather than the last. That shift explains why generators and utilities are appearing as named partners in projects that once would have been led solely by real estate developers and hardware vendors.

    Sources

    Source: Dell Technologies (DELL) Joins $15 Billion AI Data Center Project In Japan (Simply Wall St). Dell joins a power-focused AI data center project in Japan with JERA and RHAELM, slated to open in 2028.

  • Amazon’s Reported $8B Nvidia Chip Sale Tests GPUs as an Asset Class Outside Big Tech

    Amazon’s Reported $8B Nvidia Chip Sale Tests GPUs as an Asset Class Outside Big Tech

    TL;DR · 30-second read

    The Short Version

    Amazon is reportedly trying to sell about $8 billion worth of Nvidia computer chips (the specialized processors that power artificial intelligence services) to outside investors.

    That is unusual. Big tech companies normally buy these chips and keep them. Selling them would let Amazon free up cash while someone else owns the hardware.

    The catch is that these chips can lose value quickly as newer models arrive. Investors who buy in are betting the chips stay useful long enough to pay off. That bet could shape how the artificial intelligence boom gets paid for.

    Amazon is seeking to sell roughly $8 billion worth of Nvidia chips to investors, Reuters reported, citing the Financial Times. The chips are graphics processing units (GPUs), the processors that train and run most large AI models. Amazon has not publicly disclosed the structure, the buyers, the pricing or the timing of the transaction, and the effort is described as being sought rather than completed.

    Executive Summary

    The world’s largest cloud provider is reportedly looking to move a large block of AI hardware it already owns into the hands of outside investors. At $8 billion, this would be one of the largest single transfers of GPU ownership from a hyperscaler (the industry term for the handful of giant cloud operators such as Amazon Web Services, Microsoft Azure and Google Cloud) to financial buyers.

    The transaction matters less for what it does to Amazon’s AI capacity, which has not been disclosed, than for what it asks of the buyers. Anyone paying $8 billion for GPUs has to put a price on things the industry has argued about for two years: how fast these chips become obsolete, how reliably they will be rented out, and who is on the hook if they are not. The answer will show whether AI compute can be financed like aircraft or real estate, as an asset owned by one party and used by another, or whether it stays an asset that only the largest technology companies can hold.

    Why a Cloud Giant Would Sell Chips It Already Owns

    Hyperscalers have historically funded data center buildouts themselves, out of operating cash flow and corporate bonds, and owned the servers inside outright. Selling installed or purchased hardware breaks with that pattern. Amazon has not explained its reasons. Transactions of this kind usually serve one of two purposes. The first is capital recycling: convert hardware into cash that can be redeployed into new data centers, power contracts or newer chips. The second is balance-sheet management: shift ownership, and with it the depreciation (the accounting charge that spreads a purchase’s cost over its useful life), to another party.

    Which purpose applies depends on a detail Amazon has not disclosed: whether it keeps using the chips. In a sale-leaseback, a familiar structure in real estate and aviation, the seller hands over ownership and pays rent to keep operating the asset. In an outright sale, the chips leave Amazon’s fleet entirely. The two lead to very different conclusions about Amazon’s own capacity plans, which is why the reported $8 billion figure alone does not show whether Amazon is adding flexibility or reducing exposure.

    What an $8 Billion GPU Buyer Actually Has to Underwrite

    This is where the transaction becomes a test of GPUs as an investable asset. A building or a jet has decades of useful life and a deep resale market. An AI accelerator has neither. Nvidia releases new chip generations on a short cycle, and each one resets the performance customers expect per dollar and per watt. An investor holding older GPUs is exposed to that curve. The hardware may still work perfectly, but the rent it can command falls as newer chips arrive.

    That leaves a buyer two ways to earn a return, and each has a different risk. If Amazon leases the chips back, the investor is mostly underwriting Amazon’s promise to pay. That is a strong credit, and the deal starts to look like corporate lending with GPUs as collateral. If Amazon does not lease them back, the investor is underwriting the chips themselves: their resale value, the cost of hosting and powering them elsewhere, and the strength of demand from other cloud tenants. Pricing that second scenario at $8 billion would be a much bolder bet on GPU residual value, meaning what the hardware is worth after its first few years of use.

    Either way, the price investors will pay is information the market has lacked. Little public data exists on what large blocks of used AI accelerators are worth to financial buyers. A completed transaction at this size would give lenders, insurers and equipment financiers a reference point for valuing GPU collateral elsewhere in the industry.

    Who Gains if GPUs Become a Financed Asset

    If investor appetite proves deep, the most immediate beneficiaries are capital-hungry builders outside Big Tech: specialist GPU cloud providers, often called neoclouds, and data center developers whose growth depends on borrowing against hardware. A market in which a hyperscaler can sell GPUs to investors is also one in which a smaller operator can borrow against them more easily. For hyperscalers, it adds a funding channel at a time when AI capital spending is large enough to draw scrutiny from shareholders.

    The risks fall on the other side of the ledger. Moving hardware ownership to investors does not remove the obsolescence risk. It relocates it, possibly into funds and credit vehicles that are less able to absorb a fast drop in chip values than a cloud operator that can redeploy older chips to cheaper workloads. For Nvidia, the transaction involves chips already sold and does not by itself change its revenue. A deeper financing market could still widen the pool of buyers able to afford its next generation. That makes the terms of any completed deal more informative than the headline number.

    Background

    Amazon Web Services is Amazon’s cloud computing division and one of the largest buyers of AI hardware in the world. It purchases Nvidia GPUs and also designs its own AI chips, including Trainium. Nvidia is the leading supplier of the accelerators used to train and run large AI models, and it releases new chip generations on a short cycle.

    Hyperscalers have traditionally financed data centers and servers themselves and owned that hardware outright. As AI spending has grown, outside capital has moved into the sector. Specialist GPU cloud providers routinely borrow against their chips, and investors have debated how quickly AI accelerators lose value. The reported Amazon transaction brings that financing model to the scale of the largest cloud operator.

    Sources

    Source: Amazon seeks to offload $8 billion of Nvidia chips to investors, FT reports (Reuters), on Amazon’s reported effort to sell roughly $8 billion of Nvidia GPUs to investors.