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

Author: Deepak Jain

  • Wand AI Adds Zeroport to Put Agents on Air-Gapped Systems

    Wand AI Adds Zeroport to Put Agents on Air-Gapped Systems

    TL;DR · 30-second read

    The Short Version

    Artificial intelligence programs are starting to touch the systems that actually run a country: power grids, water utilities, defence networks, government records. Those systems are deliberately kept off the internet so nothing outside can reach them.

    Wand AI and Zeroport say they have a way to let automated software workers operate them anyway. The software gets no connection at all. It sends keystrokes down a wire built so that data physically cannot travel back, and receives only a moving picture of the screen, like looking through a window.

    A separate computer watches every frame and can cut the session off mid-task.

    Wand AI said in a September 9 announcement distributed over PR Newswire that it has added Zeroport to its Sovereign AI offering, incorporating Zeroport’s Fantom hardware boundary and its Moativ enforcement layer into Wand’s agent platform. The stated goal is to let government programs point autonomous AI agents at segmented and air-gapped estates, including grid, water, SCADA, defence, banking and registry systems, without creating a network route into them.

    Under the joint reference architecture, Wand’s runtime dispatches an agent session to a dedicated Fantom Core rather than to an endpoint inside the protected network. Only keystroke and mouse input travels inward across a one-way hardware break, and only a display-only pixel stream travels outward. Moativ, an on-device model running on NVIDIA Jetson Thor accelerators inside the same appliance, inspects every input and every frame and can alert, block or physically sever the session. The companies say the architecture is available to sovereign programs and enterprises now, and that adopting Zeroport is elective rather than required.

    Executive Summary

    The announcement addresses a real and specific bottleneck. National AI programs are moving past document summarisation toward operating systems of record and systems of control, and those are precisely the estates that security teams spent a decade making unreachable. Connecting an agent to them the conventional way means provisioning an interface, a credential and a network path, which recreates the exposure the segmentation was designed to eliminate. Zeroport’s argument is that the route itself should be removed rather than narrowed.

    The technical claim has two halves, and they deserve separate treatment. The first is that the access path is physical: a partition built from the chip up that carries no IP packet and terminates no network session, so an agent reaches the protected system the way a cleared human operator does, keystrokes in and pixels out. The second is that the policy governing the agent lives out of band, on hardware the governed session has no address for, which answers a genuine weakness in software guardrails that run on the same computational plane as the model they police.

    What the companies have described is an architecture and a reference design, not a deployment record. No sovereign customer, contract value, pilot result or third-party certification accompanies the announcement, and the performance figures are stated as targets and capacities rather than measured production results. The architectural idea is coherent and the problem is well framed; the evidence that it works at national scale is still forthcoming.

    The Last Mile of Sovereign AI Is an Access Problem, Not a Model Problem

    Sovereign AI has largely been discussed as a question of ownership: whose chips, whose weights, whose data centre, whose law. That framing gets a country to a capable model sitting on domestic compute. It does not get the model anywhere near the substation control interface, the water treatment historian, or the case management system where the work actually is. Air-gapping means a network is physically or logically isolated with no routable connection to the outside, and organisations do it precisely because a path into those estates is a national risk before it is an IT risk.

    Wand’s positioning treats that last mile as the binding constraint on the value of AI labour, and the logic holds. An agent restricted to systems already reachable through ordinary enterprise controls is confined to the least consequential half of a ministry’s estate. The alternatives on offer today all involve a trade the release characterises fairly: virtual private networks, zero-trust network access and privileged access management gateways narrow and log the route but preserve it, so it remains enumerable and exploitable; guardrail software asserts policy on the same plane as the model, where a sufficiently capable model can be argued out of it; and leaving the estate offline removes the route by removing the workload.

    Zeroport cites roughly two-thirds of cyber-attacks as exploiting remote access tooling and 80 percent of breaches as involving stolen or compromised credentials. Neither figure is sourced in the announcement, and both are the kind of round number that circulates widely in vendor material, so they should be read as framing rather than evidence. The underlying point survives the caveat regardless: remote access is the dominant vector in most credible threat models, and an agentic layer multiplies the number of sessions requesting it.

    Physics Governs the Route; Software Still Governs the Judgment

    The strongest part of this architecture is also the most conventional. A one-way hardware break that admits only human-interface signals and emits only pixels is a descendant of the unidirectional gateway and data diode designs that have protected operational technology perimeters for years, combined with the display-only principle behind remote desktop and browser isolation. Applying that lineage to agent sessions is a sensible extension rather than an invention, and it inherits the property that matters: there is no packet to smuggle a payload inside, and nothing for a compromised or drifting agent to pivot into. Revoking an agent removes the path outright rather than closing a session on a path that persists.

    The second claim needs more care. Moativ enforces policy from silicon the governed session cannot address, which genuinely defeats the prompt-injection and goal-drift failure modes where an agent talks its way past a rule that lives in its own context. But Moativ is itself a model making content-level judgments about what is on screen and what was typed. Physics secures the channel; a classifier decides intent. That classifier can produce false negatives, in which a harmful action reads as sanctioned, and false positives, in which legitimate work is severed. On a control system, a wrongly terminated session mid-action is not a neutral outcome.

    The controls around that risk are described more concretely than most vendors manage. Rules are written in plain language and compiled into staged detectors, dry-run against a program’s own session history before they are armed, with autonomy set per group on a five-step dial running from observe through alert, investigate and terminate to self-tune. Dry-running policy against real historical sessions is the right way to calibrate an enforcement model, and the audit trail attaches screen content, keystrokes and rationale to every autonomous action. What is absent is any published figure for detection accuracy or intervention rate.

    Pixels In, Pixels Out: What the Agent Gives Up

    Routing an agent through a screen rather than an application programming interface is a security gain purchased with capability. Agents that drive graphical interfaces are measurably less reliable than agents calling structured interfaces, because they must locate controls visually, tolerate rendering variation, and recover from states no schema describes. Every task must be expressible as keystrokes and mouse movement against a live display, and nothing structured comes back, only pixels. For supervisory control work performed by a human operator today, that is a natural fit. For bulk data extraction, batch reconciliation or anything requiring high-volume structured output, it is a poor one, and it is worth noting that the pixel-only return path is exactly what makes exfiltration hard.

    The economics scale in a way network overlays do not. Capacity grows in Cores and then in appliances, with hundreds of concurrent sessions per appliance and thousands per rack, and humans and agents riding the same rails with no shared tunnel to bottleneck. That converts a licensing question into a hardware procurement question, which is familiar territory for defence and utility buyers and appears in a capital budget rather than an operating one. The dedicated pathways target sub-50-millisecond latency, stated as a design target, which is the threshold that separates a channel usable for live control from one fit only for after-the-fact review. All inspection happens on-premises and no session content leaves the rack, a claim that matters more to sovereign buyers than almost anything else in the architecture.

    An Ecosystem Assembled Layer by Layer

    This is the third such addition Wand has announced in the same pattern, following a storage efficiency layer and a model robustness layer, and alongside existing inference privacy and confidential computing capabilities. The construction is deliberate: a sovereign platform assembled from independently adoptable components, none a prerequisite for another, each addressing a distinct objection a national buyer might raise. It is a credible route to breadth for a company founded in 2023, and it lets Wand answer a procurement checklist without building every layer itself.

    It also leaves the commercial substance of each addition undefined. The announcement describes a joint reference architecture and mutual technical fit, but not whether Zeroport hardware is resold by Wand, jointly supported, revenue-shared, or merely certified as compatible. Buyers evaluating an integration care about which entity holds the support contract when an appliance severs a session on a live grid system. It is also worth recording plainly that Fusion Fund appears as an investor in both companies, which is unremarkable in a specialised market but relevant context for how ecosystem partnerships form.

    For Zeroport, founded in 2024 in Herzliya by security and military-intelligence veterans, the partnership offers distribution into national programs that a two-year-old hardware company would otherwise reach slowly. For Wand, whose offices include Abu Dhabi and whose stated customer base spans banks, asset managers and system integrators, it supplies an answer to the question sovereign buyers ask last and hardest: what happens when the agent has to touch the thing that matters.

    Background

    Sovereign AI describes a country’s effort to run artificial intelligence on domestic compute, under domestic law, with control over the models, data and policies involved. The first wave of national programs concentrated on procuring accelerators and building data centres. The second, now underway, tries to convert that capacity into operational work inside ministries, utilities and defence estates, which raises a problem the first wave never had to solve: the highest-value systems are the ones deliberately made unreachable.

    Wand AI, founded in 2023 and headquartered in Palo Alto with offices in New York and Abu Dhabi, sells an operating system for deploying and governing AI agents alongside humans, with identity, authority, audit and governance built in. It is backed by Shasta Ventures, Fusion Fund, Thiel Capital and Palo Alto Growth Capital, and has been assembling its sovereign offering as a series of independently adoptable layers covering inference privacy, confidential computing, model robustness and storage efficiency. Zeroport, founded in 2024 in Herzliya, Israel, holds multiple patents on its Fantom platform and says it is deployed across critical infrastructure, energy, defence, financial services and government environments, with backing from lool ventures, Clarim Ventures, CyberFuture and Fusion Fund.

    Sources

    Source: WAND AI adds ZEROPORT as an Agent Containment Layer to Enable Sovereign AI on Air-Gapped and Critical National Systems — the September 9, 2026 announcement describing the joint reference architecture, the Fantom hardware boundary, the Moativ on-device enforcement layer, and the two deployment points the companies support.

  • ABB and BCG Make the Case for DC Power in AI Data Centers

    ABB and BCG Make the Case for DC Power in AI Data Centers

    TL;DR · 30-second read

    The Short Version

    • The Swiss engineering group ABB and the consultancy Boston Consulting Group published a report saying the electricity inside data centers should increasingly flow the way it does in a battery, not the way it does in a wall socket.
    • Computers, solar panels and batteries all run on that kind of power already. Every time electricity is converted from one form to the other, a little is lost as heat, like a phone charger warming up.
    • Artificial intelligence machines use so much power that those small losses now add up to real money and real capacity.
    • ABB sells the equipment, so it has a stake in the argument.

    ABB said on September 9 that it has published The Strategic Case for Hybrid AC/DC Power: Shaping the Transition to the Next Electrical Architecture, a report developed with Boston Consulting Group. The Zurich-headquartered electrification and automation group argues that direct current, or DC, is moving from the periphery to the center of industrial, commercial and digital infrastructure, and that choices made by executives and policymakers over the next two to three years will decide who shapes the transition and who simply inherits it.

    The report singles out AI data centers as one of the most immediate drivers of DC adoption, finding that 800-volt DC distribution is emerging as the defining architecture for next-generation AI infrastructure. It frames the outcome not as a contest between DC and alternating current, or AC, but as a hybrid system: AC remains the backbone of transmission and regional distribution, while DC spreads inside facilities where the major sources and loads already run natively on it.

    Executive Summary

    The argument rests on a piece of electrical plumbing that most people never see. Grids deliver alternating current, in which the flow reverses direction many times a second, because that form is cheap to push over long distances and easy to step up and down in voltage with transformers. Almost everything now growing fastest, though, is natively direct current: solar photovoltaic panels, batteries, electric vehicles, industrial automation gear and the graphics processors inside AI servers. Each handoff between the two forms runs through a converter, and each converter throws off a percentage of the energy as heat.

    ABB and BCG contend that stripping conversion steps out of the chain does three things at once: it raises efficiency, it increases the usable compute capacity behind an existing grid connection, and it simplifies the integration of on-site generation and storage. That middle point is the commercially interesting one. In markets where new grid capacity is the scarce input, an efficiency gain converts directly into more revenue-generating hardware behind the same meter.

    The report also reframes the obstacle. Concerns about DC at scale, its safety and its economics reflect outdated assumptions rather than the capabilities of current equipment, ABB and BCG conclude. What actually slows deployment, they say, is fragmented standards and a shortage of engineers and technicians trained on DC systems. That is a claim about institutions and labor markets rather than about physics, and it points to a slower, less predictable fix.

    Why AI Racks Are Rewriting the Electrical Diagram

    Conventional data center power design has been remarkably stable for two decades. Utility alternating current arrives at a substation, passes through switchgear, an uninterruptible power supply and a power distribution unit, then reaches a server whose internal supply finally converts it to the low-voltage direct current the chips actually consume. Each stage was tolerable when a rack drew a few kilowatts. The report’s contention is that at AI power densities, conventional architectures are reaching practical limits, and the accumulated conversion overhead stops being a rounding error.

    The 800-volt DC figure is the specific technical claim worth understanding. Raising the distribution voltage lets a given amount of power travel through thinner conductors, because power is voltage multiplied by current and it is current that dictates copper cross-section and resistive heating. Delivering high-density racks at low voltage means either enormous busbars or unacceptable losses. Pushing distribution to 800 volts DC and converting close to the load is the same trick the automotive industry adopted for fast-charging electric vehicles, applied to a data hall.

    The practical consequence is that the change does not stop at the rack. It touches switchgear, protection, uninterruptible power supply topology, on-site storage and the interface with the substation. That is a supply-chain shift as much as an engineering one, and it favors vendors who already hold the protection and conversion portfolios.

    The Real Prize Is the Grid Connection

    Read commercially, the strongest line in the release is not about efficiency in the abstract. It is that DC distribution lets operators maximize compute capacity within constrained grid connections. In the markets where AI capacity is being built hardest, the binding constraint is rarely capital or land. It is the megawatts a utility will actually deliver, and when they will deliver them.

    Under that constraint, the economics change character. An efficiency improvement in a power-abundant market is a cost saving measured against the electricity bill. The same improvement behind a fixed, fully subscribed grid connection is a capacity unlock, measured against the revenue of the additional accelerators it lets you energize. Those two framings can differ by an order of magnitude in value, which explains why an argument about conversion losses is landing now rather than a decade ago.

    It also explains the report’s two-to-three-year window. Electrical architecture is set early in a data center’s design and is expensive to revisit. Operators committing to buildings that will be energized late this decade are choosing their topology now, which is precisely when a vendor wants the conversation to happen.

    Standards and Skills Are the Stated Bottleneck

    ABB and BCG place the constraint on adoption in two unglamorous places: fragmented standards and a shortage of DC-specific skills. Both are credible. Direct current has no natural zero crossing, the instant of zero flow that lets a conventional breaker extinguish an arc, which is why DC protection historically demanded different hardware and different design practice. ABB says it launched the industry’s first solid state circuit breaker in 2022, and holds more than 700 DC-related patents.

    That patent position is also why the call for harmonized standards deserves to be read with clear eyes. Common standards genuinely enlarge a market and reduce buyer risk, and buyers should want them. They also tend to reward whoever arrives with the deepest existing portfolio in the standardized technology. Nothing about that is improper, and ABB is not alone in it, but a customer evaluating this report should treat it as a well-argued position paper from an interested party rather than as neutral market research.

    The skills point may be the more stubborn of the two. Standards can be negotiated on a schedule. A workforce of electricians, commissioning engineers and facility staff fluent in high-voltage DC protection is built over years, and its scarcity would constrain even operators fully convinced by the argument.

    What Is Substantiated and What Is Asserted

    The quantified evidence in the announcement is drawn from marine propulsion, where ABB says a DC-based onboard vessel power system achieved fuel savings of up to 27 percent. That is a real, specific result, and it is not a data center result. Ships run isolated power systems with variable-speed loads and their own economics; the number establishes that ABB has long-running DC field experience, not that a data hall should expect a comparable gain.

    For the data center case itself, the announcement offers direction rather than magnitude. It says 800-volt DC is emerging as the defining architecture and that efficiency and usable capacity improve, without published percentages, capital-cost comparisons or named deployments in the summary. The conclusion that scale, safety and economic objections are now outdated is likewise stated as a finding, with the supporting analysis sitting in the full report published on ABB’s site.

    None of that undercuts the thesis, which aligns with where high-density computing and on-site storage are visibly heading. It does mean the announcement is best treated as a framing document. The buyer’s question is not whether DC distribution helps, but by how much, at what installed cost premium, and with which protection and service ecosystem behind it.

    Background

    The choice between alternating and direct current was settled commercially in the late nineteenth century, when transformers made it cheap to step alternating voltage up for long-distance transmission and back down for use. Direct current never disappeared, though. It survived in specialist niches such as long submarine transmission links, traction systems and shipboard power, and it returned quietly through the back door as solar panels, batteries and electronics proliferated, each of them a direct current device tethered to an alternating current grid through a converter.

    ABB sits on both sides of that history as one of the largest electrification and automation suppliers globally, selling switchgear, drives, protection and power distribution equipment into utilities, industry and data centers. The AI buildout has made its data center segment strategically prominent, because the constraint on new compute has shifted from chips and capital toward electrical delivery. This report, produced with Boston Consulting Group, is the company’s argument that the resulting architecture will look meaningfully different from the one the industry has built for the past twenty years.

    Sources

    Source: New ABB and BCG report highlights growing role of direct current technologies in future power systems — ABB’s September 9, 2026 announcement of a report developed with Boston Consulting Group on the shift toward hybrid AC/DC electrical architectures, including 800 VDC distribution for AI data centers. The full report is published at abb.com.

  • Centrus Signs HALEU Deal to Fuel Radiant’s Kaleidos Microreactors

    Centrus Signs HALEU Deal to Fuel Radiant’s Kaleidos Microreactors

    TL;DR · 30-second read

    The Short Version

    Two American companies have agreed to build the fuel supply for a new kind of nuclear reactor, one small enough to be trucked, shipped or flown to wherever power is needed.

    Centrus Energy will make the specially enriched uranium that Radiant’s Kaleidos reactors run on. Deliveries start before 2030, and Radiant is paying money up front to help Centrus expand its enrichment plants.

    It matters because these reactors are aimed at remote sites, military bases and the power-hungry computer warehouses behind artificial intelligence.

    Centrus Energy Corp. and Radiant announced on September 9 a definitive multi-year contract under which Centrus will supply high-assay, low-enriched uranium, or HALEU, for multiple units of Radiant’s Kaleidos transportable microreactor. Centrus said deliveries will begin before the end of the decade, and that the agreement includes prepayments from Radiant that support the build-out of Centrus’s domestic commercial enrichment capacity.

    Because Centrus enriches with U.S.-origin technology and a U.S. manufacturing supply chain, the companies said the material supplied to Radiant will be unobligated, meaning it carries no foreign peaceful-use restrictions and can be used for national security applications as well as commercial ones. Neither company disclosed the contract value, the quantity of fuel involved, or the number of reactors it covers.

    Executive Summary

    The agreement extends a pattern Centrus has been building for two years: converting its position as the sole domestic producer of HALEU under a U.S. Department of Energy program into commercial offtake from advanced reactor developers. It follows a separate LEU and HALEU supply agreement with X-energy and adds Radiant to a backlog that Centrus uses to justify expansion at its Piketon, Ohio, and Oak Ridge, Tennessee, sites.

    For Radiant, the logic runs in the opposite direction. Advanced reactor developers have repeatedly found that fuel, not engineering, is the binding constraint on deployment. Kaleidos and most other microreactor and advanced designs require uranium enriched to between roughly 5 and 20 percent uranium-235, well above the 3 to 5 percent that fuels the existing light-water fleet, and until recently there was no meaningful Western commercial supply. Chief Nuclear Officer Rita Baranwal framed the deal as removing one of the biggest constraints facing advanced nuclear deployment, and said Radiant is deliberately pursuing more than one fuel path.

    The structure is the most economically interesting part. Prepayments mean a reactor developer is helping to finance an enrichment plant before that plant has produced fuel for it. That is customer capital substituting for project finance in a sector where lenders have historically been reluctant, and it signals conviction on both sides. It also concentrates risk: if either the reactor or the enrichment expansion slips, both parties are exposed.

    The Fuel Bottleneck Moves Upstream

    For most of the past decade, the advanced nuclear conversation was about reactors. Designs proliferated, regulators built new review pathways, and developers raised capital on the promise of factory-built units. The quiet problem was that almost none of those designs could run on the fuel that commercial enrichment plants actually make. HALEU sat in a gap between civilian fuel and material of proliferation concern, and the only substantial supplier was Russia.

    What the Centrus and Radiant contract represents is the bottleneck moving from availability to schedule. The question is no longer whether domestic HALEU will exist, but whether enough of it will exist, at the right time, at a price that lets a microreactor compete with a gas turbine or a grid interconnection. Centrus committing to deliveries before the end of the decade is a real date, though a soft one. It leaves several years of execution risk in a business where centrifuge cascades, licensing and security infrastructure all have long lead times.

    The prepayment mechanism also changes the incentive structure across the sector. If reactor developers can be induced to fund enrichment capacity ahead of demand, the classic chicken-and-egg standoff, where no one builds a plant without customers and no one signs customers without a plant, becomes solvable in stages. It is a model worth watching, because it implies that the credible reactor developers are those who can put capital into their own supply chain.

    Unobligated Fuel Is the Commercial Differentiator

    The most consequential technical claim in the announcement is the least discussed publicly. Uranium enriched using technology derived from foreign sources, or processed under international nuclear cooperation agreements, carries obligations restricting it to peaceful use. Material produced with wholly U.S.-origin technology does not. That distinction determines whether fuel can go into a reactor powering a military installation or a defense-adjacent facility.

    Centrus states that its AC100 centrifuge design is the only deployment-ready U.S.-origin technology available for unobligated enrichment today. That is a company characterization of a competitive landscape rather than an independently verified fact, and rivals developing domestic enrichment would be expected to contest the framing over time. It is nonetheless a meaningful position while it holds. Radiant explicitly cites commercial and national security applications for Kaleidos, and defense customers are the segment least sensitive to price and most sensitive to sourcing provenance.

    For Centrus, the strategic value is that unobligated capacity is difficult to replicate quickly. Enrichment technology is export-controlled, capital-intensive and slow to license. A supplier with a durable claim to the defense-eligible segment of the market can price differently there than in the contested commercial market, where the release’s own risk disclosures acknowledge significant competition from foreign producers who may be less cost sensitive.

    What Data Center Operators Should and Should Not Read Into This

    Radiant’s stated markets include data centers, and that is why this announcement matters beyond the nuclear trade press. Transportable microreactors are attractive to operators facing multi-year interconnection queues, the waiting lists utilities maintain for connecting large new loads to the grid, and to those building in locations where transmission simply does not exist.

    The honest read is that this is a fuel contract, not a power contract. Nothing announced here commits a reactor to a site, secures a licence for one, or names a data center customer. Radiant is described as moving from its first Kaleidos test toward commercial deployment, which places the technology before commercial operation rather than after it. Operators making 2027 and 2028 capacity decisions cannot plan around Kaleidos, and should treat microreactors as a possible answer to the early 2030s question of how to power sites that the grid cannot reach in time.

    What operators can reasonably take from it is directional. Fuel supply has been the standard objection to every microreactor pitch, and each contract of this type weakens that objection a little. The remaining objections, regulatory approval, fuel fabrication, unit economics at low output and the practical politics of siting a reactor next to a commercial building, are unaffected by this agreement.

    The Backlog Story and Its Limits

    Centrus is presenting the deal to investors as backlog growth, and backlog is the metric the equity story rests on. The company has now layered a microreactor developer on top of prior commitments including its agreement with X-energy, broadening its customer base beyond the government HALEU program that started it.

    Backlog quality, however, depends on terms that were not disclosed. A multi-year supply contract can be firm take-or-pay volume or a framework contingent on the customer’s own milestones. Prepayments suggest genuine commitment, but their size relative to the expansion programs at Piketon and Oak Ridge determines whether they are a financing event or a goodwill deposit. Centrus’s own forward-looking statements flag dependence on U.S. government appropriated funding for HALEU, its Piketon lease and successful execution of the planned expansions, which is a fair summary of what has to go right.

    The counterparty risk runs both ways and is worth stating plainly. Centrus is taking on a customer whose reactor is not yet commercially deployed, and Radiant is relying on an enrichment expansion that is not yet complete. That is not a criticism of either party so much as a description of what building a new industrial supply chain looks like. It does mean the announcement should be read as an intention backed by capital, not as fuel in a cask.

    Background

    Centrus Energy Corp., listed on the New York Stock Exchange as LEU and headquartered in Bethesda, Maryland, is the successor to the United States government’s commercial uranium enrichment enterprise. The company says it has supplied more than 1,850 reactor years of fuel to utility customers since 1998, which it equates to more than 7 billion tons of coal. In recent years it has pivoted from a trading and services business toward restoring domestic enrichment capability, producing HALEU at Piketon, Ohio, under a Department of Energy program and pursuing commercial expansion there and at Oak Ridge, Tennessee.

    Radiant develops transportable nuclear microreactors under the Kaleidos name, aimed at commercial, industrial and defense customers who need power delivered to a location rather than drawn from a grid. Its market exists because two constraints converged: rising electricity demand from artificial intelligence computing and industrial electrification, and the long queues and transmission limits that slow new grid connections. The sector’s persistent obstacle has been fuel. Advanced designs need HALEU, Western commercial supply barely existed until recently, and the material was long sourced primarily from Russia, making domestic enrichment a national security priority as well as a commercial one.

    Sources

    Source: Centrus and Radiant Announce Long-Term Partnership for Domestic HALEU to Supply Kaleidos Microreactors, the September 9, 2026 announcement from Centrus Energy Corp. detailing the multi-year HALEU supply contract, the prepayment structure and the unobligated enrichment provision.

  • ERCOT Batch Zero Clears 4.2 GW of Galaxy Digital Load

    ERCOT Batch Zero Clears 4.2 GW of Galaxy Digital Load

    TL;DR · 30-second read

    The Short Version

    Galaxy Digital, a company that builds warehouses of computers for artificial intelligence, has cleared an important hurdle in Texas.

    The state’s electricity grid operator conditionally agreed to let five of its sites plug in and draw an enormous amount of power, roughly what a large city consumes. Two sites got the firmer approval. Three still need more study.

    Why it matters: in Texas, the grid now decides who gets that much electricity and who waits. That makes the power company, not the builder, the one setting the pace.

    Galaxy Digital has secured conditional Batch Zero large load classifications from the Electric Reliability Council of Texas, the grid operator that runs most of the state’s power system, for five Texas data center projects covering roughly 4.2 gigawatts of gross power capacity. TipRanks reported that the decisions advance the company’s Helios I and Helios II campuses as Base Load, while three expansion sites were designated Studied Load, and that they lift Galaxy’s development pipeline to more than 5.7 gigawatts across four Texas sites.

    The classifications are conditional rather than final. Some projects remain subject to further capacity allocation and audit, and the decisions were granted amid heightened regulatory scrutiny of very large electricity users by Texas authorities. Galaxy also has a planned campus called Merlin in McGregor, Texas, in the pipeline.

    Executive Summary

    What Galaxy Digital announced is not a customer, a construction start or a financing. It is a place in line. Batch Zero is the first tranche of requests processed under the interconnection regime Texas built for outsized electricity users, and a classification determines whether a proposed campus is treated as a load the grid is prepared to serve or as one that still has to be studied before capacity is set aside for it.

    That distinction is the whole story. Galaxy’s two Helios campuses were classified as Base Load, the firmer designation. The three expansion sites were classified as Studied Load, which by the company’s own description leaves them exposed to further capacity allocation and audit. Roughly 4.2 gigawatts sits inside those five classifications; the pipeline figure Galaxy is promoting, more than 5.7 gigawatts, sits outside them.

    For the wider market, the announcement is a useful demonstration of how power procurement has replaced land, capital and even chip supply as the gating item for artificial intelligence infrastructure in Texas. A developer can own the site, fund the shell and sign a tenant, and still be unable to energize anything until the grid operator says how many megawatts it will actually deliver and when.

    The Interconnection Queue Is Now the Real Permit

    For most of the past decade, a data center developer’s hardest document was a zoning approval or a fiber agreement. In Texas today it is a grid classification. A gigawatt is a thousand megawatts, and a megawatt is roughly the continuous draw of several hundred homes; campuses of this size arrive on the system as the equivalent of new cities appearing overnight, which is precisely why Texas legislators moved in 2025 to give the grid operator explicit authority over how the largest new loads are studied, sequenced and, in emergencies, curtailed.

    Batch Zero is the first cohort run through that machinery, and Galaxy’s result shows what the process actually produces: not a yes or a no, but a tier. Base Load and Studied Load are not synonyms with a paperwork difference between them. One reflects capacity the operator is prepared to plan around. The other reflects a request that has been acknowledged and is still being evaluated. Galaxy’s announcement puts Helios I and II in the first category and its three expansion sites in the second, and states plainly that further capacity allocation and audit remain ahead.

    The competitive implication is that queue position is becoming a genuine asset. Developers who filed early, on sites with existing electrical infrastructure, hold something that latecomers cannot buy with capital alone. Whether that advantage is durable depends on how fast the operator works through subsequent batches, which nobody outside the process can currently forecast.

    Gross Gigawatts Are Not Leased Megawatts

    Two numbers dominate the headlines, 4.2 gigawatts and 5.7 gigawatts, and neither describes computing capacity that exists or revenue that has been contracted. The 4.2 figure is gross power capacity, measured at the point where the campus meets the grid. Real facilities spend a meaningful share of that on cooling, power conversion and losses before any of it reaches a server, so the critical load available to tenants is always lower than the gross figure, and how much lower depends on design choices the announcement does not describe.

    The 5.7 gigawatt pipeline number is broader still. It spans four Texas sites and includes projects that have not received Batch Zero classifications at all. Pipeline figures are legitimate planning disclosures, and every developer in this market publishes one, but they aggregate wildly different degrees of certainty into a single headline. A reader comparing developers on pipeline size is comparing intentions, not assets.

    The arithmetic itself invites a question: five projects received classifications, the pipeline is described as spanning four sites, and one campus, Merlin in McGregor, appears in the announcement without a stated classification or capacity. That is not necessarily inconsistent, since a single site can host multiple projects, but it means the relationship between the two headline numbers is not self-explaining.

    A Crypto Balance Sheet Funding a Utility-Scale Buildout

    Galaxy is a digital asset firm that also develops data centers, and the two businesses have very different capital rhythms. Trading, asset management and custody generate returns on a cyclical, mark-to-market basis. Gigawatt-scale campuses consume billions of dollars over multi-year horizons before producing a dollar of lease revenue, and they are typically financed with project debt underwritten against signed, investment-grade tenant contracts.

    The market has not fully resolved how to price that combination. The company carries a market capitalization of about ten billion dollars, and the most recent analyst rating cited in the announcement is a Hold with a twenty-six dollar price target, a stance that reads as neither endorsement nor rejection of the buildout thesis. That is a reasonable posture given what has and has not been established: the regulatory milestone is real and verifiable, while the conversion of classified megawatts into contracted, cash-generating capacity is not yet demonstrated by this announcement.

    The strategic logic is nonetheless coherent. Sites originally developed for bitcoin mining tend to come with substantial interconnection already in place, which is exactly the scarce ingredient artificial intelligence tenants need and cannot manufacture. Galaxy’s ambition, as it describes it, is to fold power costs into hosting and lease arrangements, which would make it a seller of energy-inclusive capacity rather than a landlord of empty shells. That is a defensible model. It is also one that depends entirely on securing power at prices that stay below what tenants will pay for it, over lease terms measured in years.

    Background

    Texas has become the center of gravity for power-hungry computing in the United States, first for bitcoin mining and more recently for artificial intelligence. Its grid, operated independently of the two large interconnections that cover most of the country, offers abundant wind and solar generation, comparatively fast interconnection historically, and a market structure that rewards flexible consumers. That combination drew mining operators to remote West Texas sites, many of which were later positioned for conversion to artificial intelligence workloads because they already held the scarce asset: a completed connection to high-voltage transmission.

    Galaxy Digital followed that path. Its Helios campus began as a mining facility and has been redeveloped as a multi-tenant site for artificial intelligence and high-performance computing, with Galaxy having previously announced long-term lease agreements with an artificial intelligence cloud provider there. The volume of such requests eventually forced a policy response: in 2025 Texas legislators gave the grid operator new authority to study, sequence and, in grid emergencies, curtail the largest new loads. Batch Zero is the first application of that framework, and it is why an interconnection classification now carries the weight a construction permit once did.

    Sources

    Source: Galaxy Digital Wins Key ERCOT Approvals, Expands Texas Data Center Pipeline to 5.7 GW — report that Galaxy Digital secured conditional ERCOT Batch Zero large load classifications for five Texas data center projects covering about 4.2 gigawatts, lifting its development pipeline to more than 5.7 gigawatts across four sites.

  • Corning’s Verizon Fiber Deal Turns Glass Into an AI Asset

    Corning’s Verizon Fiber Deal Turns Glass Into an AI Asset

    TL;DR · 30-second read

    The Short Version

    • Corning makes the hair-thin glass strands that carry internet traffic around the world. Verizon has agreed to buy billions of dollars of them.
    • Corning’s share price rose about 7 percent on the news.
    • Lumentum, which makes the lasers that push light down those strands, rose about 11 percent, even though it is not part of the deal.
    • Why it matters: the artificial intelligence building boom is soaking up the world’s supply of optical glass, so phone and internet companies now have to line up years ahead to get theirs.

    Seeking Alpha reported that Corning shares rose roughly 7% on a multibillion-dollar optical fiber supply agreement with Verizon, and a report from finance.biggo.com put the same-day gain in Lumentum Holdings at about 11% as the news pulled up the wider optical complex. Corning supplies optical fiber, cable and connectivity hardware; Verizon is one of the largest buyers of that equipment in the United States.

    Corning’s own most recent current report with the Securities and Exchange Commission, an 8-K filed September 4, does not describe the agreement. It discloses instead that Chief Financial Officer Edward A. Schlesinger will speak at the Citi 2026 Global TMT Conference on September 9 at 8:10 a.m. Eastern, and that he expects to discuss recent business performance, progress on the company’s Springboard plan, and market trends and demand drivers across its businesses. That appearance is management’s first scheduled opportunity to put numbers and duration around the Verizon relationship in front of investors.

    Executive Summary

    A long-dated fiber purchase agreement between a glass maker and a carrier would once have been a routine procurement story. It is not being read that way now. Optical fiber has become a contested input, and the reason is the enormous volume of cable being consumed inside artificial intelligence data centers, where thousands of accelerators are stitched together with optical links rather than copper. Every meter of preform, draw tower time and cabling capacity committed to that market is capacity not available to a telephone company building fiber to homes.

    Against that backdrop, a multiyear commitment from Verizon does two things for Corning. It locks in volume against a fixed manufacturing base, which is what makes capacity expansion financeable. And it signals to the market that carriers no longer assume fiber will be there when they want it at the price they expect, which is a change in bargaining posture from the last decade.

    The more revealing market reaction was Lumentum’s. Lumentum does not sell the glass Verizon is buying. Its move reflects investors treating optical capacity of any kind as a single AI infrastructure exposure, which is a thesis about the sector rather than a fact about the contract.

    Glass Became a Bottleneck Before Anyone Called It One

    Optical fiber starts as a glass preform, a cylinder of ultra-pure silica that is heated and drawn into a strand thinner than a human hair, then coated, stranded and jacketed into cable. The process is capital-intensive and slow to scale: draw towers are purpose-built, the plants are large, and adding capacity is a multiyear commitment rather than a quarterly adjustment. Supply is therefore inelastic in exactly the situations where demand spikes.

    Artificial intelligence training clusters are one of those situations. Interconnecting large numbers of accelerators requires far more optical links per rack than a traditional cloud deployment, and the cabling runs inside and between buildings on a campus consume fiber in quantities that used to belong to metropolitan network projects. That demand lands on the same industrial base that serves carriers. A hyperscale operator willing to sign for volume years out competes directly with a telephone company doing the same, and both are bidding for the output of a small number of qualified suppliers.

    A large carrier agreement is the visible consequence. When supply is comfortable, buyers keep purchasing flexible and short. When it is not, they trade flexibility for certainty and sign long. That trade is the actual news here, more than any single dollar figure.

    Why Verizon Signs Long, and What Corning Gets

    Verizon has spent years pushing fiber deeper into its network, both to reach homes directly and to feed the wireless sites that carry mobile and fixed-wireless traffic. Its move to combine with Frontier Communications extends that footprint substantially, and a larger footprint means a larger, more predictable annual appetite for cable. Carriers with build targets and regulatory or competitive deadlines cannot afford a supply gap in year three of a five-year program.

    For Corning, the value of a committed multiyear volume is not only revenue. It is the planning input that justifies keeping capacity hot and, if needed, adding to it. Optical Communications is Corning’s largest business, and the company has framed its growth ambitions publicly under the Springboard plan that its chief financial officer is scheduled to discuss at the Citi conference. A named anchor customer with a long horizon is the kind of evidence that plan needs.

    The relationship is not new. Corning and Verizon announced a multiyear purchase agreement in 2017 valued at up to roughly a billion dollars, under which Verizon committed to buy substantial annual fiber volumes. What has changed is the competitive context around the renewal. In 2017 a carrier was securing favorable terms from a supplier with capacity to fill. In 2026 the supplier has an alternative buyer with deeper pockets and less price sensitivity, and both sides know it.

    The Lumentum Move Is a Thesis, Not a Contract

    Lumentum rose more than Corning did on a deal it is not a party to. That deserves scrutiny rather than celebration. Lumentum’s business is in lasers, optical components and the transceiver modules that convert electrical signals into light at each end of a link. Corning’s is in the glass in between, plus the cable and connectivity hardware around it. A carrier buying outside-plant fiber for a residential build does not, by that act, buy Lumentum’s datacom modules.

    There is a defensible version of the read-across. If the deal is taken as confirmation that optical demand is broad rather than concentrated in a few AI campuses, then the whole photonics supply chain should carry a higher expected utilization, and component makers with exposure to data center interconnect benefit from the same underlying traffic growth. That argument is reasonable. It is also unverified by anything in this announcement, and it is worth separating from the parts of the story that rest on a signed agreement.

    For readers who buy rather than trade, the practical signal is the same either way: the pricing and lead-time environment for optical capacity is tightening, and procurement teams should assume the seller’s leverage has improved.

    Where the Upcycle Could Break

    Fiber has burned optimistic investors before. In 2023 and 2024, carriers across North America and Europe worked down inventory they had bought during the previous supply scare and cut capital spending at the same time, and optical revenue at the major suppliers fell sharply despite unchanged long-run traffic growth. Purchase agreements did not prevent that, because most such agreements are volume frameworks rather than unconditional obligations to take delivery on a schedule.

    The specific risks to watch are familiar. Carrier capital budgets are set annually and can be cut. Higher-count cable designs and denser packing let a buyer deliver the same route miles with less glass over time. Competing suppliers in Asia have capacity and price incentives to take share when Western capacity is committed elsewhere. And an artificial intelligence build cycle that slows would release capacity back into the merchant market quickly.

    None of this argues the demand is illusory. It argues that a headline value spread over an undisclosed number of years, with undisclosed minimums, is not the same thing as booked revenue, and that the market’s willingness to reprice adjacent companies on the announcement is a measure of sentiment as much as of fundamentals.

    Background

    Corning has been a materials company since 1851 and produced the first low-loss optical fiber in 1970, the invention that made long-distance optical communication practical. Optical Communications is now its largest segment, selling fiber, cable, connectors and data center connectivity hardware to telecom carriers, cable operators, and increasingly to hyperscale and colocation operators building artificial intelligence capacity. Its results have historically tracked carrier capital spending cycles closely, which has made the business both large and volatile.

    Verizon is one of the largest telecommunications operators in the United States and one of the largest single purchasers of optical equipment in the market. Its fiber strategy serves two purposes at once: delivering broadband directly to homes and businesses, and carrying traffic back from the wireless sites that support mobile and fixed-wireless service. Lumentum sits in a different part of the same supply chain, supplying the lasers and optical modules used at each end of a link, with substantial exposure to data center interconnect demand.

    Sources

    Source: Corning rises 7% on Verizon fiber supply deal (GLW:NYSE) — report of Corning’s share move on a multibillion-dollar optical fiber supply agreement with Verizon. See also Lumentum Surges 11% as Corning’s Multi-Billion-Dollar Verizon Deal Lifts Fiber Stocks, on the read-across to optical component makers.

    Primary sources: Corning Incorporated, Form 8-K filed September 4, 2026 (Item 7.01, Regulation FD Disclosure), disclosing that Chief Financial Officer Edward A. Schlesinger will speak at the Citi 2026 Global TMT Conference on September 9, 2026 and expects to discuss recent business performance, progress on the Springboard plan, and market trends and demand drivers.

  • Cisco Buys WideField Security to Harden Splunk’s Identity Layer

    Cisco Buys WideField Security to Harden Splunk’s Identity Layer

    TL;DR · 30-second read

    The Short Version

    Many of the accounts logging into a large company’s systems are not people at all. They are software: automated services, and now artificial intelligence assistants that act on their own.

    Cisco has bought WideField Security, a company that keeps track of all of those accounts, human and machine, and flags the ones behaving strangely.

    Cisco will fold the technology into Splunk, the security monitoring software it already owns. Neither company has said what the deal cost.

    Cisco Systems has acquired WideField Security Inc., an identity-security software company, in a cross-border transaction that included a separate India workstream covering WideField’s Indian subsidiary. Bar & Bench reported on September 6 that JSA Advocates & Solicitors advised Cisco on the Indian legal and regulatory aspects — due diligence on the subsidiary, structuring, review of the US and Indian transaction documents from an Indian law perspective, negotiation, conditions-precedent fulfilment and closing assistance — with a team led by joint managing partner Vivek K. Chandy. Fenwick was Cisco’s lead counsel on the global transaction; Gunderson Dettmer represented WideField. No purchase price has been disclosed.

    WideField’s platform gives organisations visibility into human, non-human and AI-driven identities, and is intended to identify risk, strengthen authentication and support real-time threat detection and response. Its technology is to be incorporated into Cisco’s Splunk platform to enhance agentic security operations centre capabilities by normalising and correlating identity, session and activity telemetry drawn from multiple sources. The deal surfaces days after Cisco filed its annual report for fiscal 2026 with the Securities and Exchange Commission on September 2, covering the year ended July 25, 2026; that filing reports Cisco’s business across four revenue categories — networking, security, collaboration and observability.

    Executive Summary

    The strategic logic is narrow and specific rather than sweeping. Cisco is not buying an identity provider that issues logins; it is buying the layer that watches them. WideField’s stated capability — pulling identity, session and activity records from many systems and reconciling them into a single, comparable picture — is precisely the input a security operations centre needs before it can automate anything. Without normalised identity data, an AI-driven analyst is guessing.

    That points the deal squarely at Splunk, which Cisco acquired in 2024 in the largest transaction in its history. Splunk’s value has always been a function of what you feed it: it ingests machine data and lets analysts search and correlate it. Identity telemetry is one of the harder categories to ingest cleanly, because every directory, cloud provider and application describes an account differently. Buying a company that has already solved that plumbing is a faster path than building it inside a platform Cisco is still integrating.

    The timing matters for a second reason. As enterprises deploy AI agents that hold credentials and act autonomously, the number of identities inside a company grows without a corresponding growth in the number of employees. Each of those agents is an account that can be over-permissioned, hijacked or simply left running. Cisco’s four reporting categories in its fiscal 2026 annual report include security, and this acquisition is an argument about where that category’s growth is meant to come from.

    Identity Has Quietly Become the Control Plane

    For most of the last two decades, enterprise security spending followed the network: firewalls, segmentation, intrusion detection. That model assumed a perimeter. Cloud adoption dissolved it. When applications, data and staff all sit outside the corporate network, the only durable checkpoint left is the identity — the account, token or key presented at the moment of access. That is why identity security has become the fastest-moving line item in many security budgets, and why a networking vendor now needs a position in it.

    WideField’s framing of the problem is the current industry consensus stated plainly: human, non-human and AI-driven identities. Non-human identities are the service accounts, API keys and machine credentials that let one piece of software talk to another; they typically outnumber employee accounts by a wide margin, rotate rarely, and are governed far more loosely. AI-driven identities extend the problem, because an autonomous agent both holds credentials and generates activity that looks like a user’s. Telling a compromised agent from a busy one requires exactly the correlation of identity, session and activity data that WideField describes.

    The competitive read is that this pushes Cisco further into territory occupied by Microsoft, Palo Alto Networks, CrowdStrike and the identity-governance specialists. It does not, on the disclosed facts, make Cisco an identity provider competing with the directories enterprises already run. The nearer analogy is a monitoring and posture layer that sits above those directories — a position that is complementary to them right up until it is not.

    Splunk Is the Reason the Deal Makes Sense

    Read the acquisition as a Splunk roadmap item and it becomes legible. Cisco’s stated intent is to enhance Splunk’s agentic security operations centre capabilities — an operations centre where AI agents triage alerts, gather context and propose or take remediation steps, rather than a room of analysts doing it by hand. The bottleneck in that design is not the model. It is data quality. An agent cannot reason about whether a login is anomalous if the same person appears as three different identifiers across a cloud provider, a single sign-on tool and an internal application.

    Normalisation and correlation is unglamorous work, and it is also the work that determines whether an automated operations centre produces useful conclusions or expensive noise. Buying a team that has built it for identity data is a defensible use of capital, and it is a different kind of purchase from the volume-driven logic that governs data platform economics, where ingesting more telemetry raises both value and cost. What Cisco has not said is whether identity telemetry processed through WideField’s technology will be priced as part of existing Splunk entitlements or as a separate module — a commercially material question for anyone already managing a Splunk bill.

    There is also an integration risk worth naming without overstating it. Cisco is still absorbing Splunk. Layering a further acquisition into a platform mid-integration can accelerate a roadmap or fragment it, and the record of large networking vendors folding security startups into existing consoles is mixed. The measure of success will be whether the capability ships as a native part of the platform on a stated timeline, not whether it appears in a slide as an adjacency.

    A Tuck-In, Judged by What Was Not Announced

    No purchase price, no share count and no closing date have been made public, and Cisco has not published deal terms alongside the legal-advisory account. For a company of Cisco’s size that pattern usually indicates a transaction below the threshold at which separate disclosure is required — a tuck-in acquired for its technology and team rather than its revenue. WideField being represented by Gunderson Dettmer, a firm whose practice is built around venture-backed companies, is consistent with a private, venture-stage target rather than a mature software business, though the companies have not characterised it either way.

    That framing should temper expectations in both directions. A tuck-in will not move Cisco’s security revenue line in the near term, and readers should be sceptical of any suggestion that it repositions the company outright. Equally, tuck-ins are how platform gaps get closed, and the gap here is a real one. The honest assessment is that the strategic rationale is substantiated by the stated integration path, while the financial significance is simply unknown because it has not been disclosed.

    Why the India Leg Was a Separate Workstream

    The most concrete detail in the public account is structural: WideField had an Indian subsidiary substantial enough to require its own diligence, its own document review under Indian law, and conditions-precedent fulfilment before closing. The supporting workstreams named — employment, direct tax and indirect tax — indicate a subsidiary with employees and tax exposure worth examining, not a dormant shell. For an acquirer, that is where deal value quietly leaks: mispriced payroll liabilities, transfer-pricing history between the Indian entity and the US parent, and equity treatment for staff whose options convert on closing.

    This is a familiar shape for enterprise security startups, many of which run engineering and around-the-clock detection operations from India while selling into North America. It also means the acquisition carries a people question as much as a technology one. Retaining the engineers who built the normalisation logic is usually the difference between a capability that ships and a codebase that is quietly rewritten by the acquirer two years later.

    For buyers in India specifically, the practical implication is modest but real: capability originating in an Indian development centre is now attached to a global platform vendor’s support, compliance and procurement machinery. Neither company has said what happens to the subsidiary’s operations, headcount or reporting lines after closing.

    Background

    Cisco Systems is a global networking and technology company that builds AI-enabled networking, security and digital infrastructure products. Its annual report for fiscal 2026, filed with the SEC on September 2, 2026 and covering the year ended July 25, 2026, reports revenue across four categories: networking, security, collaboration and observability. The security and observability lines were substantially reshaped by Cisco’s 2024 acquisition of Splunk, a data platform used to ingest and analyse machine-generated logs, and the largest acquisition in Cisco’s history. Splunk gave Cisco a foothold in security analytics to sit alongside its long-standing network hardware franchise.

    Identity security emerged as a distinct market as enterprises moved workloads to the cloud and the traditional network perimeter stopped being a meaningful boundary. The category now spans identity providers that issue and verify credentials, governance tools that decide who should have access to what, and monitoring layers that watch how identities actually behave. The rapid adoption of AI agents — software that holds credentials and acts autonomously — has expanded the number of identities inside a typical organisation without a corresponding increase in staff, and has made behavioural monitoring of non-human accounts an active area of both product development and acquisition.

    Sources

    Source: JSA advises on India leg of Cisco’s acquisition of WideField Security — Bar & Bench, September 6, 2026, on the Indian, US and global legal workstreams behind Cisco’s purchase of WideField Security and the technology’s planned integration into Splunk.

    Primary sources: Cisco Systems, Annual Report on Form 10-K for the fiscal year ended July 25, 2026, filed with the SEC on September 2, 2026.

  • Fire at TeraWulf’s Lake Mariner Exposes AI Site Safety Gaps

    Fire at TeraWulf’s Lake Mariner Exposes AI Site Safety Gaps

    TL;DR · 30-second read

    The Short Version

    In early June, a building caught fire at a large computer facility in upstate New York, built on the site of a retired coal plant. Firefighters arrived to find no working fire alarm, no built-in system to put out flames, and three fire hydrants that did not work.

    The paperwork listing which chemicals were burning had, the owner said, burned in the fire itself. The fire chief said his crew went in blind.

    The site is being converted from mining digital currency to running artificial intelligence — the same conversion happening at dozens of sites nationwide.

    Crypto Briefing reported that firefighters from the Barker Fire Department responded in early June to a fire in an unfinished building at TeraWulf’s Lake Mariner campus in Somerset, New York, and found no functioning fire alarm, no suppression system and three hydrants that did not work. According to TeraWulf, the safety data sheets that would have identified the chemicals producing the heavy black smoke had burned up in the fire. Fire Chief Steve Matisz said his crew went into the building “kind of blind,” and that he was unsure what to make of the claim that the safety sheets had been destroyed. “It’s been a difficult situation,” Matisz said.

    Lake Mariner sits on the site of a retired coal plant that TeraWulf (Nasdaq: WULF) began operating in 2022 for Bitcoin mining and is now converting to artificial intelligence and high-performance computing workloads, targeting a combined 500 to 750 megawatts across several buildings by the end of 2026. Tenants include Core42 and Fluidstack. Post-incident discussions reportedly included plans for additional hydrants and improved storage of safety data sheets. No regulatory follow-up or investigative action has been publicly reported.

    Executive Summary

    The facts as reported are narrow — one fire, one building still under construction, one rural fire department — but they touch three things that matter across the AI build-out: whether life-safety systems are in service before a building is finished, whether a site’s water supply can actually deliver, and whether anyone outside the company checks. A megawatt is a unit of electrical demand; a campus rated at several hundred megawatts draws power on the scale of a small city, and it concentrates that energy inside a handful of buildings packed with electrical gear, batteries and, increasingly, liquid cooling.

    What makes Lake Mariner a useful case is not that TeraWulf is unusual. It is that TeraWulf is typical. A retired coal plant with an existing grid connection is among the most sought-after assets in the industry right now, because the interconnection queue — the waiting line to attach new load to the electrical grid — runs years long almost everywhere else. Buying speed by inheriting a brownfield site also means inheriting its fire mains, its hydrants and its water pressure.

    For tenants such as Core42 and Fluidstack, and for the customers behind them, the operational question raised here is not about one fire. It is about what construction-phase safety looks like at a campus racing toward 500 to 750 megawatts by the end of next year, and what an insurer, a lender or a lease counterparty is told when something goes wrong.

    Construction Is the Most Dangerous Phase of a Data Center’s Life

    A finished data center is one of the better-protected buildings in commercial real estate: detection on every ceiling, suppression plumbed and pressurised, and a fire alarm panel that is tested, certified and monitored before the first server is energised. A data center under construction is a different building entirely. Suppression piping is often installed but not charged. Detection is often wired but not commissioned. Combustible packaging, temporary power, insulation and hot work — welding, cutting, grinding — all sit in the same shell at once. That is why the industry treats the construction window as a distinct risk class with its own controls, rather than assuming a partly built facility is a slightly less protected finished one.

    The reported conditions at Lake Mariner map onto exactly that window: an unfinished building, no functioning alarm, no suppression. None of that is inherently a code violation — requirements phase in as construction progresses, and there is a legitimate period in any project when systems exist but are not yet live. What the account does not establish, and what would determine whether this was routine or a lapse, is which protections were required at that stage and which were merely absent. That is a question about the project’s fire-protection plan, and it is answerable only by the company and the authority having jurisdiction.

    Brownfield Speed Comes With Borrowed Plumbing

    The three dead hydrants are the detail with the longest reach. Converted industrial sites — retired coal, steel, paper, aluminium — are attractive because the heavy electrical infrastructure is already there and the grid connection already exists. But the fire-water infrastructure is inherited too, and it was engineered for the plant that used to be there, under a maintenance regime that ended when that plant did. Private hydrants on a large campus are the owner’s responsibility to flow-test and maintain, not the municipality’s. When three of them do not work, the failure is usually not the fire itself but everything that did not happen in the years before it.

    The response profile compounds this. Repurposed heavy-industrial land tends to be rural, which means the responding department may be small, volunteer or both, with tanker shuttles rather than an urban grid of reliable pressurised mains. The specific failure the chief described — going in without knowing what was burning, because the safety data sheets were reportedly inside the building that was on fire — is the one that pre-incident planning exists to prevent. Chemical inventories are meant to be duplicated off-site and shared with the responding department in advance, precisely so that a fire cannot destroy the document that explains itself. That the post-incident discussion reportedly turned to additional hydrants and better storage of those sheets suggests the gap was recognised.

    The Risk Is Priced Somewhere, Even If It Is Not Disclosed

    Fires during construction are covered by builder’s risk policies, and completed data centers are among the most heavily underwritten assets in the market — insurers inspect them, set conditions and price on the strength of protection systems. A loss during construction is not merely an expense; it becomes part of the underwriting file for the campus, and it can shape terms on renewal and on the buildings not yet built. Where a company is financing a rapid build-out, insurance terms and lender conditions are a real cost line, not a footnote.

    There is also a schedule dimension. TeraWulf is targeting 500 to 750 megawatts across several buildings at Lake Mariner by the end of 2026, with tenants already signed. Lease agreements in this market typically carry delivery dates, and a delivery date missed by construction damage is a commercial event as well as a safety one. Nothing in the reported account establishes that this fire affected the schedule, the tenants or the target — but those are the specific channels through which a construction-phase fire reaches investors, and they are the ones worth watching.

    Who Checks, and When

    The reported absence of publicly known regulatory follow-up is the part that generalises furthest beyond this one campus. Data center fire safety is enforced locally — by building departments, fire marshals and code officials — and the enforcement capacity of a small jurisdiction is not always matched to a multi-hundred-megawatt industrial conversion arriving inside a few years. That is a structural feature of how the AI build-out has distributed itself: toward places with cheap land, existing grid capacity and, often, modest municipal staffing.

    It is worth being precise about what can and cannot be concluded. That no follow-up has been publicly reported is not evidence that none occurred; investigations are frequently non-public until they conclude, and some are never announced at all. The fair reading is narrower and still useful: three months after an incident at a campus of this scale, the public record contains a fire chief’s account and little else. For a sector asking communities to host enormous new industrial loads, that asymmetry between the size of the facility and the visibility of its oversight is itself a thing to watch.

    Background

    TeraWulf (Nasdaq: WULF) started as a Bitcoin miner, a business whose economics turn almost entirely on the price of electricity. That pushed the industry toward the same places utilities and heavy industry once favoured: sites with large, already-permitted grid connections. Lake Mariner, on the grounds of a retired coal plant in Somerset, New York, is a textbook example, and TeraWulf has operated it since 2022.

    The rise of artificial intelligence turned those assets into something more valuable. Training and running large models requires dense clusters of specialised chips, and the binding constraint on building them is power — specifically, how quickly a site can attach hundreds of megawatts to the grid. Former mining campuses already have that, which is why so many miners are converting into landlords, leasing capacity to compute providers rather than mining coins themselves. TeraWulf’s Core42 and Fluidstack tenancies at Lake Mariner, alongside a target of 500 to 750 megawatts by the end of 2026, place it squarely in that transition. The conversion also changes the building: denser electrical gear, more batteries and increasingly liquid cooling, all of which raise the stakes on fire protection.

    Sources

    Source: TeraWulf’s Lake Mariner data center fire raises safety concerns — an account of the early-June fire at TeraWulf’s Somerset, New York campus, including Barker Fire Chief Steve Matisz’s description of the conditions his crew found on arrival.

  • Hut 8 Reprices Mining Power as AI Data Center Collateral

    Hut 8 Reprices Mining Power as AI Data Center Collateral

    TL;DR · 30-second read

    The Short Version

    • Hut 8 used to make most of its money running rooms full of computers that create bitcoin. It is now renting that electricity and space to artificial intelligence companies instead.
    • It has signed a 15-year lease worth $9.8 billion for a site in south Texas. The chipmaker Nvidia is the tenant; the AI firm Anthropic is the end user.
    • Hut 8 has also lined up $7.5 billion in borrowing to build it. The shares jumped 19 percent in a day.
    • Why it matters: electricity, not computers, is now the scarce ingredient in artificial intelligence — and whoever already has it can charge a premium.

    Hut 8 Corp. (Nasdaq: HUT) is converting its bitcoin mining power base into leased artificial-intelligence capacity, anchored by a 15-year lease at its Beacon Point campus in Nueces County, Texas with a base term valued at US$9.8 billion. The site is to be leased by Nvidia under Nvidia’s very large computing agreement with Anthropic, and the campus is planned for 1,000 megawatts of total capacity. Dealroom reported that Hut 8 secured $7.5 billion in financing as the pivot accelerated; the lease terms and a 19.0% single-session share move were reported by Simply Wall St on September 7, 2026.

    The transaction reframes Hut 8 from a company whose revenue rose and fell with the bitcoin price into one selling long-duration, contracted access to energized land and grid capacity — the same asset, sold on a different contract.

    Executive Summary

    The headline number is the lease, not the loan. A 15-year base term worth US$9.8 billion averages roughly $650 million a year of contracted rent — the kind of cash-flow profile that lenders underwrite against, and the reason a company still reporting large losses can raise billions. The financing is the consequence; the lease is the cause.

    That is the structural shift worth noting. For most of the last decade, a bitcoin miner’s power portfolio was valued as an input cost — cheap electricity that made hashing marginally profitable. Under an AI lease, the same substation, the same interconnection agreement and the same acreage become collateral: a contracted, investment-grade-adjacent revenue stream that can be pledged. Miners are not being revalued because mining improved. They are being revalued because someone else needs their grid position.

    The corollary is less flattering to the miners. When roughly $7.5 billion of debt sits against a $9.8 billion base term, the counterparty’s credit and the lease’s covenants — not the operator’s ambitions — set the cost of capital and the order of claims on the cash. Hut 8 has traded volatility for visibility, and a measure of control with it.

    The Asset Was Never the Bitcoin. It Was the Interconnect.

    Hut 8 describes itself as an energy infrastructure platform integrating power, digital infrastructure and compute across the United States and Canada. That description has been true for years; what changed is the buyer. Bitcoin mining was an application that could tolerate cheap, interruptible, badly located power because the output was a fungible commodity produced anywhere. AI training and inference cannot — but it will pay handsomely for the one thing miners accumulated by accident: sites that are already connected to the grid at scale.

    Connecting a large load to a transmission network requires an interconnection agreement, and the queue for one — the utility’s study-and-approval process for new large customers — commonly runs for years in the busiest US markets. An operator holding an executed interconnection and an energized substation is holding a scarce option, and the Beacon Point lease is what that option is worth once a hyperscale tenant bids for it. This is why the repricing is spreading across the mining sector rather than being specific to one company: the balance sheet item being revalued is grid position, and several miners hold it.

    The Texas Gulf Coast location matters too. Nueces County sits in a market with abundant generation and industrial land, and Texas has become the default destination for very large new loads. It is also a region where firm power, water availability and summer grid stress are live operational questions rather than settled ones.

    A 15-Year Lease Is a Credit Instrument Before It Is a Building

    A $9.8 billion base term is not revenue in hand; it is the undiscounted sum of contracted rent over 15 years, before any renewal or escalation, and before Hut 8 spends the capital required to deliver the facility. Averaged evenly, that is on the order of $650 million a year. What makes the figure bankable is who is on the other side of it: the campus is to be leased by Nvidia under its computing agreement with Anthropic, which puts a large, well-capitalised technology company in the tenant position rather than a start-up burning venture funding.

    That structure is the point of the whole trade. A tenant of that quality converts a speculative construction project into something closer to a contracted infrastructure asset — the same reason toll roads and fibre routes get financed on the strength of their offtake agreements. It also concentrates risk. The lease reportedly covers a campus planned at 1,000 megawatts, but the disclosed figures do not establish how much of that 1,000 megawatts the $9.8 billion actually contracts. If the lease covers a portion, the remainder is a speculative build that must be leased separately; if it covers the whole campus, then the implied rent per unit of capacity is the number every competing developer will be modelling.

    There is a delivery risk that no counterparty credit can eliminate. Rent on a data center lease typically begins when capacity is delivered, not when the ink dries. Between now and then sit permits, transformers, switchgear, long-lead electrical equipment and construction labour — the same constraints slowing every other campus in the market.

    $7.5 Billion Against $9.8 Billion: Who Sets the Terms Now

    Put the two numbers side by side and the balance of power becomes visible. Roughly $7.5 billion of financing sits against $9.8 billion of base-term rent — meaning a large majority of the contracted cash flow over 15 years is, in effect, spoken for by capital providers before equity holders see anything. Whether that debt sits at the project level or at the parent, what it costs, when it matures, what covenants it carries and whether it comes with warrants or convertible features are the variables that determine how much of this deal’s economics Hut 8 actually keeps. None of that has been laid out publicly.

    This is the sharper version of the story than “miner pivots to AI.” In a mining business, the operator controlled the dials: turn machines off when power prices spike, sell coins when it suits, defer capital expenditure. Under a long-dated lease funded by multi-billion-dollar structured debt, the dials belong to the documents. Delivery dates are contractual. Debt service is fixed. The flexibility that made miners resilient through crypto drawdowns is the specific thing being exchanged for revenue visibility.

    The market’s uncertainty about how that trade lands is unusually wide. Third-party modelling published alongside the news projects Hut 8 reaching roughly $1.7 billion of revenue and $194.2 million of earnings by 2029 — which would require about 74.6% annual revenue growth and roughly a $793 million swing from a current trailing loss of $599.2 million — and derives a fair value of $161.78. More cautious estimates in the same set assume about $1.0 billion of 2029 revenue and land near $65. These are analyst and community model outputs rather than company guidance, and a spread of that magnitude on the same asset base is itself the finding: the disagreement is not about whether the lease exists, but about whether it gets built, financed and repeated.

    Nvidia Is Now in Everyone’s Capital Stack

    Nvidia’s role here is worth isolating, because it is no longer simply the company selling accelerators into these buildings. At Beacon Point it appears as the leaseholder, standing between the landlord and the ultimate AI workload. That is a chip vendor underwriting real estate risk to make sure its own hardware has somewhere to run.

    The breadth of that posture is visible elsewhere in Nvidia’s disclosures. In an 8-K filed September 3, 2026, Nvidia reported that it had entered a definitive agreement on September 2 to acquire Hugging Face, the open-model platform, for a purchase price of approximately $11.9 billion plus an equity-based retention program of up to roughly $1.0 billion, with closing expected in the first half of 2027. The filing also flags a specific vulnerability: government restrictions on open-source models, including models originating in China, could materially affect the platform and Nvidia’s results. A company deploying capital simultaneously into model distribution and into leased data center capacity is building demand for its silicon at both ends of the stack.

    For infrastructure buyers and investors, that has a practical implication. The creditworthiness underpinning a growing share of new US data center capacity increasingly traces back to a small number of very large technology balance sheets. Concentration of that kind is what makes projects financeable today, and it is also the correlation that would matter most if AI capital expenditure plans were ever revised downward.

    Background

    Hut 8 Corp. trades on Nasdaq as HUT and describes itself as an energy infrastructure platform that integrates power, digital infrastructure and compute at scale for energy-intensive uses across the United States and Canada. Historically its revenue came from bitcoin mining — running specialised computers that earn newly issued bitcoin — through its self-mining fleet and American Bitcoin operations. That business is transaction-driven: revenue moves with the bitcoin price, mining difficulty and power costs, which makes forecasting cash flow difficult.

    The wider shift began when AI compute demand collided with a shortage of grid capacity. Miners had spent years assembling sites with executed interconnection agreements and substations already in place, precisely the constraint slowing new AI data centers. Since 2024 several miners have moved to lease that capacity to AI and cloud tenants under multi-year contracts, trading commodity exposure for contracted rent. Hut 8’s Beacon Point lease in Nueces County, Texas is one of the largest examples of that conversion to date.

    Sources

    Source: Why Hut 8 (HUT) Is Up 19.0% After Pivoting Bitcoin Mining Power Into AI Data Centers — Simply Wall St, September 7, 2026, on the Beacon Point lease terms and Hut 8’s shift from mining to contracted AI infrastructure. See also Hut 8 lands $7.5B financing as AI pivot accelerates with Anthropic-linked deal, Dealroom.

    Primary sources: NVIDIA Corporation, Form 8-K filed September 3, 2026 (SEC EDGAR) — disclosing the September 2, 2026 definitive agreement to acquire Hugging Face, Inc. for approximately $11.9 billion plus an equity-based retention program of up to approximately $1.0 billion, with closing expected in the first half of 2027, and setting out related regulatory risks to open-source model distribution.

  • Google’s $3.2B TeraWulf Backstop: Who Really Funds AI Data Centers

    Google’s $3.2B TeraWulf Backstop: Who Really Funds AI Data Centers

    TL;DR · 30-second read

    The Short Version

    Google has promised to cover $3.2 billion of rent payments on a data center it will not own and will not occupy. The site is a retired coal plant near Buffalo, New York, run by a company that used to mine bitcoin.

    Without that promise, lenders would not fund the buildings. With it, construction can proceed. In return, Google holds rights to roughly 14 percent of the company.

    The borrowing stays off Google’s own books. So, largely, would the loss if the project failed.

    Startup Fortune reported that Google has backstopped $3.2 billion in lease obligations tied to TeraWulf’s AI data center campus at Lake Mariner in western New York, adding $1.4 billion in new support on top of earlier commitments. In exchange, warrants held by Google now represent roughly 14% of TeraWulf, up from about 8% months earlier.

    Google is not the tenant. Fluidstack, an AI cloud provider, holds two 10-year leases at the site covering more than 200 megawatts of critical IT load — the power available to servers themselves, excluding cooling and building overhead — expanded by a 160-megawatt building known as CB-5 to a contracted total of 360 megawatts. Those leases have been valued at $3.7 billion over the initial term and up to $8.7 billion if Fluidstack exercises its extension options. TeraWulf, a Maryland-based bitcoin miner, has spent roughly two years converting the former coal plant site outside Buffalo into an AI campus.

    Executive Summary

    The headline number is a guarantee, not a purchase. Google is not buying a data center, not leasing one, and not putting one on its books. It is standing behind someone else’s rent so that a third party’s lenders will fund construction. That single move is the reason CB-5 and the buildings behind it become financeable at all.

    The economics are straightforward once you strip out the jargon. TeraWulf cannot borrow billions against its own credit history; a former crypto miner’s balance sheet does not support project debt at that scale. Fluidstack’s ten-year lease payments can support it — but only if a lender believes those payments will arrive for a decade. Google’s backstop supplies that belief. TeraWulf gets a construction loan, Fluidstack gets capacity, and Google gets warrants, an expanding equity position, and access to compute without a capital expenditure line.

    This matters beyond one campus in New York. It is the same logic operating at far larger scale elsewhere in the sector, and it means the binding constraint on AI infrastructure is increasingly not land, turbines or chips. It is whose credit sits behind the lease.

    Credit Is the Scarce Input, Not Land

    The industry talks about AI capacity in megawatts, but megawatts are a downstream output. A developer with a permitted site, an interconnection position and a signed tenant still cannot pour concrete without several billion dollars of debt, and debt is priced against the weakest promise in the chain. TeraWulf’s site work at Lake Mariner is real; its balance sheet history is a bitcoin miner’s. Those two facts pull in opposite directions when a credit committee sits down.

    A backstop resolves the contradiction. In plain terms, Google has agreed to make the lease payments if the arrangement falls apart, which allows lenders to underwrite the loan against Google’s ability to pay rather than TeraWulf’s. The developer keeps the project, the operating risk and the upside. The hyperscaler takes on a contingent obligation — one that shows up as a commitment rather than as borrowed money — plus warrants, which are contracts giving the right to buy shares later at a fixed price. Those warrants moving Google from roughly 8% to roughly 14% of TeraWulf is the price of the credit enhancement, paid in equity instead of cash.

    For anyone tracking who builds AI capacity over the next three years, this reframes the competitive question. The winners are not necessarily the developers with the best sites. They are the developers who can attach an investment-grade counterparty to their lease stack.

    The Lease Is the Collateral

    Fluidstack’s 360 megawatts of contracted capacity at Lake Mariner is what the debt is actually secured against. A ten-year lease valued at $3.7 billion over its initial term, rising toward $8.7 billion with extensions, is a cash flow stream — and project finance exists to turn predictable cash flow streams into buildings. The tenant’s obligations, not the landlord’s assets, do the heavy lifting.

    That structure has an obvious sensitivity. The loan is sized to a tenant. If the tenant disappears, the campus is worth whatever the next tenant will pay, which in a softer market is not the same number. Google’s guarantee is the shock absorber, and its existence is a fair signal of how lenders assess a decade-long AI cloud commitment on its own: not sufficient without support. That is an observation about credit markets, not a judgment on Fluidstack’s business.

    It also explains why converted industrial sites keep winning. A retired coal plant near Buffalo already has transmission, substation land, water and a grid position that a greenfield site would spend years acquiring. Crypto miners bought many of those sites cheaply in a different cycle, which is why several of them are now landlords to AI tenants rather than operators of hash rate.

    Off the Balance Sheet, Not Out of the System

    TeraWulf’s structure is modest against the largest comparable. Meta’s Hyperion campus in Richland Parish, Louisiana — around 4 million square feet, expected to draw up to 5 gigawatts on completion in 2029 — was financed through a joint venture with funds managed by Blue Owl Capital, with Blue Owl holding 80% and Meta 20%. Morgan Stanley arranged $27 billion of debt and $2.5 billion of equity through a special purpose vehicle, a standalone company created to hold a single project’s assets and borrowings, with Pimco anchoring a bond sale rated A+ by S&P and maturing in 2049. Because Meta does not control the entity holding the debt, the debt does not appear on Meta’s balance sheet; the long lease is recorded as an operating cost instead.

    The aggregate is no longer marginal. A study cited by Forbes found the five largest US hyperscalers have moved more than $1.6 trillion in AI infrastructure debt into footnotes rather than balance sheets, and Moody’s separately estimates roughly $662 billion of hyperscaler lease commitments now sit outside reported debt — with more than $120 billion of that shift occurring in under two years. None of this is improper; lease and consolidation accounting is well-established, and the commitments are disclosed. But an investor reading only reported debt is reading a partial picture of sector leverage.

    Where the First Loss Actually Lands

    The structure relocates risk rather than eliminating it. At Lake Mariner, if the tenant walked, TeraWulf would hold a partially built campus with debt sized to a lease that no longer exists, and Google’s backstop would be the barrier between that debt and default. In the Louisiana model, private credit investors and bondholders hold the majority position and would absorb first losses, while the hyperscaler’s exposure is capped by a minority equity stake and its continuing rent obligation.

    That asymmetry is a feature, not an accident, and both sides enter it knowingly. Private credit funds are paid a spread precisely to hold decade-length demand risk that public equity investors would punish on a hyperscaler’s own books. The judgment being made across the sector is that AI compute demand holds up long enough to service twenty-plus-year paper. If it does, every party in the stack is paid as modeled. If it does not, the losses surface first in private credit vehicles and among warrant-holding partners, where they are less visible and slower to be marked than a write-down at a listed technology company would be.

    For enterprise buyers of AI capacity, there is a practical takeaway. When you sign with a neocloud or a converted-site operator, it is worth understanding who stands behind the building — because the entity on your contract may not be the entity whose credit made the building possible.

    Background

    TeraWulf is a Maryland-based company that built its business mining bitcoin — running warehouses of specialized computers that consume large amounts of electricity. When mining economics tightened, several operators discovered that their most valuable asset was not the hardware but the sites: power-rich industrial land with grid connections already in place. Over roughly two years, TeraWulf has been converting its Lake Mariner property, a retired coal plant site in western New York, into an AI data center campus leased to third parties.

    The financing model around it is newer than the buildout itself. AI campuses now routinely cost more than a mid-sized developer can borrow, so the industry has converged on project finance structures: a long lease from a creditworthy tenant, debt raised against that lease, and a special purpose vehicle or credit guarantee that keeps the borrowing separate from the hyperscaler’s own balance sheet. Meta’s Hyperion campus in Louisiana, financed through a joint venture majority-owned by Blue Owl Capital funds with bonds anchored by Pimco and rated A+ by S&P, is the largest example to date. Google’s TeraWulf backstop is a smaller variant of the same idea.

    Sources

    Source: Google’s $3.2 Billion TeraWulf Backstop Shows How AI Data Centers Get Built — reporting on Google’s lease guarantee for TeraWulf’s Lake Mariner campus, Fluidstack’s 360 megawatts of contracted capacity, and the off-balance-sheet financing structures behind large AI data center projects.

  • TSMC, Samsung Back ASML’s Next-Gen EUV Machines

    TSMC, Samsung Back ASML’s Next-Gen EUV Machines

    TL;DR · 30-second read

    The Short Version

    The world’s two largest advanced chip factories, Taiwan Semiconductor Manufacturing Company and Samsung, have agreed to buy the newest chip-printing machines from the Dutch equipment maker ASML. These machines use light to print circuit patterns onto silicon wafers, a bit like a photographic enlarger working at a scale smaller than a virus.

    ASML and the Taiwanese chipmaker also plan to switch to glass stencils twice as wide as today’s. Wider stencils mean bigger chips can be printed in one shot, and bigger chips printed in one shot are cheaper chips. Those are the chips that run artificial intelligence.

    ASML has secured commitments from Taiwan Semiconductor Manufacturing Company (TSMC) and Samsung Electronics for its newest generation of extreme ultraviolet (EUV) lithography systems, according to Bloomberg, with CNBC and Seeking Alpha reporting the same commitments and tying the demand to the artificial-intelligence buildout. Separately, Focus Taiwan and EU Today reported that TSMC and ASML are working together on 12-inch EUV photomasks, a step up from the 6-inch masks that have been the industry standard for decades, intended to widen what High-NA machines can print.

    Neither the tool commitments nor the photomask programme came with disclosed unit volumes, pricing, delivery dates or the process nodes involved. What is established is the direction: the two largest advanced-logic manufacturers on the planet have now publicly aligned themselves with ASML’s next lithography platform, and one of them is prepared to change the format of the mask itself to get more out of it.

    Executive Summary

    Lithography is the step in chipmaking where a circuit pattern is projected onto a silicon wafer. ASML is the only company in the world that builds EUV lithography systems, and its High-NA generation — ‘NA’ meaning numerical aperture, a measure of how much light the lens can collect and therefore how fine a pattern it can resolve — is the successor platform that the entire logic roadmap eventually has to migrate to. TSMC and Samsung committing to those tools removes the largest open question about High-NA: whether the volume foundries would adopt it, or ride existing EUV with multiple patterning for another generation.

    The photomask news is the more technically interesting half. A photomask is the stencil carrying the circuit pattern; today it is a 6-inch quartz plate, and its size sets a hard ceiling on how large a single chip can be. High-NA optics buy resolution by narrowing the printed field, which makes that ceiling lower, not higher — precisely the wrong direction for AI accelerators, which are among the largest chips ever manufactured. A 12-inch mask is the industry’s way of buying the field back.

    For anyone buying AI compute rather than making it, the through-line is cost per die. If large accelerators can be printed in one exposure on a next-generation tool instead of being stitched together or broken into smaller chiplets, the economics of every downstream layer — server, rack, data centre, inference price — shift in the buyer’s favour, eventually.

    The Mask, Not the Machine, Is the Constraint

    Standard lithography prints an exposure field of roughly 26 by 33 millimetres, and the size of a single chip cannot exceed it. That is the ‘reticle limit’, and the largest AI accelerators and their high-bandwidth memory stacks have been pressed against it for years. High-NA EUV improves resolution using anamorphic optics that magnify differently in each direction, and the price of that trick is a printed field roughly half the conventional size. A foundry adopting High-NA for critical layers therefore inherits a smaller canvas at exactly the moment its most valuable customers want a larger one.

    Moving the mask from 6 inches to 12 inches attacks that directly: a bigger stencil is what allows a bigger field to be projected in a single exposure. The engineering consequences ripple outward — mask blanks, pellicles, writing tools, inspection and repair equipment, scanner stages and reticle handling all assume the incumbent format. That is why a joint TSMC-ASML effort matters more than a unilateral one. Mask formats are effectively an industry standard, and standards move when the largest buyer and the sole tool supplier move together.

    AI Accelerator Economics Are Driving the Roadmap

    Historically, lithography generations were pulled forward by mobile and PC volumes, where the metric that mattered was transistors per dollar across hundreds of millions of small dies. The demand now being cited for these tools comes from a different shape of customer: relatively low unit volumes of extremely large, extremely expensive dies sold into data centres. For that customer, yield loss on a big die is brutal — defects scale with area — and every millimetre of printable field is margin.

    This is also a reminder of where the bottleneck in the AI buildout actually sits. Public attention fixes on power and cooling, and those are real constraints. But the supply of leading-edge accelerators is gated upstream by a single Dutch tool vendor’s output and by the willingness of two or three foundries to commit capital years ahead of revenue. TSMC’s capital cadence remains visible in its filings — in a 6-K filed with the SEC on September 1, the company adjusted its first-quarter 2026 cash dividend to NT$7.00000137 per common share, payable October 8, following a board-approved distribution of NT$181.5 billion — a reminder that this capital intensity coexists with substantial cash returns to shareholders.

    Who Gains, and What Has to Go Right

    ASML is the clearest beneficiary. It has no competitor in EUV, and commitments from both leading foundries convert High-NA from a technology bet into a platform with a customer base. TSMC gains optionality: it can keep extending existing EUV with multiple patterning where that is cheaper, and deploy High-NA where the layer count justifies it, while shaping the mask format that its own largest customers will design against. Samsung’s commitment is competitively significant on different terms — foundry share is won partly on the credibility of the next node, and being absent from the next lithography platform would have been a liability.

    The risks are not trivial. Larger masks mean larger, heavier, more expensive reticles moving through a machine whose overlay budget is measured in fractions of a nanometre; a 12-inch blank supply chain has to be built and qualified; and the mask ecosystem’s inspection and repair tooling has to follow. There is also a live alternative. The industry’s current answer to the reticle limit is disaggregation — splitting a large chip into chiplets and reassembling them in advanced packaging. If packaging keeps improving quickly, some of the value of a bigger field is captured elsewhere. The likeliest outcome is that both paths advance together, with larger single-exposure dies making the packaged modules themselves more capable.

    Background

    ASML, based in Veldhoven in the Netherlands, is the world’s only manufacturer of extreme ultraviolet lithography systems, the machines that print the finest features on leading-edge chips. Its equipment is therefore a chokepoint for the entire advanced semiconductor industry, and a recurring subject of export-control policy. TSMC is the largest contract chipmaker in the world, manufacturing processors designed by others, including most of the leading AI accelerators. Samsung Electronics runs both its own memory and logic products and a foundry business competing with TSMC for external customers.

    The photomask format at issue has been stable for decades: 6-inch square plates, with a standard exposure field that fixes the maximum size of any single chip. That constraint was largely academic when the industry’s economics were driven by small mobile and PC processors. It became a first-order commercial problem once AI accelerators, built as the largest practical dies and surrounded by stacked memory, became the most profitable products in the industry.

    Sources

    Source: ASML and TSMC plan larger masks to expand High-NA chipmaking — report that ASML and TSMC intend to adopt 12-inch photomasks to widen the printable field of High-NA EUV systems.

    Primary sources: Taiwan Semiconductor Manufacturing Company, Form 6-K filed September 1, 2026 (adjustment to cash dividend per share); ASML Wins Over Top Chipmakers for New EUV Machines — Bloomberg; TSMC, Samsung commit to ASML’s newest chipmaking tools as AI drives demand — CNBC; TSMC, ASML partner on 12-inch EUV photomasks for advanced chips — Focus Taiwan; ASML wins commitments from Samsung, TSMC for new chipmaking gear — Seeking Alpha; ASML and TSMC plan bigger masks that could lower chipmaking costs — Stock Titan.