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

Author: Deepak Jain

  • TAR’s $1B Series A Prices Off-Grid Power for AI Data Centers as Its Own Category

    TAR’s $1B Series A Prices Off-Grid Power for AI Data Centers as Its Own Category

    TL;DR · 30-second read

    The Short Version

    TAR, a Texas company started only this year, has raised $120 million from investors who now value it at $1 billion.

    It builds private power plants for the giant computer warehouses behind artificial intelligence tools. The plants use renewable energy and huge batteries, and never plug into the public electricity grid.

    Why it matters: these warehouses use so much electricity that many wait years to be connected. TAR says its approach lets them get power without competing with nearby homes and businesses for it.

    Austin-based TAR has raised $120 million in Series A funding at a $1 billion post-money valuation, Pulse 2.0 reported on September 11, 2026. Spark Capital led the round. TAR was founded in 2026 by Pat Becker and Leonhard (Lenny) Soenke. It builds self-contained, modular power systems that combine renewable generation with battery storage and run independently of the electric grid, aimed at AI data centers at gigawatt scale.

    The company says the money will expand its Austin headquarters and its San Francisco engineering office. It will also grow manufacturing and logistics operations in West Texas and speed up deployments already underway. Those deployments include a utility-scale project with one of the largest neoclouds, a dedicated project campus, and TAR Terminal One, a manufacturing and logistics center in West Texas.

    Executive Summary

    TAR’s raise is one of the clearest signs yet that investors see off-grid power for AI as a business category in its own right. It is no longer just a temporary fix for data centers waiting on a utility hookup. A company founded this year has reached unicorn status on its first institutional round by promising dedicated renewable-plus-battery generation that it designs, builds and operates itself.

    The thesis rests on two claims. The first is that power, not chips, is now the binding constraint on AI computing. The second is that the fastest way to supply it is to own the whole project lifecycle and automate construction. TAR pairs energy veterans from Hut 8, AES and Vistra with robotics engineers from Zipline, GrayMatter Robotics and Lucid Motors, which suggests an attempt to treat power plant construction partly as a manufacturing problem.

    What the company has not yet made public matters as much as the valuation. TAR has not disclosed installed megawatts, its customer’s name, contract terms or project timelines. The central engineering question for any off-grid system serving round-the-clock computing is how it maintains reliable supply, and TAR has not yet explained its approach.

    Off-Grid Power Gets Its Own Venture Price Tag

    For much of the data center boom, on-site generation has been framed as a bridge. It energizes a campus while the project waits in a utility’s interconnection queue, the line of projects awaiting approval to connect to the grid. A $1 billion post-money valuation for a company founded in 2026 suggests Spark Capital sees something more durable: a standalone category in which generation is built and run as a permanent, dedicated supply for computing.

    The arithmetic is instructive. A $120 million raise at $1 billion post-money implies a pre-money valuation of roughly $880 million. TAR reached that figure without publicly disclosing completed megawatts, contracted revenue or a named customer. At that level, investors are paying for team, speed and position in a market where, as Spark’s Will Reed put it, “power is becoming the main bottleneck to scaling compute.” It is a bet on the size of the problem as much as on a track record.

    The deal also points to a shift in where AI infrastructure capital goes. Energy assets have traditionally been funded by utilities, infrastructure funds and project lenders, who price them for steady, bond-like returns. A venture-style valuation for a power developer implies that some investors now treat the speed of energy delivery, and not only its cost, as a source of technology-like upside.

    Owning the Whole Stack: Faster Delivery, Concentrated Risk

    TAR’s strategy is vertical integration. It handles site selection, detailed engineering, procurement, logistics, civil construction, commissioning and ongoing operations itself, rather than coordinating a chain of outside contractors. The reasoning is that schedules tend to slip at the handoffs between firms. Co-founder Pat Becker said gigawatt-scale deployments (a gigawatt is a thousand megawatts, comparable to the output of a large power plant) “in tight time windows necessitate owning the full stack end-to-end.”

    The company’s purpose-built deployment automation stack is meant to install generation faster, at larger scale and with less field labor than conventional construction. If it works, the limit on how fast power can be built shifts away from the supply of skilled field crews and toward factory and logistics throughput. TAR Terminal One in West Texas appears designed to supply that throughput.

    The trade-off is concentration of risk. A contractor ecosystem spreads execution risk across many firms, while an integrated developer carries all of it. TAR must scale engineering, robotics, manufacturing, logistics and field operations at the same time. That is capital-intensive and operationally demanding. TAR has not yet disclosed completed projects that show its automation claims working at scale.

    The Reliability Question for Renewables Plus Batteries

    AI data centers run around the clock, and their customers expect very high uptime. A grid connection offers a backstop when on-site generation falls short, and an off-grid system has no such backstop. TAR describes its systems as combining renewable generation with battery storage and operating independently of the grid. The design therefore has to cover periods of weak sun or wind entirely on its own.

    That problem can be solved in several ways: oversizing generation, adding longer-duration storage, or including firm backup generation. Each adds cost, equipment and land. How TAR balances these choices will largely decide whether its power competes on price and reliability with grid electricity or with gas-fired on-site plants, the other common off-grid option.

    TAR’s community argument does resonate. Its systems do not draw on shared grid capacity, so the company says data center projects can add generation without competing with surrounding communities. That matters in siting debates, where the effect of very large facilities on local electricity supply and prices is a recurring concern.

    Who Stands to Gain if the Model Works

    Neoclouds are natural early customers. These are cloud providers that specialize in renting out GPU computing for AI, as opposed to the broad hyperscale clouds. They compete heavily on how quickly they can bring capacity online, which makes long grid timelines especially costly for them. TAR says it is already executing a utility-scale deployment with one of the largest neoclouds, a sign that its offer has found at least one buyer with urgent demand.

    For utilities and grid operators, the effects are mixed. Off-grid campuses ease pressure on congested transmission lines, but they also mean large new electricity users that never become utility customers. Suppliers of batteries and generation equipment, along with West Texas manufacturing and logistics workforces, stand to benefit from any sustained buildout. TAR’s competition includes other on-site power developers and the traditional engineering and construction firms whose multi-contractor model it is explicitly positioning against.

    Background

    AI data centers use far more electricity than traditional facilities, and in many regions connecting a new large user to the grid means joining a long interconnection queue and waiting for new transmission and generation. That has pushed developers toward “behind-the-meter” power: generation built on or next to the site that serves the data center directly. Such projects have often been treated as temporary bridges until a grid connection arrives.

    TAR, founded in 2026 in Austin, is pursuing the off-grid approach as a permanent model. It uses modular renewable generation and battery storage and manages projects end to end. Its lead investor, Spark Capital, is a venture capital firm. The company’s leadership draws on experience at energy and infrastructure companies including Hut 8, AES and Vistra, as well as robotics and hardware companies Zipline, GrayMatter Robotics and Lucid Motors.

    Sources

    Source: TAR Raises $120 Million Series A At $1 Billion Valuation To Build Off-Grid Power For AI Data Centers (Pulse 2.0), covering TAR’s Spark Capital-led round to expand off-grid renewable and battery power systems for AI data centers.

  • Shanghai Electric Lands First Overseas Gas Turbine Order in Malaysia

    Shanghai Electric Lands First Overseas Gas Turbine Order in Malaysia

    TL;DR · 30-second read

    The Short Version

    Shanghai Electric, a large Chinese engineering company, has won its first order outside China for the giant gas-burning engines that drive power station generators. They will go into a new power plant in Malaysia.

    Why it matters: the few American, European and Japanese firms that dominate this business have waiting lists stretching years, partly because artificial intelligence needs so much electricity. A new supplier promising machines from 2028 could give power-hungry countries another option, if its equipment proves itself abroad.

    Shanghai Electric (SEHK: 02727, SSE: 601727) announced in a press release on September 14, 2026 that it has won the contract for Unit 3 of the Sarawak Samalaju Combined Cycle Gas Turbine (CCGT) Project in Malaysia, which the company describes as a 500 MW project. It is the Chinese manufacturer’s first overseas order for heavy-duty gas turbines, the large industrial machines used in utility-scale power plants.

    The company will deliver the plant on an EPC (engineering, procurement and construction) turnkey basis and provide a 25-year long-term service agreement covering all major equipment. The customer is identified only as a Malaysian energy developer, and no contract value was given.

    Executive Summary

    The deal is broader than a turbine sale. Shanghai Electric says it will build the whole gas-fired plant and manufacture every core component itself: the gas turbines, the steam turbines, the generators, the heat recovery steam generators (boilers that capture hot turbine exhaust to make steam) and the air-cooled systems. It will also be the sole maintenance and service provider for 25 years, a bundle the company says puts it on par with the established leaders in heavy-duty gas turbines.

    The timing matters. Demand for gas-fired generation has surged, driven in part by electricity-hungry AI data centers, and the dominant turbine makers, GE Vernova, Siemens Energy and Mitsubishi Heavy Industries, are working through backlogs that push new deliveries years out. Shanghai Electric says it can deliver new units as early as 2028 and that developers in Indonesia, Thailand, the Philippines and Vietnam have expressed strong interest. If that interest turns into orders, Chinese manufacturers could become a real alternative in a market with few suppliers.

    For now this is one contract with an unnamed customer and undisclosed terms. It is a meaningful first step abroad, not yet proof of international competitiveness at scale.

    A Seller’s Market Opens a Door

    Heavy-duty gas turbines are among the hardest machines in the energy industry to make. They run at extreme temperatures, need advanced materials and cooling designs, and have to operate reliably for decades. Only a handful of companies build them at utility scale, which gives buyers little leverage when demand rises. Demand has risen sharply: utilities are replacing coal, balancing intermittent renewables and, increasingly, supplying power to large data center campuses. The incumbents have responded with fuller order books and longer waits for new units.

    That is the opening Shanghai Electric is aiming at. Its claim that new units could ship as early as 2028 is its most commercially pointed statement, because developers who cannot secure turbines cannot build plants. For a buyer facing a multi-year wait, an earlier delivery slot can outweigh a supplier’s shorter international track record. The release does not say, however, how many 2028 slots exist or whether the Malaysian project will use one of them.

    The likely early winners are developers in markets where cost and schedule matter most and where Chinese engineering and financing are already familiar. The established manufacturers are unlikely to feel much effect in the near term; one contract does not change a global order book. The bigger question is whether a credible alternative supplier eventually eases pricing power in a market that has run tight.

    The Turnkey Bet: Building It and Keeping It Running

    A combined cycle plant burns natural gas in a gas turbine, then uses the turbine’s hot exhaust to raise steam for a second turbine, getting more electricity from the same fuel. Shanghai Electric’s decision to make all the core parts in-house and wrap them in an EPC contract means the customer deals with a single counterparty for design, equipment, construction and integration. That simplifies coordination and puts responsibility for performance in one place.

    The 25-year service agreement may be the most important piece economically. Across the industry, long-term service contracts covering inspections, replacement parts and overhauls are a large and steady source of revenue for turbine makers over a plant’s life, often more durable than the original equipment sale. Securing one on its first overseas heavy-duty project gives Shanghai Electric a recurring revenue stream and, just as valuable, a site outside China where its equipment will build an operating record.

    For the customer, the trade-off cuts both ways. Single-source responsibility is simpler, but it concentrates risk: if the equipment underperforms or parts supply is disrupted, there is no second vendor to turn to for major components during the service term. The strength of that arrangement will depend on performance guarantees and penalty clauses, which have not been disclosed.

    What the Track Record Does and Does Not Show

    Shanghai Electric cites 103 heavy-duty units delivered, more than 21,000 MW of installed capacity from commissioned projects, and more than 1.3 million operating hours on units under its long-term service programs. For scale, a single turbine running around the clock accumulates about 8,760 hours a year, so 1.3 million hours is a substantial fleet history. Its current lineup has two principal models, rated at 300 MW and 78 MW.

    Because this is the company’s first overseas heavy-duty order, that record appears to have been built mainly in its home market. Domestic experience is real engineering evidence, but international buyers and lenders typically also look at availability rates, efficiency, emissions performance and how well a supplier supports equipment far from its home base. The release gives no figures on any of those, so the claim of parity with global leaders is the company’s own assessment for now.

    Southeast Asia as the Proving Ground

    Southeast Asia is a logical first market. Economies across the region are growing, electricity demand is rising, and several countries are adding gas-fired capacity as they move away from coal. Malaysia has also become a major destination for data center investment, adding to pressure on its power system, although nothing in the announcement ties the Samalaju plant to data center load.

    Shanghai Electric’s statement that developers in Indonesia, Thailand, the Philippines and Vietnam have shown strong interest should be read carefully: expressions of interest are not orders. Still, it points to a sales strategy of winning a reference project nearby and using it to win the next one. How smoothly Samalaju Unit 3 is built and runs will likely do more for that pipeline than any announcement.

    Background

    Shanghai Electric is a Chinese industrial group listed in Hong Kong (02727) and Shanghai (601727) that makes power generation equipment, including steam turbines, generators and gas turbines. Its heavy-duty gas turbine business has built a fleet of more than 100 units, mainly in its home market, and the Samalaju contract is its first heavy-duty gas turbine order outside China.

    The heavy-duty gas turbine market has long been dominated by GE Vernova, Siemens Energy and Mitsubishi Heavy Industries, whose order books have lengthened as demand for gas-fired generation has grown. Samalaju is an industrial area in the Malaysian state of Sarawak on the island of Borneo, where electricity demand from industry has driven power development.

    Sources

    Source: Shanghai Electric Secures First Overseas Heavy-Duty Gas Turbine Order for 500 MW Malaysian Project, Shanghai Electric’s announcement of its contract for Unit 3 of the Sarawak Samalaju combined cycle gas turbine project in Malaysia.

  • Fermi Founder Seeks Independent Strategic Review Before October 30 Vote

    Fermi Founder Seeks Independent Strategic Review Before October 30 Vote

    TL;DR · 30-second read

    The Short Version

    Fermi is a young company trying to build a huge private power and computing campus to run artificial intelligence systems. Its founder, Toby Neugebauer, was removed as chief executive in April but still owns about a fifth of the company.

    He has now formally asked the board to hire an outside bank to test whether selling the company, or bringing in a big partner, would be worth more than going it alone. Shareholders meet October 30.

    The stock has dropped about a fifth since April, a sign investors doubt the giant project will get built on schedule.

    Toby Neugebauer, co-founder and largest shareholder of Fermi Inc. (Nasdaq: FRMI), stepped up pressure on the company’s board ahead of its October 30 annual meeting. Neugebauer’s group, the Fermi Founder Parties, holds approximately 22% of shares outstanding. In a statement issued September 11 via PR Newswire, the group said it sent the board a letter on September 9 and a presentation on September 10. On September 10 it also submitted a shareholder proposal under SEC Rule 14a-8. The proposal asks the board to retain an independent, nationally recognized investment bank to evaluate extraordinary transactions, including a full-value sale or a majority recapitalization with a strategic partner, and to report the results to shareholders.

    Neugebauer also asked the board to voluntarily put two non-binding questions to shareholders. One would restore pre-IPO voting standards in place of a 70% supermajority adopted since April. The other would raise the company’s 2.5% ownership limit to 9.8%. The statement said Fermi shares hit an intraday low of $5.26 on September 10, down roughly 20% since Neugebauer was removed as CEO on April 17.

    Executive Summary

    The founder of an AI power-campus developer is using the formal shareholder-proposal process to force a question onto the agenda: is Fermi worth more as a standalone company or in a transaction? Neugebauer is careful to say he has never asked the board to sell. His proposal explicitly leaves the existing standalone plan in place and asks only that alternatives be tested against it by an independent banker. He also stated that his group is not soliciting proxies.

    The fight matters beyond one company. Fermi went public in October 2025 pitching Project Matador as a private utility, with its own power, transmission and water on a single scalable campus for AI compute. Neugebauer’s critique rests on specific gaps. The one disclosed tenant lease, 222 megawatts with TensorWave, is under 5% of Fermi’s stated 4.8 gigawatts of near-term power opportunity. As of September 10, he says, supporting guarantees and project financing for that lease had not been confirmed. And the company still has no permanent outside CEO.

    For investors in gigawatt-scale energy and data center developments, the episode illustrates how quickly confidence can erode. When projects are mostly still potential, the market reacts to leadership turmoil, unconfirmed financing and governance rules that limit shareholder influence.

    A Modest Ask With Pointed Timing

    Rule 14a-8 is the SEC mechanism that lets eligible shareholders place a proposal in a company’s proxy materials, the documents shareholders receive before voting. Such proposals are typically advisory. Even if one passes, the board is not legally required to act, though ignoring a strong vote carries reputational cost. The proposal is also narrowly drawn: hire an independent bank, test a sale or majority recapitalization against the standalone plan, and report back. A recapitalization means bringing in new capital that would change who controls the company. It is hard to argue against gathering that information in principle, which is likely the point.

    The timing is deliberate. Neugebauer paused an earlier proxy contest, meaning a campaign to win shareholder votes for his own positions or nominees. He now says he expected the board to finalize counterparty agreements his team had negotiated, on the terms agreed, and that execution has stagnated instead. Rather than restart a full contest, he is using lower-cost tools: a proposal, a public letter, and a request that the board voluntarily add advisory questions. That keeps pressure on without the expense and escalation of a formal vote-solicitation campaign, while positioning him to argue later that the board refused reasonable requests.

    The board’s side of the story does not appear in Neugebauer’s statement. The company has previously said it completed all five objectives of its 90-day plan, as of August 13. Shareholders will need its response to weigh the founder’s characterization of those milestones.

    222 Megawatts Against a 4.8-Gigawatt Story

    The core economic tension is scale. A megawatt is a unit of power capacity. A gigawatt is 1,000 megawatts, roughly the output of a large power plant. Fermi’s stated near-term power opportunity is 4.8 gigawatts, and its disclosed TensorWave lease is 222 megawatts, about 4.6% of that figure. For a company valued on the promise of a multi-gigawatt campus, the gap between contracted demand and stated potential is what investors are pricing.

    Financing is the second pressure point. Neugebauer says that as of the morning of September 10, Fermi had not confirmed the guaranteed agreement it previously said would support the TensorWave lease. He also says it had not confirmed that project financing was secured. In data center development, a signed lease is only as valuable as the tenant’s credit, or the guarantee standing behind it. That credit support is often what unlocks construction loans. Until those pieces are confirmed, the lease reads as a milestone rather than de-risked revenue.

    The stock-price figures in the statement also deserve a careful reading. Shares are down roughly 20% since April 17, 35% since July 2, and 20% since August 13. Taken together, those numbers imply shares traded higher in early July than on the day Neugebauer was removed. They also imply shares were near April levels by mid-August. The decline has not been a straight line from the leadership change, which complicates attributing it to any single board decision.

    Governance Rules as a Valuation Lever

    Two of Neugebauer’s requests concern who can own and influence Fermi. The first targets voting standards adopted since April. Amending bylaws now requires a 70% supermajority, and electing directors requires a majority of all outstanding shares. He wants to restore the standards discussed before the IPO: a majority of shares outstanding for bylaw changes, and a plurality of votes cast for directors. A plurality vote means the candidates with the most votes win, even without an outright majority. Higher thresholds make shareholder-initiated change harder to achieve, particularly when turnout is incomplete. Boards commonly defend such provisions as protection against a single large holder gaining effective control without paying other shareholders a premium. With a 22% holder in a dispute with the board, that argument is plausible here. Fermi’s own stated rationale is not included in the release.

    The second request concerns Fermi’s real estate investment trust (REIT) ownership limit. REITs are tax-advantaged property companies, and they must avoid concentrated ownership under federal tax rules. For that reason their charters typically cap how much any one holder can own. Neugebauer notes that peers such as Digital Realty, Equinix and Prologis use 9.8%, while Fermi’s general limit is 2.5%. A lower cap restricts how large a position any single institution can build. That can thin out the pool of large, long-term buyers, and some investors view that as a drag on valuation. Raising it would widen the potential shareholder base. Whether that would change the stock’s trajectory more than delivering contracts and financing would is an open question.

    Leadership Uncertainty and Counterparty Confidence

    Neugebauer argues that counterparties considering committing billions to Project Matador want to know who is running the company. After a four-month search, Fermi named board member Lee McIntire as CEO. Neugebauer contends that the choice showed no outside candidate would take the role, and says the company itself described McIntire as a temporary solution. The first point is his inference rather than an established fact; searches end with internal appointments for many reasons. The underlying concern is still reasonable. Large tenants, lenders and equipment suppliers weigh management continuity when signing long-dated commitments.

    Neugebauer’s own claims warrant scrutiny too. The contention that Fermi is a one-of-a-kind asset, and significantly undervalued, is an argument, not a demonstrated valuation. He says counterparties had agreed to terms while he led the company, but those terms have not been made public. His request that the board release all written communications with him since the IPO could clarify the record for both sides. His stake also gives him an obvious interest in a higher share price. That interest aligns with other shareholders on value, though not necessarily on control.

    Background

    Fermi Inc., doing business as Fermi America, went public on Nasdaq under the ticker FRMI on October 1, 2025. It was capitalized around Project Matador, which the company describes as a private utility campus. The campus would supply its own power, transmission and water for AI computing at scale, with the stated aim of avoiding added burden on public infrastructure and local ratepayers. The company is organized with a real estate investment trust ownership structure. It reports a near-term power opportunity of 4.8 gigawatts and a disclosed 222-megawatt lease with TensorWave.

    Co-founder Toby Neugebauer was removed as CEO on April 17, 2026. Since then, according to Neugebauer, the board adopted a 70% supermajority to amend bylaws and a majority-of-outstanding-shares standard to elect directors. On July 2 it removed the option of a dual-path strategic process, and on August 13 it announced completion of its 90-day plan. Neugebauer launched and later paused a proxy contest. He has also publicly criticized a $375 million convertible note offering, which he tied to former CFO Miles Everson’s resignation from the board. Board member Lee McIntire has since been named CEO.

    Sources

    Source: Fermi Founder Parties File Formal Proposal for Independent Strategic Review of Extraordinary Transactions, Press Board on Restoring Texas-Style Governance Ahead of October 30 Annual Meeting, a statement from Toby Neugebauer and affiliated entities on their shareholder proposal and governance requests to Fermi’s board.

    Primary sources: Fermi Founder Parties File Formal Proposal for Independent Strategic Review of Extraordinary Transactions, Press Board on Restoring Texas-Style Governance Ahead of October 30 Annual Meeting (PR Newswire, September 11, 2026).

  • Broadcom Stock Slips Despite Soaring AI Chip Outlook: Why Investors Hesitate

    Broadcom Stock Slips Despite Soaring AI Chip Outlook: Why Investors Hesitate

    TL;DR · 30-second read

    The Short Version

    Broadcom, one of America’s biggest chipmakers, reported sales that nearly doubled from a year earlier. It also predicted its artificial intelligence chip business will roughly quadruple by 2028. Its stock still slipped about 3%.

    One worry: much of that future growth depends on just three buyers. Those are Google and the chatbot makers OpenAI and Anthropic. If any of them spends less, Broadcom feels it fast.

    Broadcom’s chips power many of the artificial intelligence tools people now use. That makes its fortunes a rough gauge of whether the spending boom lasts.

    Broadcom reported fiscal third-quarter results on September 2 and filed its quarterly report (a 10-Q) with the SEC on September 10, covering the three months ended August 2, 2026. Morningstar reported that the shares fell 3% after the release, even though, by Morningstar’s calculation, the August-quarter results implied 86% year-on-year growth and October-quarter guidance implied 93%.

    Management also issued an outlook running through 2028. It calls for artificial intelligence revenue to quadruple from 2026 levels and for non-GAAP earnings per share, a measure that excludes items such as acquisition-related charges, to nearly triple to $30. Broadcom said OpenAI and Anthropic will become its two largest customers starting in 2027. It also pointed to a multiyear agreement with Google covering its custom AI chips.

    Executive Summary

    On paper, this was an exceptional report. Broadcom designs custom AI accelerators, which are chips built to one customer’s specification rather than sold off the shelf. It told investors that demand for them will keep compounding for at least two more years. The company supplies Google’s eighth-generation inference TPU, the tensor processing unit Google uses to run trained AI models for users. It expects enormous volume ramps at OpenAI and Anthropic. Morningstar, which rates the stock five stars with a $650 fair value estimate, said the 2028 guidance beat its own model.

    The muted stock reaction is the more interesting signal. Morningstar noted the shares had fallen nearly 30% since the previous quarter. It read that decline as the market pricing in a risk that Broadcom misses its 2028 targets or that growth slows afterward. Long-dated guidance, however bullish, asks investors to trust a handful of customers’ spending plans years in advance. For many investors, that trust now comes with conditions: which customers, on what terms, and at what profitability.

    For the wider infrastructure market, Broadcom’s outlook is a direct read on how aggressively the largest AI developers intend to build out computing capacity. It also shows how much of that capacity will run on custom silicon rather than general-purpose GPUs.

    Great Numbers, Skeptical Market

    Broadcom’s guidance is unusually specific and unusually far out. Most chip companies guide one quarter ahead. Broadcom put figures on 2028: AI revenue up fourfold from 2026 and roughly $30 of non-GAAP earnings per share. Morningstar calls the guidance credible and argues underlying AI infrastructure demand will keep rising through 2030. The market’s reaction was more cautious. A 3% post-earnings dip followed a slide of nearly 30% since the prior quarter.

    That gap between guidance and share price reflects a familiar tension in AI hardware. Investors reward growth, but they discount growth that depends on a few enormous purchase decisions. When a company’s outlook already assumes a quadrupling, even strong confirmation can fail to move the stock. The debate shifts from whether demand exists to whether it holds beyond the guided window, and at what price.

    Morningstar’s valuation shows how much future growth is embedded in its $650 estimate. That figure implies adjusted price-to-earnings multiples of 56 times fiscal 2026 and 34 times fiscal 2027, falling to 22 times on 2028 estimates. Put simply, the estimate looks reasonable only if the 2028 earnings arrive.

    Diversifying Away From Google, Into Frontier Labs

    Google has long been Broadcom’s anchor customer in custom accelerators. Two developments this quarter complicate that picture. First, Google split its TPU program: Broadcom supplies the inference chip, while MediaTek won the training chip, which is used to build the models in the first place. Morningstar argues the inference part is higher in volume and complexity. Still, the split shows that even Broadcom’s largest customer is willing to multisource.

    Second, Broadcom expects OpenAI and Anthropic to overtake Google as its biggest customers from 2027. That diversifies Broadcom away from a single buyer. It also shifts exposure toward frontier-model developers, whose infrastructure budgets depend on how their own businesses and funding evolve. Morningstar acknowledges investors may worry about this spending but argues it is supported by improving AI model economics. That claim will be tested over the next two years rather than settled by this report.

    Morningstar assigns Broadcom a High Uncertainty Rating and names customer concentration as its primary risk. The AI accelerator business rests on a handful of high-spending customers. Shifts in any one customer’s spending patterns can swing results and sentiment.

    The Margin Question Behind the Growth

    Broadcom is known for profitability. Morningstar credits it with free cash flow margins consistently above 40%, meaning over 40 cents of every revenue dollar becomes cash after capital spending. It projects free cash flow exceeding $50 billion a year by fiscal 2027. Much of that reputation was built on a mix of networking chips and infrastructure software, including VMware, acquired in 2023.

    As custom accelerators grow to dominate revenue, investors will want to know whether profitability holds as the mix changes. Custom chips are negotiated with a few very large buyers who have real bargaining power, as Google’s decision to bring in MediaTek suggests. Morningstar’s bear case also flags moderating growth in Broadcom’s legacy software lines. That makes the economics of the AI business more important to the overall margin picture, not less.

    None of this contradicts the demand story. It explains why demand alone may no longer be enough: investors are asking what each dollar of AI revenue is worth, not just how many dollars are coming.

    Supply, Debt and Leadership

    Broadcom depends heavily on TSMC to manufacture its chips, so any capacity constraint could limit how quickly it ships to customers. Morningstar believes Broadcom would be a priority TSMC customer because of its scale and long relationship. Even so, a quadrupling of AI revenue requires a matching step-up in manufacturing allocation.

    The balance sheet looks manageable. As of October 2025, Broadcom held $16 billion in cash against $65 billion of gross debt, roughly half tied to VMware. Its gross debt-to-adjusted EBITDA ratio fell from 3.5 times after the deal to 1.5 times by the end of fiscal 2025. EBITDA stands for earnings before interest, taxes, depreciation and amortization. Leadership is a longer-term consideration. Morningstar identifies CEO Hock Tan as integral to Broadcom’s acquisition and operating strategy; he is contracted with the company through 2030, within the window in which the AI ramp is meant to play out.

    Background

    Broadcom is a semiconductor and infrastructure software company led by CEO Hock Tan, who has built it largely through acquisitions run for high efficiency. The largest recent deal was the 2023 purchase of virtualization software maker VMware. Broadcom reports revenue from products and from subscriptions and services, across the Americas, Asia Pacific and EMEA. In recent years its growth engine has become custom AI accelerators: chips it co-designs with large technology companies, beginning with Google’s TPU program.

    Custom silicon has become a major alternative to off-the-shelf GPUs for companies spending heavily on AI computing. Its appeal is performance and cost tuned to one customer’s workloads. Broadcom’s traditional networking chips and VMware software still generate substantial cash flow, but the AI chip business is now the primary driver of its valuation. Its results are therefore closely watched as a barometer for AI infrastructure spending.

    Sources

    Source: Broadcom Stock Fell 3% After Earnings Despite Immense AI Chip Demand (Morningstar), Morningstar’s analysis of Broadcom’s fiscal third-quarter results and 2028 outlook.

    Primary sources: Broadcom Inc. Form 10-Q for the quarter ended August 2, 2026 (SEC EDGAR, filed September 10, 2026).

  • Nebius–Palantir Deal: A Neocloud Moves Inside the Software Perimeter

    Nebius–Palantir Deal: A Neocloud Moves Inside the Software Perimeter

    TL;DR · 30-second read

    The Short Version

    Palantir, a company that sells data-analysis software to large organizations, has picked Nebius, a company that rents out powerful computers for artificial intelligence, as its preferred partner for running that work.

    The pitch: businesses can train their own artificial intelligence on their own private data, using machines they control, instead of handing that data to a big tech provider.

    Nebius’s stock jumped on the news. Neither company has said how much money is involved or exactly when it starts.

    On September 8, 2026, Palantir Technologies and Nebius Group announced a strategic partnership under which Palantir named Nebius its preferred sovereign AI infrastructure partner. According to the joint press release, which Nebius furnished to the SEC in a Form 6-K the same day, Palantir will bring Nebius compute and inference endpoints inside the Palantir enterprise perimeter after an integration period. Eligible Palantir commercial customers will then be able to run and adapt open AI models on Nebius infrastructure. The companies also said they will work together to deploy new compute capacity for Palantir customers, including modular data centers at sites where power is already available.

    Seeking Alpha reported that Nebius shares rose about 8% following the announcement, bringing their year-to-date gain to roughly 175%. Neither company disclosed financial terms, capacity commitments, or a timeline.

    Executive Summary

    The Nebius–Palantir partnership pairs a neocloud with an enterprise software platform. A neocloud is a newer cloud provider built mainly to rent out GPU capacity for AI. The deal gives Nebius something pure infrastructure providers often lack: a direct channel into large commercial organizations that already run Palantir’s Foundry and AIP software. For Palantir, it adds a named compute layer to what it calls its Sovereign AI Operating System. That system is pitched to customers who want to train models on proprietary data without giving up control of the data, the models, or the hardware they run on.

    The announcement matters because it tests a central question for GPU capacity providers: can they move beyond selling compute by the hour and capture a share of the higher-margin software relationship? The release answers only part of that question. Nebius gains preferred status and placement inside Palantir’s environment. Palantir, however, keeps the authorization, isolation, and data layer, which is the part customers interact with day to day. How revenue is split, and how much capacity is involved, has not been disclosed.

    Distribution, Not Just Capacity

    Most neoclouds compete on how quickly they can bring GPUs online and how cheaply they can rent them. That is a hard business to differentiate. Hardware is largely sourced from the same supplier, and large AI labs negotiate aggressively on price. The Palantir arrangement gives Nebius a different kind of asset: placement inside a software platform that enterprise buyers have already approved, integrated, and budgeted for. In enterprise IT, getting onto an approved-vendor list can take longer than building a data center. Being embedded in a platform the customer already trusts shortens that path considerably.

    The joint release frames Palantir’s choice around Nebius being built for AI from the ground up, with infrastructure and software designed together, rather than adapted from general-purpose computing. That claim reflects how the partners describe their choice. It is not an independent benchmark. It does, however, match the positioning neoclouds have used against the large hyperscalers (Amazon Web Services, Microsoft Azure, Google Cloud), whose AI offerings sit alongside broad general-purpose cloud businesses.

    Who Captures the Software Value?

    The editorial question is whether this deal moves Nebius up the stack. The honest answer is: partially. Under the announced structure, Palantir’s AIP, Ontology, Foundry, and Apollo provide what the release calls the authorization and isolation layer. That is the governance and data-modeling software that decides who can see what and how models are applied to business operations. Nebius supplies the compute and inference endpoints underneath. Inference endpoints are the services that run a trained model and return its answers. In that arrangement, Palantir owns the customer-facing software relationship, and Nebius becomes a preferred engine inside it.

    That still has value for Nebius. Inference is recurring, usage-driven demand. It tends to be more durable than one-off training runs, and enterprise customers who fine-tune open models on their own data will need to keep running them. But the term used is “preferred,” not “exclusive.” Nothing in the release prevents Palantir from adding other infrastructure partners later. The release also does not indicate whether Nebius bills customers directly or is paid through Palantir. That distinction largely determines how much of the economics Nebius actually captures.

    Sovereignty as a Commercial Product

    “Sovereign AI” has typically meant national governments wanting AI infrastructure under domestic control. This announcement applies the idea to companies instead. The release describes offering commercial organizations that have not previously had a sovereign option a way to run their own models on trusted infrastructure. In practice, that means an enterprise can take an open model, whose weights are publicly available, adapt it with proprietary data, and keep both the data and the resulting model. The alternative is sending that data to a closed model hosted by a third party.

    The release explicitly targets Palantir’s commercial customers. It does not mention government or defense workloads, even though Palantir has deep roots in government. Any expansion into public-sector use would likely require separate security accreditations. Palantir CEO Alex Karp said Palantir’s ontology and Nebius’s infrastructure “will undergird the sovereignty our partners are demanding.” Nebius CEO Arkady Volozh said organizations need “both the performance of large-scale AI infrastructure and control over their data and models.” The release also claims that domain-adapted open models can outperform general-purpose closed models for specific uses. That is plausible for narrow tasks, but it is an assertion rather than a demonstrated result.

    Modular Builds and the Power Question

    The most operationally concrete line in the announcement concerns speed to capacity. The companies said they will use modular data-center deployments at sites where power is already available. Modular data centers are prefabricated units that can be installed far faster than a conventional purpose-built facility. The emphasis on existing power reflects a constraint across the industry: grid connections for large new loads can take years to secure. Sites with power already in place are among the scarcest assets in AI infrastructure.

    This approach reduces some risks and introduces others. Modular deployments can serve customers faster, but they are typically smaller than campus-scale builds. They also depend on securing GPUs, cooling, and network connectivity at each location. The release names no sites, megawatt figures, or deployment schedule. Until those details appear, the capacity commitment is best read as a stated intention, not a quantified plan.

    Background

    Nebius Group N.V. is an Amsterdam-headquartered AI cloud company listed on Nasdaq under the ticker NBIS. It emerged from the 2024 restructuring of Yandex N.V. after that company divested its Russia-based businesses. Founder and CEO Arkady Volozh co-founded Yandex. Nebius describes itself as building a full-stack platform for AI development, from data and model training to production deployment. It is part of the neocloud cohort of GPU-focused providers that has grown rapidly alongside demand for AI compute. As a foreign private issuer, it reports material events to the SEC on Form 6-K.

    Palantir Technologies, listed on Nasdaq as PLTR, built its reputation on data-integration software for government and defense agencies and has since expanded into commercial markets. Its Foundry, Ontology, Apollo, and AIP products let organizations model their operations, manage data access, and apply AI to business workflows. The Nebius partnership extends that software into a named compute layer that Palantir markets as sovereign AI.

    Sources

    Source: Nebius And Palantir: Implications Of The New Partnership (NASDAQ:NBIS) — Seeking Alpha, analysis of the Nebius–Palantir AI infrastructure partnership and the market reaction.

    Primary sources: Nebius Group N.V. Form 6-K filed September 8, 2026 (SEC EDGAR); Exhibit 99.1: Palantir and Nebius partner to deliver a complete sovereign AI stack to Palantir customers (September 8, 2026).

  • Applied Digital’s Harwood, ND AI Data Center (Polaris Forge 2): Site, Power, Cost

    Applied Digital’s Harwood, ND AI Data Center (Polaris Forge 2): Site, Power, Cost

    TL;DR · 30-second read

    The Short Version

    Two buildings north of Fargo, North Dakota are nearly finished, and together they are bigger than the largest building in the state, an Amazon warehouse two miles away. About 1,800 workers are on site.

    The buildings will hold computers that run artificial intelligence, and they will draw about as much electricity as a small city. The local council approved the project in September 2025, and part of it is supposed to be running by the end of this year.

    The company building it, Applied Digital, has not said which businesses will rent the space.

    InForum reported that Applied Digital’s Polaris Forge 2 campus in Harwood, North Dakota is nearing completion, with roughly 1,800 workers finishing two buildings of 907,000 square feet each. The Harwood City Council approved the $3 billion project on September 10, 2025. Part of the campus is on pace to become operational by the end of 2026, with the remainder opening in early 2027, according to Nick Phillips, Applied Digital’s executive vice president of public affairs and real estate acquisition.

    The site is a 280-megawatt facility drawing power from Cass County Electric, and Applied Digital is paying $112 million for the substation and transmission upgrades that serve it. The company has not disclosed which companies will occupy the buildings. It expects 200 to 250 permanent employees once the campus opens, on 400 acres of which about 300 have been graded.

    Executive Summary

    Harwood is Applied Digital’s third North Dakota project, after Jamestown and the Polaris Forge 1 campus at Ellendale, with a fourth under construction near Center. Ellendale, announced in late 2024, is partially operational. The pattern across all four is the same wager: build large blocks of power-dense capacity in a cold, cheap-power state and race to energize them while demand for artificial intelligence compute outruns supply.

    What makes Harwood worth studying is the calendar. From a municipal approval on September 10, 2025 to first operational capacity at the end of 2026 is roughly fifteen months for a 280-megawatt campus. In a market where developers routinely quote three to five years between site control and delivered power, that interval is the entire product. “Time to power” is the industry’s shorthand for how long a tenant waits between signing and switching on machines, and it has become the variable that decides which sites win contracts.

    The unresolved half of the story is commercial rather than physical. Steel, generators, cooling loops and a substation are visible and verifiable. Leases are not. Applied Digital has not named a tenant for Harwood, has not disclosed how much of the 280 megawatts is contracted, and has not detailed how the $3 billion build is financed. Those omissions do not undermine the construction progress, but they determine whether the speed converts into revenue.

    Fifteen Months From Vote to Volts

    The construction detail in the account is mostly a lesson in sequencing. Eight generator rooms in each building hold twelve units apiece, which is 96 generators per building and 192 across the campus, a level of on-site backup that signals a design intended for tenants who will not tolerate interruption. Roughly 86,000 electrical connections and several miles of copper cable have to be terminated and tested before any of it carries load. Workers wear helmets with augmented reality systems that check whether components are installed in the right place and whether crews are tracking to schedule, which is less a gadget than an admission that verification, not fabrication, is the thing most likely to slip.

    Phillips describes the split inside each building as about 25% data halls and 75% supporting systems. That ratio is the clearest single indicator of how the economics have shifted. In a conventional enterprise facility, the computer room dominates the floor plan. In a building designed for dense artificial intelligence racks, the electrical rooms, switchgear, uninterruptible power supply gear and cooling plant swallow three-quarters of the envelope. Applied Digital is not really building a warehouse for servers. It is building a power conversion and heat rejection plant that happens to have servers in the middle of it.

    The arithmetic underlines the point. If 280 megawatts is measured at the computers, spread across roughly 453,000 square feet of data hall, that is on the order of 600 watts per square foot, an order of magnitude above the densities most existing commercial data centers were designed for. Applied Digital has not specified whether the 280-megawatt figure is the load delivered to equipment or the total drawn from the utility, and the distinction changes that number materially. Either way, the campus is firmly in the density range that only recently became buildable at scale.

    The $112 Million Sentence

    The most consequential figure in the account is not the $3 billion headline but the $112 million Applied Digital says it will pay for the substation and transmission upgrades. That is roughly $400,000 per megawatt of connection, and it is the developer, not the utility’s other customers, footing the bill. Grid connection cost allocation is the live political question in nearly every jurisdiction now absorbing large computing loads, because the alternative structure, in which network upgrades are socialized across the rate base, is what makes residents worry about their bills. Harwood drew protesters on exactly that concern.

    Applied Digital’s counter-claim is that its facilities do not raise electricity bills, and Phillips cited more than $46 million in savings for customers across North Dakota and South Dakota since the Ellendale project began. The underlying mechanism is plausible in principle: a very large, very steady customer spreads a utility’s fixed costs across more kilowatt-hours, which can lower the per-unit cost everyone pays. But the claim as stated is not yet verifiable by a reader. It does not identify which utility’s customers, over which period, measured against which counterfactual, or whether the figure is an avoided increase rather than a reduction. Those are answerable questions, and regulated utilities generally have to answer them in rate proceedings.

    It is worth being precise about what each side is arguing, because they are not arguing about the same thing. The savings claim is retrospective and concerns Ellendale. The objection at Harwood is prospective and concerns what happens when a second large load lands on a different distribution system. A past result under one set of contracts does not settle the future question, and neither does a protest settle whether the mechanism works. The $112 million commitment is the strongest evidence Applied Digital has offered that it intends to carry its own interconnection costs, which is the part of the dispute that can actually be documented.

    Air, Not Water, and Why North Dakota Rewards That Choice

    Cooling is where siting logic becomes visible. Branden Cummins, the company’s civil structure architecture superintendent, described a closed loop that circulates fluid through six data halls in each building, carries it to the roof, and uses large fans to shed the heat before recirculating it. Because the loop is sealed, Applied Digital says water consumption will be comparable to a household. That is a meaningful claim in an industry now facing real scrutiny over evaporative cooling, which chills equipment by boiling away water and can consume millions of gallons a year at this scale.

    The trade is not free. Rejecting heat to air rather than to evaporating water costs electricity for fans, and the penalty grows with ambient temperature. In a Texas or Arizona summer, that penalty is severe enough that many operators accept the water consumption instead. In eastern North Dakota, where cold hours dominate the year, the fan energy penalty is small and the water savings are close to pure gain. The closed loop is not primarily an environmental gesture; it is the design that the climate makes cheapest, and the environmental benefit follows from the geography.

    Noise and setback receive the same treatment. Silencing equipment and sound barriers on the generator plant, with the nearest home about half a mile away and the town of Harwood more than a mile off, reflect a siting choice made before mitigation was needed. Developers increasingly buy distance rather than engineer around its absence, because generator testing and rooftop fan arrays are the two complaints most likely to follow a facility through its operating life.

    A Builder Without a Named Tenant

    Applied Digital is a merchant developer, meaning it builds capacity and then leases it, as opposed to a hyperscaler building for its own use. That model turns on a single question: how much of the capacity is committed before it is energized. The company has not disclosed which companies will occupy Harwood. Until it does, an observer cannot distinguish between a campus built against signed leases and one built on conviction about demand.

    The local economics are more legible. Roughly 1,800 construction workers will give way to 200 to 250 permanent employees, a ratio close to eight to one that is typical of the sector and worth stating plainly: data centers are enormous construction projects and modest employers. The durable local benefit is tax base rather than payroll. Polaris Forge 1 at Ellendale paid about $4.3 million in property tax for 2026, Harwood is receiving no tax incentives or abatements according to the company, and Phillips estimated Applied Digital would become the largest single taxpayer in the state by 2029, while acknowledging that Harwood’s own assessment is not yet estimated because the building is not finished.

    The absence of abatements is the notable item. Large data centers are frequently recruited with property tax exemptions that hollow out the fiscal case communities were sold, and a project that forgoes them changes the arithmetic for the county. It also suggests the developer’s binding constraint was schedule rather than subsidy. When capacity delivered in 2026 is worth materially more than the same capacity delivered in 2029, a year of negotiation over abatements costs more than the abatements are worth.

    Background

    Applied Digital is a Texas-based developer that builds and leases data center capacity rather than operating computing services itself, a model the industry calls merchant development. Its North Dakota expansion began with a facility in Jamestown, followed by the Polaris Forge 1 campus at Ellendale announced in late 2024 and now partially operational, the Harwood project approved in September 2025, and a fourth site under construction near Center, about 40 miles northwest of Bismarck.

    The expansion tracks a broader shift in where computing gets built. Training and running large artificial intelligence models requires racks that consume several times the power of conventional servers, which has turned electrical capacity, rather than fiber or floor space, into the binding constraint on the industry. Developers have responded by moving to regions with available generation, cold climates that reduce cooling costs, and utilities able to energize large loads quickly. That has brought industrial-scale construction to rural communities with no prior exposure to it, along with recurring local debates over electricity rates, noise, water and tax base that Harwood has seen in full.

    Sources

    Source: An inside look at the Harwood data center as it nears completion — InForum’s report from Applied Digital’s Polaris Forge 2 construction site in Harwood, North Dakota, with detail on schedule, power, cooling, staffing and taxes from company executives.

  • Hydro One Files $1.3B North Shore Link With Ontario Regulator

    Hydro One Files $1.3B North Shore Link With Ontario Regulator

    TL;DR · 30-second read

    The Short Version

    Ontario’s largest electricity company wants to string a new high-voltage power line across the province’s north, from a station near Wharncliffe to the city of Sault Ste. Marie. It would cost $1.3 billion and switch on in 2029. The provincial regulator has to approve it first.

    Nine First Nations have been offered the chance to own half of the line itself, not merely to be consulted about it.

    Why an ordinary person should care: factories, mines and computing sites cannot open where there is no spare electricity. This line is being planned before those customers show up.

    Hydro One Networks Inc. and Hydro One Sault Ste. Marie LP filed a joint leave-to-construct application with the Ontario Energy Board on September 11, seeking approval to build the North Shore Link, according to the company’s announcement. The project is a new 230-kilovolt transmission line running from Mississagi Transformer Station near Wharncliffe to the Third Line Transformer Station in Sault Ste. Marie, which Hydro One Sault Ste. Marie LP owns.

    Hydro One put the investment at $1.3 billion and said the line is expected to be in service in 2029. The work also includes upgrades at Mississagi and an expansion of nearby Algoma Transformer Station, with the Sault Ste. Marie partnership handling upgrades at Third Line. Nine potential First Nation partners have been offered the opportunity to take a 50 per cent equity stake in the transmission line component through Hydro One’s First Nation Equity Partnership Model.

    Executive Summary

    A leave-to-construct application under Section 92 of the Ontario Energy Board Act is the gate every transmission project in the province has to pass through. It obliges the applicant to lay out route, design, timing and cost, and it puts the need for the line in front of a regulator, intervenors and ratepayer advocates. Hydro One has now placed a $1.3 billion northern reinforcement into that process.

    The stated rationale is broad rather than specific. Ryan Docherty, Hydro One’s executive vice president for capital portfolio delivery, framed the project around rising provincial demand, aging infrastructure and more frequent severe weather, and described the line as strengthening the link between Ontario’s northeast and northwest. No customer, industrial project or contracted load appears in the announcement.

    That is the interesting part. Transmission takes the better part of a decade from concept to energization, while a large industrial or computing load can commit to a site in months. A regulator that waits for signed commitments before approving wires guarantees the wires arrive late. A regulator that approves ahead of them puts capital at risk of sitting underused. The North Shore Link asks the Ontario Energy Board to make that call on a $1.3 billion scale.

    Building Wires Before the Load Signs

    The economics of transmission run backwards from the economics of the customers it serves. A 230-kilovolt line, meaning a circuit that moves power at 230,000 volts so it can carry bulk energy over long distances with modest losses, is a multi-year construction project with a multi-decade asset life. Hydro One’s own timeline puts the North Shore Link in service in 2029, roughly three years after filing, and that clock only starts once the regulator says yes. Large electricity consumers do not work on that horizon. They pick a site, want power within a couple of years, and go elsewhere when the answer is a queue position.

    This is why the need case in a filing like this one tends to rest on system arguments rather than customer arguments. Hydro One’s announcement cites demand growth, asset age and storm resilience, and says the line strengthens the northeast-to-northwest connection. Those are real engineering rationales and they stand on their own. They are also, notably, rationales that do not require anyone to have signed a connection agreement.

    The honest framing is that the filing is a bet on capacity being wanted rather than proof that it is already claimed. That bet can be entirely sound. Ontario has publicly signalled load growth across electrification, industry and computing, and pre-building headroom is cheaper than retrofitting a constrained corridor later. But it is a bet, and the Section 92 proceeding is precisely where the strength of the underlying forecast gets tested by parties whose job is to protect ratepayers.

    Fifty Per Cent, Nine Nations

    The equity structure is the most concrete commitment in the announcement. Nine potential First Nation partners have been offered the chance to invest in a 50 per cent stake in the transmission line component of the project, through a model Hydro One has applied to other northern lines. This is ownership, with the returns that regulated transmission assets generate, rather than a consultation process or a community benefits fund.

    For a project developer, the practical value is schedule. Linear infrastructure across northern Ontario crosses traditional territories, and objections surfacing late in permitting are among the most reliable ways to lose a year. Aligning economic interest at the planning stage does not eliminate that risk, but it changes the negotiating posture from opposition to partnership on projects where the model takes hold.

    Two things are worth noting evenly. The word Hydro One uses is potential, and the offer covers the transmission line component rather than the station work. Whether all nine participate, on what financing terms, and how the equity is funded are not addressed in the announcement. Those are ordinary details for a structure at this stage, and they are also the details that determine whether the 50 per cent figure translates into real returns for the communities involved.

    What This Means for Anyone Shopping for Power in Algoma

    For site selectors, whether in steel, mining, manufacturing or compute, the useful signal is not the dollar figure. It is the direction of capacity in the Sault Ste. Marie area and the date. A reinforced 230-kilovolt path into Third Line Transformer Station, plus expanded capability at Algoma Transformer Station, changes what a region can physically absorb. Until the line energizes, the constraint is whatever exists today.

    That produces an awkward window. Anyone evaluating a northern Ontario site now is evaluating against a grid that gets materially better in 2029 if the regulator approves and construction holds, and does not if either slips. Buyers negotiating connection today should be asking how their service depends on this project and what happens to their in-service date if the proceeding runs long or the build encounters the transformer and equipment lead times that have stretched across the industry.

    For investors, the read is simpler. Hydro One reported $39.7 billion in assets and $9 billion of revenue for 2025, and invested $3.4 billion in its networks that year. The North Shore Link sits alongside filings the company made earlier in 2026 for projects in central, eastern and Ottawa-area Ontario. This is a regulated utility adding to a rate base through a steady cadence of approvals, and the pace of that cadence matters more to the equity story than any single line does.

    Background

    Hydro One Limited, listed on the Toronto Stock Exchange under the ticker H, is Ontario’s largest electricity transmission and distribution provider, serving 1.5 million customers through its subsidiaries. Its transmission business owns the high-voltage backbone that moves power across a province stretching more than a thousand kilometres from the Manitoba border to the St. Lawrence, including the long, sparsely populated northern corridors that connect the northwest to the load centres of the south. Hydro One Sault Ste. Marie LP is the Hydro One entity that owns Third Line Transformer Station, the northern end point of the proposed line.

    Every transmission expansion in Ontario runs through the Ontario Energy Board under Section 92 of the Ontario Energy Board Act, 1998. The proceeding is adversarial by design: the applicant files route, design, timing and cost, and intervenors including consumer groups and industrial customers can challenge whether the investment is needed and whether the cost is prudent. Sault Ste. Marie sits in the Algoma district, a northern industrial region where heavy industry, mining and, increasingly, electrification of everything from transport to industrial heat all compete for the same finite grid capacity.

    Sources

    Source: Hydro One seeks approval from the Ontario Energy Board to build the North Shore Link — Hydro One’s September 11, 2026 announcement of a joint leave-to-construct application for a $1.3 billion, 230-kilovolt line into Sault Ste. Marie, targeted for service in 2029.

  • TSMC Reportedly to Double CoWoS Packaging Capacity by 2028

    TSMC Reportedly to Double CoWoS Packaging Capacity by 2028

    TL;DR · 30-second read

    The Short Version

    The chips that run artificial intelligence are not held up by the factories that print them. They are held up by the assembly step at the end, where a processor and its memory are bonded onto a single slab so they can talk to each other fast enough.

    Taiwan Semiconductor Manufacturing Company owns most of the world’s capacity for that step. Analysts in Taiwan say it plans to roughly double that capacity by 2028. Until it does, anyone building an artificial intelligence data center is standing in a queue. Intel is selling a rival method, and smaller assembly firms are absorbing the overflow.

    Wccftech, citing a Taiwanese media report that quotes analysts, said TSMC plans to expand its chip-on-wafer-on-substrate (CoWoS) advanced packaging capacity to 260,000 wafers per month by the end of 2028, roughly double the approximately 130,000 wafers per month the same forecast puts it at by the end of 2026. The expansion is said to center on TSMC’s Arizona campus and its AP7 packaging facility in Taiwan.

    The same forecast estimates that Intel’s competing EMIB-T packaging, converted into CoWoS-equivalent 12-inch wafer terms, could reach 15,000 to 20,000 wafers per month in 2027 and 40,000 to 45,000 in 2028, and says the CoWoS shortage has pushed work toward other packaging firms including Amkor, UMC and ASE. TSMC has not published a capacity target of its own. In a Form 6-K filed with the Securities and Exchange Commission on September 10, the company reported August 2026 consolidated revenue of NT$514.81 billion, up 53.3 percent from a year earlier.

    Executive Summary

    Advanced packaging is the step where a logic die and its stacks of high-bandwidth memory are mounted together onto a shared substrate so they behave as one component. CoWoS is TSMC’s version, and it is the packaging used for the accelerators that fill AI data centers. For roughly three years it, rather than leading-edge wafer manufacturing, has been the narrowest point in the supply chain for those parts.

    A doubling of that capacity by 2028 would be the single largest change to AI hardware availability that any one company can make. It would not, however, double accelerator output in a straight line. Package sizes are growing, the newest designs consume more interposer area and more memory stacks per unit, and a wafer of packaging substrate therefore yields fewer finished accelerators each generation than it did the year before. Capacity that doubles against demand that also rises per unit is not the same as supply that doubles.

    TSMC’s own disclosure does not address packaging capacity. What its September 10 filing does show is demand running hot: August revenue rose 53.3 percent year over year, and revenue for the first eight months of 2026 was up 39.3 percent. That is consistent with a company selling everything it can finish, which is the condition under which a capacity forecast of this kind becomes credible without being confirmed.

    The Bottleneck Moved Downstream

    For most of the history of the chip industry, the scarce resource was the front end: photolithography tools, leading-edge wafer starts, yield on a new process node. Packaging was the cheap, unglamorous back end. AI accelerators inverted that. A modern accelerator is not one chip but an assembly, a large logic die surrounded by stacks of memory, all sitting on a silicon interposer threaded with vertical connections called through-silicon vias. The density of those connections is what delivers the memory bandwidth the workload needs. That assembly step is CoWoS, and there is no abundant substitute for it.

    The practical consequence for anyone financing a data center is that wafer capacity is no longer the number to watch. A foundry can start more wafers than it can finish into packages, and unfinished silicon does not train models. The forecast of 130,000 wafers per month at the end of 2026 rising to 260,000 by the end of 2028 is, in effect, a two-year schedule for how quickly the industry can convert build-ready halls into loaded racks.

    Read it with care, though. Interposer area per accelerator has been climbing with every generation as designers add memory stacks and reticle-stitched logic. If the average package consumes meaningfully more of a wafer in 2028 than it does today, unit output grows by considerably less than the wafer count suggests. Doubling the input is a ceiling on supply growth, not a promise of it.

    Arizona Can Print the Chips but Cannot Finish Them

    The most operationally interesting detail in the forecast is geographic. Wafers produced at TSMC’s Arizona campus are flown to Taiwan to be packaged, because the packaging capacity sits on the island. Siting part of the expansion in Arizona alongside the AP7 facility in Taiwan would begin to close that loop, turning an American fab from a wafer source into something closer to a finished-component source.

    That matters for reasons beyond logistics. A round trip across the Pacific adds days of cycle time and a customs and freight cost to every high-value wafer, and it means that the resilience argument for building fabs outside Taiwan is currently incomplete. A domestic fab whose output must still transit Taiwan to become a usable product does not remove the concentration risk that justified it. TSMC’s own filing shows the Arizona entity is treated as a significant subsidiary, with parent guarantees extended to TSMC Arizona among other affiliates, but the filing says nothing about packaging there.

    For data center operators and the governments underwriting them, this is the question worth asking of any onshoring announcement: does the site produce a wafer or a shippable part? Until packaging follows fabrication, the answer for Arizona remains the former.

    Intel’s EMIB-T and the Value of a Second Source

    Intel’s EMIB-T takes a different architectural route. Instead of building the whole assembly on a large silicon interposer, it embeds small silicon bridges inside the organic substrate and places the logic and memory dies on top. Less silicon is used, which can mean lower cost and fewer of the handling and warpage problems that come with very large interposers, but the approach historically trades away some of the interconnect density that the largest, most bandwidth-hungry designs demand.

    The forecast puts EMIB-T at 40,000 to 45,000 CoWoS-equivalent wafers per month by 2028. Against 260,000, that is a meaningful second source rather than a competitive threat, on the order of a sixth of the leader’s volume. The claim that it suits custom accelerators of the kind Google and Amazon design is plausible on architecture, since in-house cloud silicon is generally tuned for a known workload and a known cost target rather than for maximum bandwidth per package. It is also unverified: neither Intel nor any cloud provider has announced such an arrangement.

    The spillover to Amkor, UMC and ASE follows the same logic. When the preferred supplier is sold out, work migrates to whoever can perform an adjacent version of it, and outsourced assembly and test firms have been the natural landing place. Buyers get optionality and a check on pricing. The risk is that second-source packaging is not a drop-in substitute. Requalifying a part on a different packaging architecture is an engineering program, not a purchase order, and the customers best placed to do it are the ones designing their own silicon.

    What Is Substantiated and What Is Not

    The demand side of this story rests on disclosed figures. TSMC’s September 10 filing with the Securities and Exchange Commission reports August consolidated revenue of NT$514.81 billion, a 10.1 percent increase on July and 53.3 percent on August 2025, with the first eight months of 2026 up 39.3 percent year over year. Growth at that rate at TSMC’s scale is difficult to explain without sustained AI accelerator demand behind it.

    The supply side is a forecast, not a commitment. The 260,000 figure, the 130,000 baseline, the Arizona and AP7 split and the EMIB-T range all come from analysts rather than from the companies. None of the three firms named has published a packaging capacity target, a capital budget attached to one, or a schedule. Analyst capacity models in this industry have a reasonable track record on direction and a poorer one on timing, because packaging expansions depend on tool delivery, cleanroom construction and qualification cycles that slip quietly.

    The sober reading is that the direction is well supported and the magnitude is an estimate. Treat 2028 as the earliest plausible date for materially looser accelerator supply, treat the doubling as a ceiling rather than a plan of record, and treat any procurement model that assumes it as carrying schedule risk that TSMC has not underwritten.

    Background

    TSMC is the world’s largest contract chip manufacturer, building processors designed by other companies. It commercialised CoWoS in the 2010s as a niche technique for high-performance parts, long before AI accelerators turned it into the scarcest resource in the industry. Today the packaging step determines how many finished accelerators reach the market, because a logic die without its memory bonded alongside it is not a usable product.

    The competitive response has come from two directions. Intel has pushed its own bridge-based packaging as part of its foundry ambitions, targeting customers who want a supplier outside Taiwan or a lower-cost assembly for custom silicon. Outsourced assembly and test firms such as Amkor, ASE and UMC have absorbed work that TSMC cannot schedule. None of this has yet changed the basic shape of the market, in which one company’s packaging line sets the pace at which AI data centers can be filled.

    Sources

    Source: Report: TSMC to Double CoWoS Capacity by 2028 as AI Chip Shortage Spills Over to Rivals, which relays a Taiwanese analyst forecast for TSMC’s advanced packaging capacity and for Intel’s competing EMIB-T volumes.

    Primary sources: Taiwan Semiconductor Manufacturing Company, Form 6-K filed September 10, 2026 (August 2026 revenue report), filed with the US Securities and Exchange Commission.

  • Court Rejects Coal Plant Order, Narrowing DOE Emergency Power

    Court Rejects Coal Plant Order, Narrowing DOE Emergency Power

    TL;DR · 30-second read

    The Short Version

    A 64-year-old coal plant in western Michigan was supposed to shut down in May 2025. The federal government ordered it to keep burning coal anyway, saying the power was needed to stop blackouts.

    A federal appeals court has now ruled that order was unlawful, because there was no real emergency. Keeping the plant open has cost about $259 million so far, and Midwest households and businesses may end up paying much of it in their electricity bills.

    The government is still issuing similar orders at other plants, so the fight is not over.

    The U.S. Court of Appeals for the District of Columbia Circuit ruled Friday that the Energy Department exceeded its authority when it forced Michigan’s J.H. Campbell coal-fired power plant to operate past its scheduled May 2025 retirement. The Associated Press reported that a unanimous three-judge panel sided with environmental groups and the states of Michigan, Illinois and Minnesota, finding there was no genuine emergency under the Federal Power Act. Writing for the panel, Judge Cornelia Pillard described the statute’s emergency provision as “essentially a narrow, last-resort backstop” and called the reversal of the plant’s “long and carefully planned retirement” disruptive.

    The 64-year-old plant in West Olive, operated by Consumers Energy, has stayed online under successive orders signed by Energy Secretary Chris Wright, who argued the unit was needed for regional reliability. Keeping it running has cost roughly $259 million so far, according to new financial filings. Consumers Energy said it is reviewing the ruling and will keep the plant operating under a more recent Energy Department directive that extends through mid-November.

    Executive Summary

    The decision is the first appellate ruling to test how far the Energy Department can stretch emergency authority to keep generation online, and it lands squarely against the department. The court did not weigh in on whether the Campbell plant’s output was useful last winter. It held that usefulness is not the legal test: the emergency provision applies only when immediate, last-resort action is required and when the Energy Department in particular must act, rather than the states responsible for making sure enough generation exists to meet demand.

    That distinction matters well beyond one coal unit. Since President Trump’s January 2025 executive order declaring a national energy emergency, which cited demand growth from artificial intelligence and data centers, the administration has used the same authority to hold roughly half a dozen coal plants and one Pennsylvania oil-and-gas plant past their closure dates. The ruling gives challengers in Indiana, Colorado, Florida and Washington a precedent written by the court that most federal energy appeals run through.

    It also sharpens a question the industry has largely deferred. If aging fossil units cannot be conscripted by federal order, then the burden of serving new data center load falls back where the court says it belongs, on state regulators, regional grid operators and the developers themselves. Emergency orders were never a capacity plan. They are now a considerably less reliable one.

    A Backstop, Not a Planning Tool

    The Federal Power Act’s emergency provision, Section 202(c), lets the Energy Department order a power plant to run during a sudden crisis. It was written for hurricanes, fuel shortages and deep freezes, not for structural shortfalls that everyone can see coming years out. The D.C. Circuit read it that way, holding that the authority is “triggered only when there is a need for immediate, essentially last-resort action and the circumstances require action by DOE in particular, as opposed to action by the state or states responsible for resource adequacy.” Resource adequacy is the industry term for having enough generation on hand to meet peak demand, and in the United States it is primarily a state and regional grid operator responsibility.

    The second half of the holding is the part utilities will read twice. Reversing a long-planned retirement, the court said, is itself disruptive. That reflects operational reality. A plant scheduled to close sheds staff, winds down coal supply contracts, defers environmental capital spending and is replaced in the resource plan by something else. Restarting that machinery is not free, and the court treated the disruption as a cost of the order rather than a neutral act of caution.

    It is worth being precise about what the court did not decide. The panel reviewed statutory authority, not engineering. The Energy Department’s specific claims, that Campbell delivered more than 650 megawatts daily from January 21 to February 1 and that coal output in affected regions rose 25 percent at the peak of the winter storms, were not rejected as facts. Its broader claim that the orders “likely saved hundreds of lives” is a causal inference the department has not published supporting analysis for. And if the output genuinely was needed during those storms, that points to a resource adequacy gap which the ruling says states and grid operators, not emergency orders, are supposed to close.

    The $259 Million Question

    The hard number in this story is the cost: roughly $259 million to keep one 64-year-old unit running past its retirement date. Costs from reliability orders of this kind are typically recovered through regional rates, which is why the challengers, and Michigan’s attorney general, framed the case around Midwest ratepayers footing the bill for a plant their own utility had planned to close. Sierra Club attorney Sanjay Narayan called the ruling a victory for families “paying to keep this old, expensive and dirty power plant online.”

    Advocacy framing aside, the underlying economics are not really contested. A unit gets a retirement date because its running costs, maintenance burden and emissions compliance no longer pencil out against alternatives. An emergency order does not change that arithmetic; it transfers the loss to someone else. Neither Consumers Energy nor the Energy Department has published the plant’s operating economics under the orders, so the $259 million figure is the only quantified measure of what the intervention has cost, and the ruling leaves open who ultimately absorbs it.

    For large electricity buyers, this is the cost-allocation fight in miniature. Reliability insurance for a region has to be paid for by somebody. When it is bought through an emergency order rather than a capacity market or a bilateral contract, the payer is chosen administratively and litigated afterward. That is a poor foundation for a ten-year power budget.

    What AI Load Growth Actually Needs

    The January 2025 executive order invoked artificial intelligence and data center growth as justification for a national energy emergency. Set the scale side by side. Campbell’s contribution during the storm period was a bit over 650 megawatts, which the Energy Department described as enough for hundreds of thousands of homes. It is also roughly the appetite of a single large AI training campus of the kind being announced across the Midwest and Mid-Atlantic. Holding a handful of aging units online buys time measured in months and capacity measured against one or two hyperscale projects.

    The practical takeaway for developers is that announced retirement dates just became more credible as planning inputs. Siting a campus on the assumption that a nearby fossil unit will be kept alive by federal order is now a materially riskier bet. The binding constraints remain what they were: interconnection queues, the multi-year backlog of requests to connect new plants and large loads to the grid, plus transformer and turbine lead times. None of those are addressed by an emergency order.

    The court’s federalism framing pushes the pressure toward state commissions and regional grid operators, and through them toward large customers. Expect more of what is already emerging: on-site or behind-the-meter generation, long-term contracts that fund firm capacity directly, and demand flexibility arrangements in which a data center agrees to curtail during system peaks in exchange for faster service. Those are slower and more expensive than a federal directive, but they do not expire when a panel of judges rules.

    The Orders Keep Coming

    The administration’s response was immediate. Hours after the ruling, Secretary Wright issued a new emergency order keeping the TransAlta-operated Centralia plant in Washington online; it had been scheduled to close at the end of 2025. Similar orders cover plants in Indiana, Colorado and Florida, and a Pennsylvania oil-and-gas plant has been directed to keep its turbines running as a hedge against Mid-Atlantic shortages. Consumers Energy, for its part, will keep Campbell running through mid-November under a separate directive while it reviews the decision.

    Each order stands on its own record, so the ruling does not automatically void the others. It does hand challengers a template, and because the D.C. Circuit hears most federal energy appeals, its reasoning carries unusual weight. The Energy Department has not said whether it will seek rehearing or Supreme Court review, which is the near-term question for anyone modeling how long this authority remains usable.

    For plant owners, an emergency order is revenue with a legal tail. Payments made under an order later found unlawful invite disputes over recovery and potential disallowance, and the uncertainty itself imposes planning costs on utilities trying to decide whether to retire, retrofit or resell a marginal unit. The cleanest reading of Friday’s decision is not that coal must close on schedule. It is that the federal government cannot unilaterally decide it should not.

    Background

    The J.H. Campbell plant has operated on Michigan’s west side for 64 years and is run by Consumers Energy, the regulated utility subsidiary of CMS Energy. Like much of the aging coal fleet in the Midwest grid region, it was slated for closure as operating costs, maintenance demands and cleaner alternatives shifted the economics. Its retirement date of May 2025 had been in utility resource plans for years before the Energy Department intervened.

    Section 202(c) of the Federal Power Act, the provision at issue, has historically been used sparingly and briefly, typically during hurricanes, fuel disruptions or extreme cold snaps. That changed after President Trump’s January 2025 executive order declaring a national energy emergency, which cited electricity demand growth from artificial intelligence and data centers. Since then the department has used the authority to hold roughly half a dozen coal plants past their closure dates, plus an oil-and-gas plant in Pennsylvania, producing litigation in several states of which the Michigan case is the first to reach an appellate decision.

    Sources

    Source: Federal court rejects Trump order keeping Michigan coal plant open — Associated Press report on the D.C. Circuit ruling that the Energy Department exceeded its authority in ordering the J.H. Campbell coal plant to operate past its scheduled retirement.

  • Anthropic Flags Houthi-Linked Attempts to Use AI for Weapons

    Anthropic Flags Houthi-Linked Attempts to Use AI for Weapons

    TL;DR · 30-second read

    The Short Version

    Anthropic, an American artificial intelligence company, says people in the part of Yemen controlled by the Houthi armed movement tried to use its chatbot for help developing advanced weapons.

    Why an ordinary person should care: the same software that drafts emails and writes computer code can be asked for dangerous things too, and for now the company selling it is also the one policing it.

    Anthropic says it caught the activity. It has not said how many people were involved, what they got out of it, or when any of it happened.

    The Associated Press reported that Anthropic said users in Houthi-held Yemen attempted to use its artificial intelligence models to help develop advanced weapons. The activity originated from territory controlled by the Houthi movement, which governs much of northern Yemen including the capital, Sanaa, and which has been the subject of United States terrorism designations in recent years.

    Anthropic’s account, as reported, describes attempts rather than outcomes. The company has not publicly quantified the activity, named the weapons categories involved, or detailed how accounts in a conflict zone reached a commercial model that its own usage policy bars from weapons work.

    Executive Summary

    Frontier artificial intelligence providers have spent two years arguing that their models could, in principle, lower the skill barrier for building dangerous things. This disclosure describes an attempt to do exactly that, from a specific place, by users a model provider says it identified. That is a meaningful shift in register: from risk modelling to an incident.

    For the infrastructure industry, the significance is not the headline about Yemen. It is that a software vendor with no physical presence in the country became, by default, the enforcement point for a national-security question. Model providers now perform a function that looks less like content moderation and more like sanctions screening, with the detection happening inside the inference layer rather than at a border or a bank.

    That responsibility does not stay with the model provider. The same user journey touches cloud hosting, payment processing, identity verification, content delivery and transit networks, each operated by a different company with its own compliance obligations and its own log data. Anthropic’s disclosure is a useful prompt for every one of those operators to ask what it would have seen, and whether it would have recognised what it was looking at.

    Misuse Stops Being a Thought Experiment

    Until recently, the debate over catastrophic artificial intelligence risk ran almost entirely on hypotheticals and red-team exercises, where a lab hires experts to attack its own system. Critics reasonably noted that no one had produced a concrete case of a hostile actor using a commercial model for weapons work. A disclosure of attempted weapons development by users in a conflict zone changes the shape of that argument, even without a single confirmed capability gained.

    The honest reading is narrow. Attempted misuse is evidence that the intent exists and that commercial models are considered worth trying, which was always the more likely of the two uncertainties. It is not evidence that the models delivered anything a determined group could not have found in technical literature. Those are separate claims, and conflating them is how both the alarmed and the dismissive camps get this wrong.

    What is newly hard to dispute is the operational point. Access controls on frontier models are now load-bearing parts of a counter-proliferation regime, and they are maintained by private companies on commercial timelines. Any security professional who has watched a fraud team tune detection thresholds knows what that means in practice: the line moves, it is probabilistic, and it is only as good as the signals feeding it.

    Who Owns the Control Point?

    A request that reaches a model from a sanctioned or conflict-affected region passes through a long chain. Someone sold the connectivity. Someone hosted the inference, possibly a hyperscale cloud provider reselling model access, meaning a large cloud platform offering a third party’s model through its own interface. Someone processed the payment, or accepted a prepaid alternative. Someone provided the intermediary address that made the traffic look ordinary, whether a commercial virtual private network or a low-cost server in a neutral jurisdiction.

    Each link in that chain has compliance duties that were written for a different kind of transaction. Export-control rules and sanctions screening evolved around shipped goods, wire transfers and named entities. They map awkwardly onto a metered conversation with a model, where the controlled item is not the software but the knowledge it assembles on request, and where attribution to a place depends on signals that are trivially easy to disguise.

    The practical consequence for operators is that geographic risk is becoming a question about traffic and identity, not just about where racks sit. Providers that already run robust know-your-customer processes, reseller governance and abuse telemetry are better positioned than those that treat artificial intelligence workloads as ordinary compute. Expect enterprise buyers, particularly in regulated sectors, to start asking their cloud and model vendors how misuse detection works, how often accounts are actioned, and what gets escalated to governments.

    The Uplift Question Decides the Policy

    Everything downstream of this story depends on a single unresolved empirical question: does a frontier model materially help someone build a weapon they could not otherwise build? That is the uplift question, and it is the axis on which regulation, liability and procurement will turn. If uplift is marginal, access controls are a reputational and legal necessity but not a strategic one. If uplift is real for some categories, then model access belongs in the same policy family as controlled precursors and specialised machine tools.

    Neither the companies nor the public evidence base has settled this, and it would be misleading to pretend an attempted-misuse disclosure settles it. Classified assessments likely exist; published ones are thin. Buyers should therefore treat vendor safety claims as statements about process rather than outcomes, and read them the way they read a penetration-test summary: useful, bounded, and not a guarantee.

    There is a defensive corollary worth naming for critical-infrastructure operators. The same capabilities being tested for weapons work apply to the dull, effective attacks that actually take down plants, grids and pipelines: reconnaissance, industrial-protocol familiarisation, convincing phishing in a target’s own language. Adversary capability should now be modelled as artificial-intelligence-assisted by default, independent of how the weapons debate resolves.

    A Region That Already Matters to the Network

    Yemen sits beside one of the most consequential stretches of digital infrastructure on the planet. The Red Sea and the Bab al-Mandab strait carry subsea cable systems linking Europe, the Gulf, East Africa and Asia, and cable damage in those waters has previously degraded intercontinental capacity and forced traffic onto longer, costlier paths. Connectivity planners have treated the corridor as a concentration risk for years, which is why new routes around it keep getting announced.

    That context makes this story less abstract for infrastructure readers than a story about chatbot prompts might sound. The same geography shows up in two different risk registers at once: physical exposure of cables and shipping lanes, and now conduct-based exposure at the application layer. Operators with assets, landing stations or partners in the region will find both on the same page of their next board report.

    None of this argues for treating a country as uniformly hostile. Yemen is also a place where millions of civilians rely on the same internet access and the same commercial tools as everyone else, and blunt geographic blocking imposes real costs on them while barely inconveniencing a well-resourced actor with an intermediary server. The uncomfortable conclusion is that effective controls have to be behavioural and account-level, which is harder, slower and far less satisfying to announce.

    Background

    Anthropic is an American artificial intelligence company and the developer of the Claude family of large language models, systems trained on vast text corpora to generate text and code on request. It sells access directly and through major cloud platforms, and has positioned safety research and published usage restrictions as central to its commercial identity. Alongside its peers, it has increasingly published accounts of detected misuse of its models, a practice that borders on the threat-intelligence reporting long familiar from cybersecurity vendors.

    Yemen has been divided by civil war for over a decade, with the Houthi movement controlling the north and a recognised government and allied forces holding parts of the south. The conflict has drawn in regional and international actors and made the surrounding waters a security concern for commercial shipping and the subsea cable systems that share those routes. For technology companies, the country sits in the category of jurisdictions where sanctions exposure, limited verification infrastructure and genuine civilian need for digital services all overlap.

    Sources

    Source: Users in Houthi-held Yemen tried to develop advanced weapons with AI, Anthropic says — Associated Press report on Anthropic’s disclosure of attempted weapons-development use of its models by users in Houthi-controlled Yemen.