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

Author: Deepak Jain

  • Digital Realty’s $3B Hyperscale Fund Target Ties AI Build Pace to Investor Appetite

    Digital Realty’s $3B Hyperscale Fund Target Ties AI Build Pace to Investor Appetite

    TL;DR · 30-second read

    The Short Version

    Digital Realty, a company that builds data centers and rents them out, wants to raise $3 billion from outside investors to pay for very large ones in the Americas.

    Data centers are warehouse-sized buildings full of computers that run online services and artificial intelligence. The biggest ones cost so much that even large owners share the bill with investors.

    Why it matters: how much money investors actually put in will help decide how many of these buildings get built, and how fast.

    IPE Real Assets reported that Digital Realty Trust (DLR) is targeting $3 billion for a fund dedicated to hyperscale data centers in the Americas. Hyperscale facilities are the very large buildings and campuses leased to the biggest cloud and technology companies. The fund would pool capital from outside investors to finance them alongside Digital Realty’s own balance sheet.

    The figure is a fundraising target, not a completed raise. No first close, named investors, specific campuses or capacity figures have been disclosed.

    Executive Summary

    Digital Realty is seeking $3 billion of outside capital for a vehicle focused on hyperscale data centers across the Americas. For a publicly listed real estate investment trust, a private fund is a way to fund capital-intensive campus development without relying only on new shares or corporate debt, while keeping a role as manager and co-owner.

    The significance lies in who ends up setting the pace. For the projects routed through this vehicle, the scale of construction depends on how much investors commit and on what terms, not only on Digital Realty’s own capacity to borrow or issue stock. That gives limited partners, the institutions that supply fund capital, a meaningful influence over how much hyperscale capacity this structure can finance.

    What the target does not yet show is equally important: how many megawatts the fund covers, which markets, whether tenants are already signed, and whether grid power is secured. Those details will determine whether $3 billion translates into near-term capacity or a longer pipeline.

    A $3 Billion Target Is a Ceiling Investors Set

    A fundraising target is an ambition, not a bank balance. When a sponsor such as Digital Realty says it is targeting $3 billion, it is telling prospective investors, typically pension funds, insurers and sovereign wealth funds known as limited partners, how large it hopes the vehicle will be. Commitments arrive over one or more closings, and the final total can land below, at or above the target. The committed figure, not the headline, determines how much development the fund can underwrite.

    That is the mechanism behind the headline. For whatever share of Digital Realty’s Americas hyperscale pipeline flows through this vehicle, the pace of building (how many campuses break ground, and when) depends on how much outside equity arrives and on the terms investors accept. Borrowing can stretch fund equity further, but lenders generally size loans against the equity beneath them, so investor appetite still sets the base. Digital Realty’s balance sheet remains a funding source. The fund adds a second one whose size the company does not fully control.

    The effects reach beyond the REIT. Hyperscale tenants, the cloud and AI companies that lease whole buildings or campuses, gain a capacity source that does not hinge solely on one listed company’s share price and credit. Utilities and equipment suppliers planning around these campuses gain another variable to track: capital availability alongside power availability.

    Why a Listed Landlord Shares the Bill

    Real estate investment trusts must pay out most of their taxable income as dividends, which limits the cash they can retain for construction. Large hyperscale developments are capital-intensive and take years to reach full rent, so funding them only through retained cash, new shares or corporate debt either dilutes existing shareholders or raises leverage. A private fund offers a third route: outside investors supply much of the equity, while the sponsor typically keeps a stake, manages the assets and collects fees.

    The trade-off is that Digital Realty would share the economics of the assets it places in the vehicle. Fee income and a smaller capital outlay per project come at the cost of a smaller ownership slice. For shareholders, the question is whether fees and faster growth outweigh the returns passed to fund partners, a balance that depends on terms that have not been made public.

    For institutions, a dedicated vehicle offers direct exposure to hyperscale data centers through an established developer-operator, rather than indirect exposure through listed REIT shares. How strongly they respond to the $3 billion target will be an early signal of how much private capital remains available for this segment.

    What Would Make the Fund Matter

    The dollar figure matters less than what those dollars buy. The details that would show the fund’s real reach (megawatts of capacity, specific campuses, pre-signed leases, secured grid connections) have not been disclosed. A fund backing pre-leased capacity with power already secured carries very different risk from one financing land that still awaits utility service and tenants.

    The “Americas” label also deserves attention. It suggests a mandate that may extend beyond the United States, but no markets have been named. Power availability, permitting timelines and customer demand vary widely across the region, and the fund’s geographic mix will shape both its returns and where new hyperscale capacity actually appears.

    Background

    Digital Realty Trust, traded under the ticker DLR, is a real estate investment trust that develops, owns and leases data centers to enterprises and large cloud providers. As a REIT, it distributes most of its taxable income to shareholders, which shapes how it funds new construction.

    The hyperscale segment, meaning very large facilities built for the biggest cloud and AI customers, is one of the most capital-intensive parts of digital infrastructure. Operators across the industry have increasingly paired their own balance sheets with joint ventures and private funds to finance campus-scale development, sharing ownership in exchange for access to institutional capital.

    Sources

    Source: Digital Realty targets $3bn for Americas hyperscale data centre fund (IPE Real Assets), on Digital Realty’s fundraising target for an Americas hyperscale data center vehicle.

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

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

    TL;DR · 30-second read

    The Short Version

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

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

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

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

    Executive Summary

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

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

    Why a Cloud Giant Would Sell Chips It Already Owns

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

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

    What an $8 Billion GPU Buyer Actually Has to Underwrite

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

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

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

    Who Gains if GPUs Become a Financed Asset

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

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

    Background

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

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

    Sources

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

  • Equinix Ireland’s 36% Profit Fall Masks 9% Growth and a Grid Risk to AI-Era Expansion

    Equinix Ireland’s 36% Profit Fall Masks 9% Growth and a Grid Risk to AI-Era Expansion

    TL;DR · 30-second read

    The Short Version

    • Equinix runs giant buildings full of computers that firms like Microsoft, Google and Amazon use. Its Irish business saw profit before tax fall by more than a third last year, to 4.4 million euros.
    • That sounds alarming, but the drop mostly reflects a one-off payment the company received the year before. Sales actually grew 9%.
    • The real worry is electricity. The company’s own directors warn that power suppliers may not be able to give it more in the places it wants to grow.
    • In Ireland, where these computer buildings already lean heavily on the national grid, power may decide how fast they can expand.

    RTÉ reported that newly published accounts for Equinix (Ireland) Ltd, the Irish arm of the California-based data centre operator, show pre-tax profit fell 36% to €4.4m in 2025, even as revenue rose 9% from €75.3m to €82.4m. The decline was due mainly to €2.9m of dividend income booked in 2024 that did not recur. Operating profit fell 6% to €5.43m, gross margin rose from 23% to 25%, and EBITDA (earnings before interest, tax, depreciation and amortisation) came to €13.2m.

    During the year, Equinix bought BT Group’s Irish data centre business, including facilities in Citywest and Ballycoolin, Dublin, for €61.27m, a deal the Competition and Consumer Protection Commission cleared in August 2025. Headcount rose from 101 to 110, and the directors flagged the risk that electricity providers may not be able to supply further capacity where the company expects to expand.

    Executive Summary

    Equinix’s Irish subsidiary reported a sharp fall in pre-tax profit for 2025, to €4.4m from the prior year, on revenue that grew 9% to €82.4m. The directors attribute the fall mainly to a €2.9m dividend received in 2024 that was not repeated. The operating business tells a steadier story: gross margin improved to 25% on growth in recurring colocation and interconnection revenue, and operating profit slipped only 6% to €5.43m.

    That steadier story carries a cost signal. Staff costs rose by about a quarter, from €14.18m to €17.8m, while headcount grew 9%, absorbing roughly half the year’s €7.1m revenue gain and thinning the operating margin even as gross margin widened. The company also added capacity through its €61.27m purchase of BT’s Irish data centre business.

    The most consequential line in the accounts is not a number. The directors say they intend to keep increasing capacity, and they name electricity supply as the risk to doing so. In Dublin, where grid connections for new data centres have been restricted for years, that makes power availability, not customer demand, the variable to watch for Equinix, its customers and anyone pricing data centre capacity in Ireland.

    A Dividend Explains the Headline, Not the Business

    The 36% headline compares two years that are not like for like. In 2024, Equinix (Ireland) Ltd booked €2.9m of dividend income, money received rather than earned by running data centres, and that income did not recur in 2025. The directors identify it as the main reason for the fall. Measured on operations alone, the business held up: operating profit fell 6% to €5.43m, and after €1m of net interest the company reported pre-tax profit of €4.4m.

    The cash-generation picture is firmer still. EBITDA, a rough proxy for the cash the operations throw off before financing and accounting write-downs, was €13.2m. The gap between that figure and operating profit is mostly €7.7m of depreciation, the non-cash charge that spreads the cost of buildings, generators and cooling plant over their working lives. Heavy depreciation is normal in a capital-intensive asset class like data centres and says little about demand, which the directors describe as having “remained solid.”

    Wider Gross Margins, Thinner Operating Margins

    Gross margin rose from 23% to 25%, which the directors credit to growth in recurring interconnection and colocation revenue. Colocation, renting space, power and cooling so customers can house their own servers, brought in €67.25m, more than four-fifths of revenue. Interconnection, directly linking customers to networks, clouds and each other inside the same building, contributed €9.44m, about 11%. Managed infrastructure added €3.6m and non-recurring revenue €1.85m, so nearly all income is recurring. Interconnection is valued in the colocation model because each new link adds revenue while using little extra floor space or power.

    Below the gross line, costs grew faster than sales. Revenue rose by about €7.1m, while staff costs rose by €3.6m, roughly 25%, against a 9% rise in headcount, from 101 to 110 people (79 engineering and technical, 29 sales and administration, two directors). On the reported figures, operating margin fell from roughly 7.7% to about 6.6%. The accounts do not break down the staff cost increase; the year also included the integration of BT’s Irish data centre business, but the company has not linked the two.

    Power, Not Demand, Sets the Pace of Expansion

    The 9% revenue growth matters less than what it takes to repeat it. The directors say the company “intends to continue to increase its capacity,” and in the same accounts they name the risk that electricity providers may not be able to provide further capacity in locations where it expects to expand. In Dublin that is a live constraint: Ireland’s grid operator, EirGrid, has for several years restricted new data centre connections in the region, and the energy regulator has moved toward requiring new large energy users to bring their own generation or storage. For an operator that wants to grow, a grid connection, not a customer contract, is the scarce input.

    The link to the AI buildout runs through power density. Equinix markets its 282 data centres worldwide as “AI-ready,” and the Irish directors list rising “requirements for power and cooling” among the market’s growth drivers. Denser computing, of which AI training and inference are the leading example, draws more electricity per rack, so customer growth consumes grid capacity faster than it consumes floor space. The Irish directors do not attribute demand to AI specifically; they cite internet traffic, financial services computing, cloud and software-as-a-service. But whatever the workload, more compute per site pushes power to the front of the queue.

    Two consequences follow for the people who buy and finance this capacity. First, capacity that is already connected and powered becomes more valuable, which is consistent with the directors calling theirs “premium” and with the widening gross margin; buying BT’s operating Citywest and Ballycoolin facilities added capacity without waiting on a new connection, though the company has not framed the deal that way. Second, energy costs are moving toward customers: the directors say price increases are embedded in the budget and partly hedged, that “communication has started towards the customers,” and that a task force is looking for ways to cut primary energy use. Dublin tenants should expect power terms, not just space, to drive renewals.

    Background

    Equinix Inc., headquartered in California and listed under the ticker EQIX, owns and operates 282 data centres in 75 locations worldwide. Its model centres on colocation, renting secure, powered and cooled space for customers’ servers, and interconnection, which lets those customers link directly to networks, cloud platforms and one another inside the same facility. Its client list includes Oracle, Nvidia, Google Cloud, Microsoft, Dell Technologies and Amazon Web Services.

    Dublin is one of Europe’s main data centre hubs, home to large cloud and technology operations, and also one of its most power-constrained. Ireland’s grid operator has restricted new data centre connections in the Dublin region for several years, making existing grid capacity a strategic asset. Equinix’s Irish subsidiary runs five Dublin sites and, in 2025, added BT Group’s Irish data centre business, including facilities in Citywest and Ballycoolin.

    Sources

    Source: Pre-tax profits at data centre builder Equinix down 36% (RTÉ), a report on the 2025 accounts of Equinix (Ireland) Ltd.

  • Sharon AI’s 9.95% GPU Loan Shows Lenders Underwrite AI Chips Through Their Contracts

    Sharon AI’s 9.95% GPU Loan Shows Lenders Underwrite AI Chips Through Their Contracts

    TL;DR · 30-second read

    The Short Version

    An Australian company called Sharon AI has borrowed 356 million US dollars and pledged the computer chips it is buying as security, much like a mortgage uses the house.

    The chips are powerful Nvidia processors that run artificial intelligence. The loan costs about 10 percent a year, and Goldman Sachs is among the lenders.

    The twist: lenders get repaid from what customers pay to rent those chips. So the loan really rests on Sharon AI’s customer deals, and it is only the first of several loans behind a plan to install 68,000 chips by mid-2027.

    SharonAI Holdings Inc. (NASDAQ: SHAZ), an Australian provider of AI computing infrastructure, announced on October 1, 2026 that it has entered its first GPU-backed debt facility: a US$356 million committed, senior secured loan held in a special purpose vehicle and priced at a fixed 9.95%, excluding fees. The facility is secured against the GPUs and their associated cash flows, and proceeds will fund compute capacity dedicated to customer contracts.

    Lenders include Goldman Sachs and select large private credit funds alongside Australian, Asian and global investors, with Jarden Australia acting as sole financial advisor and arranger. Sharon AI called it the first in an expected series of GPU financings supporting more than 68,000 NVIDIA GPUs deployed by mid-2027, and said it has now secured more than US$2.6 billion of debt and equity capital over ten months against a customer contract book worth more than US$8.8 billion.

    Executive Summary

    Sharon AI has opened a new funding channel for its GPU fleet: debt secured directly by the chips and the customer revenue they generate. The US$356 million facility is modest against the company’s 68,000-GPU build-out plan, but Sharon AI presents it as the template for a series of similar financings, shifting part of its hardware funding from equity to borrowed money at a fixed 9.95% rate before fees.

    The significance lies in the structure. Because the security package combines hardware and contracted cash flows, the lenders are effectively underwriting Sharon AI’s customer contracts, with GPU resale value as the fallback. That makes the quality, length and enforceability of those contracts, rather than raw chip count, the variable that determines how much Sharon AI can borrow and at what price.

    The deal also lands days after Sharon AI disclosed a change of auditor to Ernst & Young and announced a confidential-computing collaboration with VAST Data aimed at regulated customers, two developments that bear on how lenders and investors will judge the company’s reporting and its contract pipeline.

    The Chip Is the Collateral, the Contract Is the Repayment

    Sharon AI describes the facility as senior secured and GPU-backed, held in a special purpose vehicle (SPV), a separate legal entity set up to own specific assets and borrow against them. The security is “the GPUs and associated cash flows,” and proceeds will fund compute “dedicated to customer contracts.” Senior secured means these lenders are paid first and hold a direct claim on the pledged assets.

    Put those pieces together and the lenders’ position becomes clear. In normal operation, the debt is serviced from what customers pay to use the specific GPUs inside the SPV. The chips themselves matter mainly if something goes wrong: if a customer stops paying or the SPV defaults, lenders can claim the hardware and its revenue stream. From the lender’s seat, each GPU is worth largely what its attached contract is expected to earn over the life of the loan, with resale value as the backstop.

    That has consequences beyond Sharon AI’s balance sheet. Graphics processors are a fast-moving asset class, and each new NVIDIA generation can change what older chips fetch on the secondary market. A lender relying on resale value alone would carry that risk directly. Tying security to contracted cash flows moves the central credit question from what a chip will be worth at the end of the loan to whether the customer will keep paying, which makes the strength and duration of the customer book the main driver of how cheaply an AI cloud operator can finance hardware.

    What a Fixed 9.95% Buys, and What It Leaves Out

    The 9.95% rate is fixed, so Sharon AI’s interest bill will not rise if benchmark rates do, a useful match for revenue that comes from multi-year contracts at agreed prices. The figure explicitly excludes fees, however. Arrangement and commitment fees on structured facilities can be meaningful, so the all-in cost of this borrowing is higher than 9.95% by an amount the company has not stated.

    Chief Executive James Manning said the structure is “designed to enhance return on equity.” The mechanism is simple: if the GPUs earn more than the cost of debt, borrowing lets shareholders keep the difference on a smaller equity base. The arithmetic also runs in reverse. If utilization, pricing or customer payments fall short, leverage magnifies losses for equity holders, who rank behind senior lenders. Whether the claim holds depends on contract economics Sharon AI has not published.

    The lender roster adds context. Goldman Sachs and “select large private credit funds” sit alongside Australian and Asian investors. Private credit, meaning lending by investment funds rather than banks, commonly provides asset-backed loans of this kind, and a committed facility (money promised up to a limit) gives Sharon AI drawable capacity as deployments come online.

    $356 Million Is the Opening Tranche of a 68,000-GPU Plan

    Sharon AI calls the facility “the first in an expected series of GPU financings” behind a scheduled build-out of more than 68,000 NVIDIA GPUs by mid-2027. It has not said how many of those chips this US$356 million covers, but the framing is clear: this is a repeatable structure, not the whole bill.

    Two company figures set the scale. Sharon AI says it has secured more than US$2.6 billion of institutional debt and equity over the past ten months, and that its customer offtake, meaning contracted commitments to buy compute, stands at a total contract value (TCV) above US$8.8 billion. TCV sums the full value of contracts across their entire terms, so it is not annual revenue, and its worth to a lender depends on contract length, cancellation rights and counterparty strength. Sharon AI lists hyperscalers (the largest cloud platforms), AI-native companies, government, enterprise and research organizations among its customers, without naming those behind this facility.

    The company is also building product features for regulated buyers. On September 23 it announced a collaboration with VAST Data to offer VAST DataEnclave, a confidential-computing environment in which, the companies say, customer data and model weights remain encrypted even from Sharon AI as operator, targeting banks, government agencies and other regulated organizations. If such features help win long-dated contracts from institutional customers, they feed directly into the cash flows that secure facilities like this one.

    Reporting Discipline Matters More Once Debt Is Layered In

    Lenders monitor borrowers through their financial reporting, so audit quality becomes more important as debt is added. In an 8-K filed September 24, Sharon AI disclosed that its Audit Committee approved ending HoganTaylor LLP’s engagement on September 20 and engaged Ernst & Young on September 23 for the fiscal 2026 audit, following a competitive process involving several firms.

    The filing states there were no disagreements with HoganTaylor and that its audit reports for 2024 and 2025 carried no adverse or qualified opinions. It does list one reportable event: a material weakness, meaning a significant gap in internal controls over financial reporting, identified by HoganTaylor and disclosed in the 10-K filed March 31, 2026. Moving to a Big Four auditor is a common step for a growing public company, and the filing gives no indication of a dispute. It does not say whether the material weakness has been remediated, and for a borrower planning a series of GPU-backed facilities, reliable controls over cash flows and asset records are part of what lenders are underwriting.

    Background

    SharonAI Holdings Inc. trades on Nasdaq under the ticker SHAZ and describes itself as an Australian “neocloud,” a specialist provider that rents GPU computing power for AI rather than offering general-purpose cloud services. Through its AI Factory platform and a network of technology and co-location partners (data center operators that host its equipment), it serves organizations training and running AI models, with an emphasis on sovereign infrastructure that keeps data within national borders. It reports to the SEC as an emerging growth company.

    Companies in this segment typically need large sums up front to buy GPUs before customer revenue arrives, which makes financing structure central to their growth. Sharon AI says its ambition is gigawatt-scale compute capacity across Australia, New Zealand and the broader Asia-Pacific, a gigawatt being roughly the output of a large power station.

    Sources

    Source: Sharon AI Enters Into GPU-Backed Debt Facility, Expanding Funding Flexibility for AI Factory Deployments, Sharon AI’s announcement of its US$356 million GPU-backed SPV debt facility.

    Primary sources: SharonAI Holdings 8-K filed September 24, 2026 (change in certifying accountant); SharonAI Holdings 8-K/A filed September 23, 2026; Exhibit 99.1 to 8-K/A: Sharon AI Collaborates with VAST Data on the Launch of DataEnclave; SharonAI Holdings 8-K filed September 23, 2026; Exhibit 99.1 to 8-K: Sharon AI Collaborates with VAST Data on the Launch of DataEnclave.

  • Eaton’s 5,000 m² Austrian Expansion Lands at the Low-Voltage End of AI’s Power Chain

    Eaton’s 5,000 m² Austrian Expansion Lands at the Low-Voltage End of AI’s Power Chain

    TL;DR · 30-second read

    The Short Version

    Eaton, one of the world’s biggest makers of electrical equipment, has finished enlarging its factory in Schrems, Austria. It added floor space nearly the size of an American football field in under ten months.

    The new space makes small circuit breakers, the switches in a fuse box that cut power when something goes wrong, on fully automated lines. It also builds more of the larger equipment that spreads power around a building.

    Why it matters: the computer warehouses behind artificial intelligence need large amounts of this gear. Eaton has not said how much more it will make.

    Eaton is marking the completion of a 5,000-square-metre expansion of its manufacturing site in Schrems, Austria, on 1 October 2026, according to a company announcement published by the Financial Times. The new space houses fully automated production of miniature circuit breakers (the small protective devices that sit in electrical distribution panels and trip when a circuit overloads or faults) and expanded assembly capacity for low-voltage systems.

    Eaton says the project was completed in less than ten months and ties it to rising European demand from electrification, digitalisation and critical infrastructure that depends on highly available electrical power. The company did not disclose the investment amount, production volumes or headcount.

    Executive Summary

    Eaton, which reported $27.4 billion in 2025 revenue, has added 5,000 m² of manufacturing space at its Schrems plant in Austria. The expansion combines fully automated miniature circuit breaker production with a manufacturing execution system, software that links machines, internal logistics and enterprise systems in real time, and end-to-end traceability for every product that leaves the line.

    The announcement matters to data center builders because of where the capacity sits. Miniature circuit breakers and low-voltage systems form the downstream end of a facility’s electrical chain, the layer that distributes and protects power close to racks, cooling and building systems. Large AI facilities use this equipment in volume, and Eaton names data centers among its core markets.

    The announcement covers less ground than a headline about European power capacity might suggest. It says nothing about medium-voltage switchgear or transformers, and Eaton has not quantified the added output. The expansion adds European supply at the breaker and panel level. Eaton has not shown that it relieves the larger, grid-side equipment queues.

    Capacity at the Low-Voltage End of the Power Chain

    To see where this expansion matters for AI data centers, follow electricity from the grid to a server. Power arrives at high voltage and is stepped down by transformers. It then passes through medium-voltage switchgear, the heavy equipment that switches and protects large feeders, and into low-voltage switchboards and distribution panels. Finally, individual circuit breakers protect each circuit feeding racks, cooling units and building systems. Eaton’s Schrems expansion adds fully automated miniature circuit breaker production and more low-voltage systems assembly, which places it at the downstream end of that chain.

    That placement shapes how buyers should read the news. A large data center contains a great many protected circuits, so breakers and distribution equipment are consumed in volume. More European output at this layer is relevant to electrical contractors and developers who source panels and protective devices regionally. The announcement does not mention medium-voltage switchgear or transformers. Those are the larger, more engineered items that tend to dominate discussions of power-equipment lead times on AI projects.

    The practical read is that any easing this expansion brings will show up at the breaker and panel level. Projects whose schedules depend on grid-side equipment should not treat it as relief on that front. The same product families also serve Eaton’s residential, commercial, industrial and utility customers. Data center demand therefore competes for this output with every other electrification project in Europe.

    Ten Months to Build, With Software Coordinating the Floor

    Eaton says the expansion was completed in less than ten months. How fast a supplier can add capacity is one of the few supply-side variables it controls directly. A short build suggests Eaton can respond within a single planning cycle when it decides a product line needs more output.

    The more distinctive element is the factory’s digital layer. A manufacturing execution system (MES) is the software that tells machines what to make, tracks work in progress and connects production to inventory and ordering systems. Eaton describes Schrems’ MES as linking manufacturing, intralogistics and enterprise systems in real time. Combined with end-to-end traceability, meaning each unit’s production history can be followed, this addresses what the company calls increasingly demanding performance and compliance requirements. Operators of critical facilities typically need documentation of exactly what was installed and how it was made.

    Full automation also fits the product. Miniature circuit breakers are high-volume, standardised devices, the kind of item where automated lines can raise output without a proportional increase in labour. Eaton has not said what output the new lines deliver. The economic logic is sound, but its effect on supply cannot yet be measured.

    Regional Manufacturing as a Supply-Chain Argument

    Eaton makes its case in supply-chain terms: faster response to changing customer needs, lower logistics costs, a smaller carbon footprint from proximity to European markets, and greater resilience in critical supply chains. For European data center developers, regionally made equipment can mean shorter shipping legs and less exposure to cross-border disruption. Those advantages matter most when construction schedules are tight.

    Scale needs perspective. Eaton operates across 180 countries with $27.4 billion in 2025 revenue, so a 5,000 m² addition at a single site is incremental relative to the whole company. What the expansion offers is a signal of direction. Eaton is adding European capacity in product lines that data centers consume, and it has done so quickly. That is not the same as a step change in regional supply.

    The workforce element addresses a quieter constraint. Eaton points to partnerships with local schools and its Connected Learning Center, opened in 2024, as ways of building the automation and digital-manufacturing skills the site needs. Those skills determine whether new capacity can be staffed and sustained as well as installed.

    Background

    Eaton is a Dublin-based power management company founded in 1911. It makes electrical protection and distribution equipment for data center, utility, industrial, commercial and institutional, residential, machine building, aerospace and mobility customers. It reported $27.4 billion in revenue in 2025 and serves customers in 180 countries.

    Inside a data center, power flows from the grid through transformers and medium-voltage switchgear, then into low-voltage switchboards, distribution panels and circuit breakers that protect individual circuits. Every layer must be delivered before a facility can energise. Supply of this equipment has therefore become a planning concern for AI builds, and capacity additions by major manufacturers are watched closely.

    Sources

    Source: Eaton Strengthens European Manufacturing With Expanded Production Facility in Austria – Company Announcement (Financial Times): Eaton’s announcement of a 5,000 m² expansion at its Schrems, Austria, site adding automated miniature circuit breaker production and low-voltage systems assembly capacity.

  • Broadcom’s Up-to-$42B Anthropic Loan Puts AI Compute Credit Risk on a Chipmaker

    Broadcom’s Up-to-$42B Anthropic Loan Puts AI Compute Credit Risk on a Chipmaker

    TL;DR · 30-second read

    The Short Version

    Broadcom, one of the world’s biggest chip makers, has agreed to lend Anthropic, the company behind the Claude chatbot, up to $42 billion, according to a financial filing. The money is for Anthropic to rent Broadcom’s own chips.

    Think of a carmaker lending you the money to lease its cars. If you stop paying, the carmaker takes the loss.

    That is the real story: some of the risk that an artificial intelligence company cannot pay its bills now sits with the company selling it the hardware.

    Broadcom has agreed to lend Anthropic up to $42 billion so the AI developer can lease Broadcom’s chips, Reuters reported in an exclusive, citing a filing. The arrangement makes Broadcom both the supplier of Anthropic’s AI computing hardware and the source of the money used to pay for it.

    The figure is a ceiling, not a fixed sum. How much Anthropic will draw, over what period, and on what terms has not been made public by either company.

    Executive Summary

    The headline number is large, but the structure is what matters. In a typical hardware deal, a chipmaker sells, gets paid, and any question about whether the buyer can afford the purchase belongs to the buyer and its lenders. Here, the chipmaker is also the lender. If the facility is drawn, Broadcom holds a claim on Anthropic’s future ability to pay, not just a completed sale.

    For Anthropic, the facility offers a way to secure large amounts of AI computing capacity without paying for it upfront. For Broadcom, it helps lock in demand for its chips with one of the most prominent AI model developers. For everyone else watching AI infrastructure spending, it is a concrete example of compute financing moving onto a supplier’s balance sheet, and it raises fair questions about collateral, pricing and how the resulting revenue should be read.

    When the Chip Vendor Becomes the Bank

    The mechanism is simple. Broadcom extends credit of up to $42 billion; Anthropic uses it to lease Broadcom chips; Anthropic repays over time from the money it earns selling AI products. Every dollar drawn becomes a dollar of exposure for Broadcom to a single borrower, and that exposure is only as good as Anthropic’s future revenue. This is what the term credit risk means here: the chance that a borrower cannot repay in full and on time.

    That risk now sits on a chipmaker’s books rather than solely with banks, bond investors or specialist equipment lessors, the parties that normally absorb it. The people affected are Broadcom shareholders and creditors, who inherit exposure to the economics of AI model development; credit analysts, who must now assess a semiconductor company partly as a lender; and Anthropic itself, whose funding for compute is tied to one supplier’s balance sheet.

    None of this makes the arrangement unsound. A loan tied to hardware may be secured by that hardware, and a large, profitable chipmaker can be a more patient lender than a bank. But the risk does not disappear because the lender is also the seller. It changes hands.

    Why Lease, and Why Let the Supplier Pay

    Leasing chips, rather than buying them, lets a customer spread the cost over the period it uses the hardware instead of paying everything upfront. For an AI developer whose computing needs are growing quickly, that matches cash going out to the revenue the chips help generate. Financing the lease through the chip supplier can also be faster than arranging a syndicated loan, and the supplier has unusual insight into the value of the hardware being financed.

    For Broadcom, the logic is demand certainty. Custom AI chips are designed over long cycles, and a customer committed through a multi-billion-dollar financing facility is a customer less likely to switch suppliers. In effect, the loan helps secure the order book. The trade-off is that a portion of Broadcom’s future chip revenue would be funded with Broadcom’s own capital.

    What Credit and Equity Analysts Will Look For

    The first question is revenue quality. When a vendor finances its own sales, the revenue it reports is only fully realized once the customer repays. Investors will want to see how Broadcom accounts for the leases and the loan, and how much of any reported AI revenue is tied to this facility.

    The second question is collateral. AI chips lose value as newer generations arrive, so hardware that secures a loan today may be worth considerably less within a few years. A lender’s protection depends on how fast the loan amortizes relative to how fast the hardware depreciates.

    The third is precedent. Customer financing by telecom equipment makers around 2000 is the case credit analysts most often cite, because losses mounted when financed customers failed. The comparison is a question to test, not a verdict: the outcome depends on the borrower’s revenue, the lender’s concentration, and the security behind the loan, none of which have been fully disclosed here.

    The Ripple Effect Beyond the Two Companies

    Chips are only one part of AI computing capacity. Leased accelerators still need data center space, cooling, networking and, above all, electricity. As reported, this facility is tied to chips, so operators and utilities serving any resulting deployment will still look to other parties for commitments on buildings and power.

    For competing chip suppliers and AI cloud providers, the deal sets a visible reference point. Customers negotiating large hardware orders may ask whether financing is on the table. One transaction does not establish an industry pattern, but it does show that supplier balance sheets are now part of how frontier AI compute gets paid for.

    Background

    Broadcom is a large US semiconductor and infrastructure software company. In AI infrastructure, it is best known for designing custom accelerator chips for major cloud and AI companies and for the networking silicon that links thousands of processors together. Custom chips are an alternative to general-purpose graphics processors, and they tie a customer to a specific design partner over multi-year development cycles.

    Anthropic, founded in 2021, builds the Claude family of AI models. Like other developers of frontier AI models, it needs enormous amounts of computing capacity to train and serve its models, which makes access to chips, and the financing to pay for them, a central business constraint.

    Sources

    Source: EXCLUSIVE: Broadcom to lend Anthropic up to $42 billion to lease its chips, filing says, Reuters’ report on Broadcom’s financing facility for Anthropic’s chip leases.

  • Only 190 of Amazon’s 690 Nuclear MW Are New. The Bigger AI Stake Is a 2034 License.

    Only 190 of Amazon’s 690 Nuclear MW Are New. The Bigger AI Stake Is a 2034 License.

    TL;DR · 30-second read

    The Short Version

    Amazon has signed a 20-year deal to buy about a third of the electricity from Calvert Cliffs, Maryland’s only nuclear power plant, which today produces enough for more than 1.3 million homes. The plant’s owner, Constellation Energy, says the deal unlocks more than $3 billion of investment.

    The twist: most of that power already exists. Just over a quarter of Amazon’s share will come from new equipment, due between 2030 and 2032. The plant’s two reactors have federal operating licenses that expire in 2034 and 2036, long before the contract ends. The power also stays on the shared grid rather than being walled off for Amazon.

    Constellation Energy and Amazon signed a 20-year agreement on Sept. 30, 2026 under which Amazon will buy 690 megawatts (MW) of output from the Calvert Cliffs Clean Energy Center in Maryland, Mugglehead Investment Magazine reported. The 690 MW includes a planned 190 MW capacity increase at the 1,790 MW station, expected to come online between 2030 and 2032. Constellation says the agreement is expected to enable more than $3 billion in Maryland infrastructure investment.

    Amazon said electricity from Calvert Cliffs will continue flowing to the PJM regional grid. The companies also signed a separate retail supply agreement covering Amazon operations across PJM’s 13 states and the District of Columbia. Constellation did not disclose a price per megawatt-hour or a spending schedule for the investment.

    Executive Summary

    Amazon, one of the hyperscalers (the largest cloud and AI computing operators), has committed to two decades of nuclear power from Maryland’s only nuclear plant. The contract covers 690 MW, roughly a third of the station’s output once a planned 190 MW uprate is complete. Constellation ties the deal to more than $3 billion of state investment, to relicensing the plant for another 20 years, and to potential new clean energy plants at the site.

    The deal is less about new megawatts than the headline suggests. About 500 MW of the contract comes from reactors already running, and the new 190 MW will not arrive until 2030 to 2032. The larger consequence is financial: a 20-year revenue stream for a station whose two reactor licenses expire in 2034 and 2036 strengthens the case for keeping all 1,790 MW running into the 2050s. The structure matters too. The plant stays connected to the shared grid instead of being dedicated to an Amazon campus built on site.

    What remains open is the economics. There is no disclosed price, start date, investment schedule, or relicensing application date. Constellation has also named no design or budget for the new plants it says the deal will help advance.

    Most of the 690 Megawatts Already Exists

    A megawatt measures generating capacity. Calvert Cliffs’ current 1,790 MW supplies the equivalent of more than 1.3 million homes and about 80 per cent of Maryland’s clean energy, by Constellation’s count. Amazon’s 690 MW contract includes the planned 190 MW increase, so roughly 500 MW of it is output from reactors that are already operating. The 190 MW is an uprate, meaning engineering changes that let existing reactors produce more power. It is not a newly ordered reactor.

    For anyone modeling new electricity supply for data center growth in PJM, the timing is the point. In the near term, the deal assigns existing carbon-free output to a named buyer. The net addition to the grid, about a tenth more capacity at the station, arrives between 2030 and 2032.

    Amazon’s earlier PJM nuclear agreement follows a similar pattern. Talen Energy’s June 11, 2025 contract covers 1,920 MW from the existing Susquehanna plant in Pennsylvania, reaching full quantity no later than 2032 and running through 2042. Talen treats any extra output from uprates as a separate possibility. Two contracts with one buyer do not make a market-wide trend. Still, both of Amazon’s disclosed PJM nuclear deals are anchored on reactors that already exist, with new capacity small or optional.

    Why the 2034 License Date Is the Bigger Stake

    The U.S. Nuclear Regulatory Commission (NRC) lists Unit 1’s operating license as expiring July 31, 2034 and Unit 2’s as expiring Aug. 13, 2036. Both licenses were renewed in 2000. A 20-year contract signed in 2026 runs into the mid-2040s at the earliest. If deliveries begin soon after signing, more than half of the term falls after Unit 1’s current license ends. The contract can be fully performed only if the NRC grants subsequent license renewal. That process extends reactor operation from 60 to 80 years and requires its own safety and environmental reviews.

    The arithmetic shows where the stake lies. The uprate adds 190 MW. Operation beyond the mid-2030s depends on relicensing for all 1,790 MW, more than nine times as much capacity. Constellation says the agreement helps provide the revenue certainty needed to relicense the plant for another 20 years. That is the mechanism: long-dated contracted revenue is what justifies the cost of a renewal application and the plant investment that goes with it. “Amazon’s commitment supports the long-term operation of Calvert Cliffs for generations to come,” chief executive Joe Dominguez said.

    Several parties depend on that outcome. Amazon needs the plant running to receive its contracted supply. Maryland gets most of its clean electricity from this one station. PJM would lose far more capacity if the plant closed than the uprate adds. A power sale does not renew a license, though, and Constellation has given no date for an application. The deal’s most consequential effect is a stronger financial case for keeping the station open. The regulatory decision is still years away.

    A Hyperscaler Pays for Nuclear Without Taking It Off the Grid

    Amazon said Calvert Cliffs’ electricity will keep flowing to PJM, the operator of the regional grid serving 13 states and Washington, D.C. That footprint includes Northern Virginia, the world’s largest data center market. The main alternative is co-location, in which a data center is built at the plant and draws power directly, “behind the meter,” before it reaches the grid. Co-location has drawn regulatory scrutiny in PJM over its effect on other customers.

    Calvert Cliffs instead uses what the industry calls a grid-side power purchase agreement (PPA), a long-term contract to buy a plant’s output. The plant sends its electricity into the grid, and Amazon’s facilities are served through a separate retail supply agreement. Under a structure like this, the buyer underwrites the plant’s revenue while its output stays available to the regional system. That reduces the friction co-location attracts.

    The trade-off is that Amazon’s sites still depend on the grid’s ability to deliver power where its data centers are located. The nuclear contract secures a source of electricity, but it does not settle questions of local grid connections or transmission capacity.

    The $3 Billion Figure Still Needs a Schedule

    Constellation says the deal will “enable” more than $3 billion of Maryland investment. That figure is not described as Amazon’s direct spending, and no spending timeline was given. It is also unclear how the sum divides among the uprate, relicensing work, and the potential new clean energy plants, for which Constellation named no design, equipment order, budget, or completion date.

    Without a price per megawatt-hour, outside observers cannot compare the contract with market power prices or judge how much of the plant’s future revenue it locks in. Constellation’s shares rose after the announcement, which suggests investors value the long-term revenue certainty. Overall, the announcement is concrete on duration, volume, and the uprate window, and thin on economics.

    Background

    Constellation Energy, spun off from Exelon in 2022, operates the largest fleet of nuclear plants in the United States. Calvert Cliffs, on Maryland’s Chesapeake Bay shore, has two reactors totaling 1,790 MW. The company is also working to restart an 835 MW reactor in Pennsylvania under a separate long-term agreement with Microsoft. That unit received an NRC environmental finding on Sept. 25, 2026.

    Since 2024, hyperscalers have turned to existing nuclear plants for around-the-clock, carbon-free power for cloud and AI data centers. Many of these deals sit inside PJM, the grid serving 13 states and Washington, D.C., where data center demand has grown fastest. Amazon’s 1,920 MW contract with Talen Energy’s Susquehanna plant, signed in June 2025, is the closest comparable.

    Sources

    Source: Constellation Energy signs 20‑year Amazon deal for 690 MW at Calvert Cliffs (Mugglehead Investment Magazine): report on the Sept. 30, 2026 Constellation–Amazon power purchase agreement covering Calvert Cliffs’ output and a planned 190 MW uprate.

  • AI Data Center Water Risk Runs Through 50,000 Utilities’ Cyber Defenses

    AI Data Center Water Risk Runs Through 50,000 Utilities’ Cyber Defenses

    TL;DR · 30-second read

    The Short Version

    A cybersecurity company called Cyware is teaming up with WaterISAC, the group that helps American water utilities warn each other about hackers.

    The goal is simple: when one water system spots an attack, the warning should reach the others fast, including thousands of tiny rural systems with little or no security staff.

    Why it matters beyond your tap: many of the huge computer buildings that power artificial intelligence rely on the local water supply to stay cool and to fight fires. If the local water plant is hacked, those buildings are exposed too. No price or timeline was announced.

    Cyware announced on Sept. 30, 2026, in a release distributed via PR Newswire, a partnership with WaterISAC, the information-sharing organization for the US water sector. Under the arrangement, WaterISAC will use Cyware’s threat intelligence platform to collect, enrich and distribute cyber threat data and vulnerability alerts to water and wastewater utilities.

    WaterISAC said its existing partnership with the National Rural Water Association already lets it serve more than 20,000 of the sector’s smallest utilities, and that Cyware was selected after an evaluation of available platforms. No financial terms, contract length or rollout schedule were disclosed.

    Executive Summary

    WaterISAC, the hub through which US water utilities, security experts and government partners share threat warnings, is moving its intelligence work onto Cyware’s platform. The stated aim is to replace manual processes for gathering and enriching threat data with automation, so that alerts reach members faster and larger utilities can receive intelligence directly.

    The deal matters because the US water sector is extraordinarily fragmented. The release counts more than 50,000 utilities, many of them small, and argues that each one is a potential point of failure. Collective defense, the idea that a threat seen by one member protects all of them, only works if warnings reach the weakest participants in a form they can act on.

    For data center operators, the announcement is a reminder that a facility’s physical resilience depends on infrastructure it does not control. Cooling systems that draw on municipal water, and fire-suppression systems fed by the local supply, inherit the cyber posture of whichever utility serves the site.

    Why a Water-Sector Deal Belongs on a Data Center Risk Register

    Data center resilience planning usually focuses on what sits inside the fence: backup generators, redundant cooling loops, dual fiber paths. Water is often treated as a commodity input. But many facilities, especially those using evaporative or other water-based cooling, draw on the local municipal supply, and building fire-suppression systems commonly depend on it as well. The Cyware release itself lists fire suppression alongside drinking water and sanitation as a core function of water systems.

    That dependency runs through operational technology, or OT: the control systems that run pumps, valves, chemical dosing and treatment at a water plant. Cyware’s release states that adversaries targeting OT now move at machine speed. If a utility’s controls are disrupted, pressure and supply can be affected for every customer on that system, including a data center campus. The operational conclusion is that the relevant cyber defender for a site’s water is not a national agency or a large vendor, but one of more than 50,000 local utilities, each with its own level of security staffing and maturity.

    For operators and developers, that makes the local utility’s cyber posture a legitimate site-selection and resilience input, weighed alongside on-site water storage and cooling design. It also means that sector-wide efforts like WaterISAC’s affect data center risk indirectly: a better-informed local utility is a more dependable upstream supplier.

    The Small-Utility Problem Is the Real Test

    The most revealing detail in the announcement is structural. WaterISAC says its partnership with the National Rural Water Association already lets it serve more than 20,000 of the smallest utilities. WaterISAC Executive Director Tom Dobbins described the Cyware platform as helping WaterISAC’s analysts deliver more real-time insights to members and letting the organization extend intelligence directly to larger utilities.

    Read together, those statements suggest a tiered model: analysts work in the platform, larger utilities may connect directly, and smaller utilities receive curated output. That is a reasonable design, since a rural system with no dedicated security staff cannot ingest a raw threat feed. But it also means the speed gains Cyware describes will be felt most by larger members and by WaterISAC’s own analysts. Whether a warning reaches every participant regardless of size, as Cyware’s Tom Stockmeyer framed the goal, depends on how that last step is handled.

    What the AI Claims Establish, and What They Do Not

    Cyware positions itself around agentic AI, meaning software that automates steps an analyst would otherwise perform, such as collecting, enriching and routing threat data. The release says this will help WaterISAC significantly reduce manual effort and deliver intelligence more efficiently. Those are plausible benefits for an intelligence team, but the release offers no baseline, no target metrics and no description of how automated outputs are validated before distribution.

    It is also worth separating intelligence speed from response speed. A warning delivered in seconds helps only if the recipient can act on it. For a small utility, the bottleneck is often people and budget rather than information. The partnership addresses the information side; the response side remains with each utility. Cyware has also said it is supporting NRECA on Department of Energy-funded research into automated collective defense for electric cooperatives, which places the company across both utility inputs that data centers depend on most: power and water.

    Background

    The US water sector is unusually fragmented: more than 50,000 utilities, ranging from large metropolitan systems to small rural operations with minimal staff. WaterISAC serves as the sector’s information-sharing hub, one of a family of industry-specific ISACs that pool threat warnings so members can learn from one another. Through its partnership with the National Rural Water Association, WaterISAC says it reaches more than 20,000 of the smallest utilities.

    Cyware is a threat intelligence software vendor that unifies intelligence management, sharing and collaboration, and orchestration and automation. It markets itself around agentic AI and collective defense, and counts enterprises, government agencies and ISACs among its customers. The WaterISAC deal extends that ISAC-focused strategy into a sector whose resilience underpins other infrastructure, including data centers that rely on local water for cooling and fire protection.

    Sources

    Source: Cyware and WaterISAC Join Forces to Strengthen Water Sector Cyber Defense, Cyware’s Sept. 30, 2026 announcement of its threat intelligence partnership with WaterISAC.

  • A 100 MW Space-Solar PPA for AI Is Signed. Its Bankability Hinges on a 2027 Demo.

    A 100 MW Space-Solar PPA for AI Is Signed. Its Bankability Hinges on a 2027 Demo.

    TL;DR · 30-second read

    The Short Version

    A startup called Virtus Solis has signed a 20-year deal to sell electricity collected by satellites in space and beamed down to a receiving station it plans to build in Illinois.

    The buyer, Brae Systems, runs underwater computer centers for artificial intelligence. The starting amount of power, delivered for a full year, would cover the electricity use of tens of thousands of American homes.

    The catch: the beaming technology has not yet been publicly shown working. Its first public demonstration is scheduled for March 2027. That test, not the signature, will show whether the deal can happen.

    Virtus Solis Technologies and Brae Systems announced on September 30, 2026, from Troy, Michigan, and Chicago, that they have closed a 20-year power purchase agreement (PPA) under which Virtus Solis will supply 100 megawatts (MW) of round-the-clock, space-based solar power, or 876,000 megawatt-hours (MWh) a year, through a dedicated receiving station to be built in Illinois. Brae holds an option to expand to 250 MW (2.19 million MWh a year) within three years of initial commercial operations. The companies describe the deal as the world’s first commercial space-based solar PPA and a multi-billion-dollar lifetime gross contract commitment.

    The electricity is intended for Brae’s subsea GPU data centers. Virtus Solis says it will hold its first public technical demonstration in March 2027, is supported by a $2 million corporate development agreement with ARPA-E, and expects this to be the first of a series of commercial offtake agreements over the next several months.

    Executive Summary

    The agreement pairs two unconventional approaches to AI infrastructure: satellites that collect sunlight in orbit and beam it to Earth as radio-frequency energy, and data centers placed underwater so surrounding water does the cooling. Brae gets a 20-year commitment of firm, carbon-free power, while Virtus Solis gets a named, long-term buyer, the kind of contracted revenue energy developers use to raise capital.

    The significance lies in the sequencing. The contract is signed before the technology’s first public demonstration, and the release itself names March 2027 as the next major proof point for Virtus Solis’s unit economics. No commercial operations date, price, launch schedule or financing plan is disclosed. The PPA therefore sets up a dated test of whether orbital baseload power can move from a signed offtake to a financed, operating asset.

    For AI infrastructure planners, the deal is worth watching less as a near-term power source than as a signal: developers facing grid constraints are willing to contract for power from technologies that do not yet operate commercially.

    A Contract Signed Ahead of the Hardware

    A power purchase agreement is a long-term contract in which a buyer commits to purchase electricity from a specific generator on agreed terms. In conventional energy, the PPA usually comes first: a developer secures a creditworthy buyer, then uses that contracted revenue to raise the debt and equity needed to build. The Virtus Solis–Brae agreement follows the same sequence, but at an earlier stage than is typical, because the generating asset is an orbital solar system that has yet to hold its first public technical demonstration.

    The release gives some anchors: a 20-year term, a receiving station in Illinois, and a March 2027 demonstration. It does not give a commercial operations date, a satellite launch schedule or a price per megawatt-hour. Even the expansion option is keyed to “initial commercial operations,” a milestone whose date is not stated.

    Virtus Solis says more offtake agreements will follow over the coming months. For an early-stage generator, stacking contracted demand is a rational strategy, since each signed buyer strengthens the case to investors. The value of that stack, however, rests on the performance of the technology underneath it.

    Why Beam Efficiency, Not the Signature, Sets Bankability

    Space-based solar power collects sunlight on satellites, converts it into radio-frequency energy, beams it to a ground receiving station, and converts it back into electricity. Losses at each conversion step determine how much collected energy actually reaches the customer, and therefore how many satellites, launches and dollars are needed per megawatt delivered. That end-to-end figure is known as wireless power transfer (WPT) efficiency, and it is the central variable in the economics of the whole asset class.

    The release itself identifies this as the hinge. It calls the March 2027 demonstration “the next major proof point” for Virtus Solis’s unit economics (its cost per unit of power delivered) and says the company expects to show record WPT efficiencies. The $2 million ARPA-E agreement (ARPA-E is the US Department of Energy’s high-risk energy research agency) supports development of “a new class of power electronics” that orbital generation requires. In the company’s own framing, the cost basis is still being established.

    Infrastructure lenders typically underwrite two separate questions: can the buyer pay, and can the asset produce? A 20-year PPA speaks to the first. The second remains open until the efficiency figures are demonstrated publicly, which is why the contract’s practical meaning for financing arrives with the 2027 demonstration rather than with this week’s announcement. The outcome affects Virtus Solis’s ability to raise construction capital, Brae’s long-range power planning, and every subsequent buyer weighing a similar offtake.

    What the Numbers Do and Don’t Establish

    The annual energy figures are exact multiples of capacity: 100 MW running every one of the 8,760 hours in a year equals 876,000 MWh, and 250 MW equals 2.19 million MWh. The stated volumes therefore assume full output around the clock. The release separately cites a “guaranteed 99.9% uptime,” which would allow roughly 8.8 hours of downtime a year, so the headline volumes are best read as a nameplate ceiling rather than a delivery forecast.

    The “multi-billion-dollar” lifetime value can be bounded, though not pinned down, because no price is disclosed. The base 100 MW over 20 years totals about 17.5 million MWh; reaching even $2 billion on that volume alone would require an average price of roughly $114 per MWh. At the full 250 MW for 20 years, about 43.8 million MWh, $2 billion corresponds to roughly $46 per MWh. Either the contracted price is relatively high, or the multi-billion characterization assumes Brae exercises its expansion option, which is Brae’s choice to make.

    None of this contradicts the companies’ statements. It clarifies that the most prominent figures, 250 MW and multi-billion dollars, describe the contract’s upper bound rather than its committed base.

    The Case for Brae, and the Delivery Path

    Brae’s rationale tracks genuine pressures on AI infrastructure. Clusters of GPUs, the chips that train and run AI models, draw large, constant electrical loads and shed significant heat, and in many regions new data centers wait years for grid connections. Brae’s model addresses cooling by submerging compute modules, and the PPA addresses supply with power that does not depend on the local grid’s generation mix. CEO Vishnu Indukuri said the agreement gives Brae “predictable power costs and zero carbon emissions, completely insulated from terrestrial grid curtailment.”

    The geography raises a question the companies have not yet answered. Power is to arrive at a receiving station in Illinois, while Brae’s compute is subsea, and the release does not say where Brae’s modules will sit or how electricity will travel from the station to them. If delivery runs over existing transmission lines, the degree of insulation from curtailment (grid operators ordering generators or loads to reduce output) will depend on how that transmission is contracted.

    If the technology performs, the beneficiaries are AI operators seeking firm, clean power without waiting in grid interconnection queues. If timelines slip, Brae will need other supply in the interim, and the release does not describe a backup arrangement.

    Background

    Space-based solar power has been studied since the late 1960s. Its appeal is that sunlight in orbit is not interrupted by clouds or night in the way ground-based panels are; its obstacles have been the cost of launching hardware and the efficiency of beaming energy to Earth. Virtus Solis Technologies, based in Troy, Michigan, is pursuing the concept with mass-manufactured modular satellite arrays and radio-frequency wireless power transmission to terrestrial receivers.

    Underwater data centers have also been tested before, most visibly in Microsoft’s Project Natick, which operated a sealed server module on the seabed off Scotland’s Orkney Islands. Brae Systems is commercializing the approach for dense GPU clusters, arguing that water cooling and freedom from land constraints suit the heat and power demands of AI workloads.

    Sources

    Source: Virtus Solis Signs Landmark 20-Year Space Solar PPA with Brae Systems to Deliver 250 MW of Orbital Baseload Clean Power, the companies’ joint announcement of the 20-year orbital solar supply agreement for Brae’s subsea GPU data centers.

  • Talen’s $1,208M in PJM Capacity Revenue: AI Forecasts Pay Before Servers Arrive

    Talen’s $1,208M in PJM Capacity Revenue: AI Forecasts Pay Before Servers Arrive

    TL;DR · 30-second read

    The Short Version

    Talen Energy, a company that owns power plants including a large nuclear station in Pennsylvania, has lined up about $1.2 billion in payments from the grid operator serving much of the eastern United States.

    The money is not for electricity sold. It pays Talen to keep its plants ready for the grid’s busiest days. Those payments are set in advance, and they have climbed because the grid expects many new artificial intelligence data centers. So power plants get paid before many of those data centers exist.

    Electricity customers ultimately cover the bill. The same week, Talen named a new chief executive and a $1.5 billion plan to buy back its own shares.

    ScanX reported that Talen Energy (NASDAQ: TLN), the Houston-based independent power producer, has secured $1,208 million in capacity revenues from PJM Interconnection, the operator of the largest power grid in the United States. Capacity revenues are payments a generator receives for committing its plants to be available during a future delivery year, separate from what it earns selling electricity.

    The figure lands the same week as an 8-K filed September 29 in which Talen named President Terry Nutt as chief executive effective January 1, 2027, succeeding Mac McFarland, and announced $1.5 billion of accelerated share repurchases inside an upsized $3.0 billion buyback authorization running through the end of 2028.

    Executive Summary

    Talen’s $1,208 million in PJM capacity revenue is money earned for availability, not output: PJM pays generators a fixed price for each megawatt they commit to have ready during a delivery year, and the size of that payment is set by an auction that buys enough supply to cover forecast peak demand. That forecast has been pushed sharply higher by expected data-center growth, much of it tied to artificial intelligence.

    The practical consequence is that AI demand is already turning into cash for existing merchant power plants, ahead of the data centers it anticipates. For Talen, which sells into wholesale markets rather than earning regulated rates, a large, pre-set revenue line behaves much like a contract. In the same week, the company’s board committed $1.5 billion to an accelerated share repurchase that Talen said would retire more than 10% of its shares at the current stock price, and handed the chief executive role to its former chief financial officer.

    Talen has not explicitly linked the capacity revenue to the buyback, and important details, including delivery period, cleared megawatts and plant mix, remain open. But taken together, the two announcements illustrate how the AI power story is flowing through grid market design and into generator balance sheets.

    Paid on the Forecast, Not the Plug-In

    PJM’s capacity market works through an auction, known as the Base Residual Auction, held ahead of each delivery year. PJM estimates how much generation it needs to cover forecast peak demand plus a safety margin, then buys that commitment from power plants. Every plant that clears receives the auction price for each megawatt, for every day of the delivery year, as long as it is available when called. Total revenue is simple arithmetic: cleared megawatts times the price times the days in the year. Talen’s $1,208 million is the product of that arithmetic applied to its fleet.

    The demand side of that auction is a forecast, and PJM’s recent load forecasts have been revised upward largely on expected data-center growth. PJM’s independent market monitor has attributed a large share of recent capacity price increases to forecast data-center demand, and prices for the 2026/27 and 2027/28 delivery years reached a temporary ceiling approved by federal regulators. In other words, a meaningful part of what generators are being paid reflects servers that are planned, not yet running.

    That is the operational point. Once an auction clears, the commitment and the price are locked for that delivery year. If some forecast data-center campuses slip or are never built, the payments still flow. The forecast risk sits largely with the utilities and customers who fund capacity through their bills, while generators such as Talen hold a cleared price. This is how a fleet anchored by the Susquehanna nuclear plant can bank revenue tied to the AI buildout without signing a data-center contract for every megawatt it owns.

    Contract-Like Cash for a Merchant Fleet

    Talen is a merchant generator: it sells power at market prices rather than under regulated utility rates, so its energy revenue rises and falls with natural gas prices, weather and demand. Capacity revenue is different because it is known before the delivery year begins. For lenders and shareholders, that predictability makes it behave more like a contract than a commodity bet.

    Talen’s September 29 disclosures show how it is deploying its cash. The company entered $1.5 billion of accelerated share repurchases, in which a bank delivers shares up front and the final count is settled later based on the average stock price. Talen said the ASRs are uncollared, meaning there is no price band limiting how many shares the fixed dollar amount ultimately buys, and that at the current stock price they would retire more than 10% of shares outstanding by the end of the first quarter of 2027. The board also raised remaining buyback capacity to $3.0 billion through December 31, 2028, leaving $1.5 billion beyond the ASRs.

    Nutt described the ASRs as buying stock at “an attractive free cash flow yield” while “preserving liquidity and strategic flexibility.” Talen did not tie that decision to the capacity figure in its release. Still, a large revenue line fixed in advance is exactly the kind of visibility that lets a board commit a fixed dollar amount to an uncollared buyback with confidence.

    The Price-Signal Debate, Asked of Both Sides

    Generators and PJM have argued that high capacity prices are a legitimate scarcity signal: they keep existing plants from retiring and encourage new supply as demand grows. Consumer advocates and state officials, including Pennsylvania’s governor, whose challenge preceded the temporary price cap and floor, have argued that customers are paying elevated prices on forecasts that may prove too high, and that payments to existing plants do not add new supply quickly.

    Both positions deserve pointed questions. For generators: how much of the elevated capacity revenue is going into new or uprated megawatts, rather than returned to shareholders? Talen’s most recent capital-allocation announcement is a buyback, not a construction program. For critics: if prices were held well below scarcity levels, what would keep existing plants running as load grows, and how confident can anyone be that a lower forecast would be more accurate? The reliability of large-load forecasts has itself become contested, because a single data-center project can appear in more than one utility’s plans.

    A Finance-Led Succession for a Capital-Allocation Story

    The incoming chief executive’s résumé fits the business the capacity figure describes. Nutt, 50, has been Talen’s President since December 2025 and was chief financial officer from July 2023 through December 2025. Before Talen he was chief financial officer of EDF Trading North America and held risk-management and controller roles at Vistra. Capacity auctions, hedging and capital returns are core skills in that background.

    The board framed the transition as continuity in strategy, operations and capital allocation. McFarland stays chief executive through December 31, 2026, then serves as senior advisor until his retirement on March 1, 2027. Rubric Capital Management, which described itself as Talen’s largest actively managed shareholder, and Energy Capital Partners both publicly backed the choice, signaling that the shareholder-return emphasis is likely to persist.

    Background

    Talen Energy is a Houston-based independent power producer whose fleet is anchored by the Susquehanna nuclear plant in Pennsylvania, one of the largest nuclear stations in the United States. The company emerged from a Chapter 11 restructuring in 2023 and has since become one of the most closely watched power stocks of the AI era, after selling a data-center campus adjacent to Susquehanna to Amazon and later expanding its power-supply relationship with the company.

    PJM Interconnection runs the grid and wholesale market for 13 states and Washington, D.C. Its capacity market, which pays plants to be available rather than only for power they produce, has become a focal point in the debate over who should bear the cost of AI-driven load growth, after recent auctions cleared at a temporary ceiling approved by federal regulators.

    Sources

    Source: Talen Energy secures $1,208 million in PJM capacity revenues (ScanX) reports Talen’s capacity revenue from PJM Interconnection.

    Primary sources: Talen Energy Form 8-K filed September 29, 2026 (SEC EDGAR); Talen Names Terry Nutt Chief Executive Officer and Announces $1.5 Billion Accelerated Share Repurchase, Exhibit 99.1 (SEC EDGAR).