Tag: On-Site Power

  • At Nebius’s 300 MW AI Site, Bloom’s First Power Block Is Waiting on the Building

    At Nebius’s 300 MW AI Site, Bloom’s First Power Block Is Waiting on the Building

    TL;DR · 30-second read

    The Short Version

    The electricity showed up before the computers did. Bloom Energy, which builds fuel cells (machines that turn natural gas into electricity without burning it), says the first slice of power is ready at a huge artificial intelligence data center being built in Vineland, New Jersey.

    It took about five months from signing the deal. The building it will feed is not open yet.

    The full site is planned to use as much electricity as a mid-sized city. It is part of a deal to supply computing power to Microsoft. The question now is whether the rest of the power, and the building, arrive just as fast.

    Bloom Energy (NYSE: BE) has the first fuel-cell power block ready for customer use at Nebius’s (NASDAQ: NBIS) 300-megawatt data center in Vineland, New Jersey. Founder and CEO KR Sridhar said so on the company’s October 22 third-quarter earnings call, Blockspace Media reported. The data center itself is not yet online. Nebius is bringing the site up in phases, and Sridhar said its first block is “almost getting ready.” The two companies signed their agreement on May 14. According to a Nebius filing, the contract covers about 250 MW of guaranteed capacity from about 328 MW of installed fuel cells. It runs across three expected phases with 10-year supply terms, and monthly service fees total up to $2.6 billion.

    The same day, Bloom reported record third-quarter revenue of $1,235.8 million, up 138.1% year over year, and raised its full-year 2026 revenue guidance to $4.4–4.6 billion. Nebius separately disclosed a $3.5 billion, 364-day senior secured bridge facility from Citi to fund data center development and GPU procurement.

    Executive Summary

    Bloom Energy says it has taken on-site power for Nebius’s Vineland AI campus from a signed contract to delivered capacity in about five months. By Sridhar’s account, that was roughly three months to win local planning approval and two more to have the first block ready. The site is tied to Nebius’s multiyear agreement with Microsoft. Bloom’s equipment replaces an earlier plan for 32 natural-gas engines that drew resident objections over emissions.

    The significance is the order in which things arrived. For the first phase at Vineland, electricity is no longer the item that decides when AI servers switch on. The data hall is. That reverses the usual story of AI construction, in which buildings wait on power. It also shifts the remaining schedule and financing pressure toward the data hall and GPU side of the project. Nebius is funding that side partly with a 364-day bridge loan.

    The claim covers one block. Bloom has not said how large that block is or when the remaining phases follow. Vineland is now a reference project for fuel-cell power at AI scale, and the next two phases are where it gets tested.

    Power Arrived First at Vineland

    The timeline is unusually specific. Nebius and Bloom signed on May 14. On August 17, Vineland’s Planning Board approved developer DataOne’s 587,980-square-foot Phase 2 expansion. The approval included multiple fuel-cell areas and a tank for liquefied natural gas (LNG), which is gas chilled to a liquid so it can be stored on site. By Bloom’s October 22 earnings call, Sridhar said the first block of power was “ready for customer use.” That is roughly five months from contract to first power for on-site generation at an AI campus.

    The building that power will serve is not online yet. Sridhar said Nebius brings the 300-megawatt site up “block by block” and that its first block “is almost getting ready.” He added: “we are already there for the first block.” A megawatt is a million watts, and a site this size draws the electricity of a mid-sized city. In construction terms, the pacing item is whatever sets the date when GPU servers start drawing current. For phase one, that is no longer electricity. It is the data hall itself: the shell, cooling, electrical fit-out, and the delivery and installation of GPU servers. That reverses the usual framing of AI buildout, in which power is the scarce input and the building waits on it.

    The claim has limits. “Ready for customer use” is Bloom’s own description, given on its own earnings call, and Bloom has not said how many megawatts the first block holds. One block at one campus is a data point, not an industry trend. Still, the near-term schedule question at Vineland now sits on the data hall side, not Bloom’s. That matters to Nebius, to Microsoft as the customer behind the campus, and to the lenders behind the power plant.

    Three Months of Approval, Two Months of Build

    Sridhar’s own split of the timeline is revealing. Planning board approval took about three months, and delivering the first block took two more. On his figures, local approval took longer than installation. That matches the risk list in Bloom’s third-quarter investor presentation. It names “air permitting delays which affect the ability to finance the use of Energy Servers” alongside utility interconnection and other installation delays.

    Vineland’s history points the same way. The site’s earlier plan called for 32 natural-gas engines, and residents objected over emissions. The fuel-cell design replaced it. Bloom’s units make electricity from natural gas through an electrochemical reaction rather than by burning it. The expansion was then approved in August. That sequence does not prove the technology change caused the approval. It does show that the choice of generator and its local acceptance were tied together, and that the zoning hearing can take longer than the equipment.

    Behind-the-meter power means generating electricity at the campus instead of drawing it from the utility grid. For developers weighing it, the practical lesson from Vineland is about where schedule risk sits. The local permit and the fuel plan matter at least as much as the hardware.

    Power on a Service Contract, GPUs on a Bridge Loan

    Nebius is buying power as a service rather than buying a power plant. Under the contract, it pays monthly fees of up to $2.6 billion over the term, and Bloom installs, operates and maintains the systems. Bloom puts the project investment at $1.7 billion. Industrial Development Funding is the lead developer, and Oaktree Capital Management supplies infrastructure capital. Morgan Stanley structured the tax-equity financing and is its only investor; tax equity is capital from investors who use a project’s tax credits. MUFG Bank led the senior debt. The capital for the generators is long-dated and spread across several firms that specialise in project finance.

    Nebius’s own financing for the building and the servers is shorter-dated. A foreign issuer files a 6-K as its current report to US regulators. In one dated October 22, Nebius disclosed a $3.5 billion senior secured bridge facility from Citi. It was signed October 17 and matures October 16, 2027. The interest rate is Term SOFR, a benchmark lending rate, plus 2.75%, rising 0.25% every 90 days to a 3.50% cap. In the accompanying press release, Nebius said proceeds go to “data center development and GPU procurement.” COO Ophir Nave said the company can still draw on “asset-backed financing based on investment-grade customer commitments of more than $40 billion of long-term contracts.”

    The facility funds Nebius’s global build-out, not Vineland specifically. Even so, the contrast is clear. At Vineland, the power is funded through multi-party project finance with 10-year supply terms. The data halls and GPUs Nebius is building draw partly on 364-day money with a rising margin. Neither structure is a fault. The point is that, with power ahead for phase one, the campus’s remaining sensitivity to cost and timing sits mostly on Nebius’s side of the ledger.

    What Blocks Two and Three Have to Prove

    Sridhar’s promise went beyond the first block: “We will not be the bottleneck to them turning on their data center their entire 300 megawatts.” The disclosed contract guarantees about 250 MW from about 328 MW of installed units, and the companies have not reconciled that with the 300-megawatt site figure. The installed capacity exceeds the guaranteed output by roughly 78 MW. The gap between 250 MW and 300 MW could be explained by a grid connection, by a different way of measuring the site, or by that installed headroom.

    Bloom has the scale to try. Third-quarter product revenue rose 179.4% to $1,073.6 million. Gross margin reached 33.8%, up 453 basis points (4.53 percentage points). CFO Simon Edwards called the guidance increase the third this year. Bloom’s own risk factors, however, include its ability to expand production capacity, supply constraints, and “the availability of financing for data center developments.” If phases two and three arrive at the pace of the first, Vineland strengthens the case that on-site fuel cells can keep up with AI campus schedules. If they slip, the five-month figure stays a single-block result.

    Background

    Bloom Energy, based in San Jose, California, makes fuel-cell systems it calls Energy Servers. They generate electricity on site from natural gas without combustion. The company has increasingly sold to hyperscalers, the largest cloud and AI computing operators, and to data center developers who need power faster than local utilities can connect it. Bloom joined the S&P 500 in 2026 and reported record third-quarter results on October 22.

    Nebius Group, headquartered in Amsterdam and listed on Nasdaq, is an AI cloud company. It rents GPU computing capacity to developers and enterprises and is expanding its data center footprint worldwide. Its Vineland, New Jersey, campus is being developed by DataOne and is tied to Nebius’s multiyear agreement with Microsoft. As a foreign private issuer, Nebius reports material events to US regulators on Form 6-K.

    Sources

    Source: Bloom Energy says first power block is ready at Nebius’s 300 MW New Jersey data center (Blockspace Media): Bloom’s CEO on first fuel-cell power at Nebius’s Vineland AI campus and the contract’s terms.

    Primary sources: Nebius Group Form 6-K, October 22, 2026 (Citi bridge facility); Nebius secures $3.5 billion financing facility with Citi (Exhibit 99.1); Bloom Energy Form 10-Q for the quarter ended September 30, 2026; Bloom Energy Form 8-K, October 22, 2026; Bloom Energy Reports Record Third Quarter 2026 Financial Results and Raises Full Year 2026 Guidance (Exhibit 99.1); Bloom Energy Q3 2026 supplemental investor presentation (Exhibit 99.2).

  • Talen-Backed 29-Building Campus Shows Why Data Center Power Plans Wait on Zoning

    Talen-Backed 29-Building Campus Shows Why Data Center Power Plans Wait on Zoning

    TL;DR · 30-second read

    The Short Version

    A power company, Talen Energy, and a business partner want to build up to 29 huge computer buildings on about 900 acres in a small Pennsylvania township. Buildings like these run cloud services and artificial intelligence.

    The largest version would have floor space equal to about 60 football fields. It would also have its own power plant. The developers have offered the township up to $50 million in community payments.

    The local zoning board did not rule on the project. It gave the township’s elected leaders 60 days to act on rules for projects like this. A local deadline, not the power supply, now sets the schedule.

    The Upper Mount Bethel Township Zoning Hearing Board in Northampton County, Pennsylvania, voted 4-1 on Monday, Oct. 19, to postpone a hearing on Slate Belt Holdings’ challenge to the township’s zoning ordinance. The hearing now resumes Dec. 17. In the meantime, the board gave township supervisors 60 days to take action on data center regulation, The Morning Call reported. Slate Belt Holdings is a partnership between Talen Energy and River Pointe Logistics.

    The challenge covers 23 data center buildings and two substations on 754 industrially zoned acres west of River Road. A second challenge, filed Oct. 8 and due to be heard by November, covers six more buildings, battery storage and a power plant on roughly 152 acres east of the road. Land development plans filed Oct. 13 show two layouts across the roughly 900 acres. One has 29 buildings totaling 3.5 million square feet. The other has six buildings totaling nearly 1.5 million square feet.

    Executive Summary

    The postponement does not decide anything. It does move the project’s next date onto a township calendar. The zoning board chose not to rule on whether the township’s ordinance is valid as applied to data centers. Instead it handed the issue back to the elected supervisors. According to the board’s solicitor, they are expected either to advance the long-delayed data center ordinance or to act on the memorandum of understanding the developer has proposed. A memorandum of understanding is a negotiated, non-final agreement between the parties.

    This matters beyond one township because of what the project bundles. The plan pairs hyperscale data center buildings with two substations, battery storage and an on-site power plant. Hyperscale means very large facilities built for cloud and large-scale computing customers. Developers increasingly treat that combination as the scarce asset in data center buildout. Here it is held up at an earlier gate: whether local land-use law permits the buildings at all, and on what terms. Two and a half years after the township’s zoning officer was first asked that question, the township has not passed a data center ordinance.

    The Gate Is a Township Ordinance, Not a Power Plant

    Power is often described as the binding constraint on large data center projects. That means grid connections, substations and on-site generation. Slate Belt’s plan addresses power directly. The western parcel includes two substations, which step high-voltage transmission power down for use on site. The eastern parcel adds battery storage and a power plant. Having a power producer as a partner signals that the developers see energy supply as part of the product, not something they hope to secure later.

    None of that sets the timetable right now. The next operative date is Dec. 17, and a five-member zoning hearing board chose it. The legal question is narrow but decisive. Slate Belt’s attorney, Shawn Gallagher, argues that an April 2024 determination by the township’s zoning officer makes a data center a use permitted by right. That would mean approval without discretionary review. The township’s newly hired solicitor, Bill Jones, counters that the April 2024 plans did not include square footage. Under the township’s ordinances, total square footage determines whether a developer needs a special exception, which is a discretionary approval with its own hearing and conditions.

    That is why the two layouts matter. The gap between the 29-building, 3.5-million-square-foot option and the six-building, roughly 1.5-million-square-foot option is about 2 million square feet. On the township’s reading, that kind of size difference changes the approval path. For anyone who builds, supplies or finances a campus like this, the scale of the investment depends on how a township defines a land use. That is the mechanism behind the headline. Securing power and securing permission are separate tracks, and the slower one sets the pace.

    What $50 Million Has and Has Not Bought

    Slate Belt’s August proposal offered up to $50 million in donations. In exchange, the township would end the legal fight, work with the developer on data center zoning changes and eventually approve its plans. An updated $50 million community benefits agreement came with the Oct. 13 filing. Community benefits agreements are common in large developments. They package payments or commitments to a host community, usually in return for cooperation on approvals.

    The offer is substantial, but it is conditional by design. As described, the money depends on outcomes the township would have to deliver, including zoning changes and approvals. That explains why it has not settled the question. Supervisors have split on strategy. Some argue that negotiating directly with the developer is the best way to win environmental protections. Others want to pass the township’s own ordinance first. Several versions of the deal have gone back and forth without a resolution. The zoning board’s move now puts both paths on the same 60-day clock.

    For developers watching from elsewhere, one lesson is clear from this case. A large benefits package can open a negotiation, but it does not replace a legal basis for approval. Where the underlying ordinance is unsettled, the money becomes one bargaining chip among several. It does not shortcut the process.

    A Deadline With an Undefined Ask

    The 60-day instruction is firm on timing and vague on substance. Ron Angle, the alternate member who made the motion, said he expected supervisors to “do the job that the citizens of this township elected them to do.” Supervisors Chair John Bermingham told The Morning Call he did not know what he was supposed to do in 60 days. Solicitor Chad DiFelice’s reading, either advance the ordinance or act on the memorandum, is the clearest guide available. It still leaves supervisors to choose between two strategies they have been divided on.

    The process is also gaining participants. The Delaware Riverkeeper Network was granted standing on behalf of 27 members who live within half a mile of the proposed campus. One non-member resident in that radius was also admitted. Standing gives these parties formal rights to present evidence and argument. A second validity challenge covering the eastern parcel, which holds the power plant, is scheduled separately for November. The developer’s spokesperson said that challenge seeks to clarify how zoning applies to the full buildout.

    The supervisors’ draft ordinance went to the Lehigh Valley Planning Commission for review on Sept. 21. An earlier March 2025 draft, written after Amazon showed interest in a data center deal, was never finalized. Each delay so far has carried a cost for everyone involved. The developer has faced litigation and uncertainty. Residents have had to organize legal representation. Supervisors have had to defend an ordinance whose application to this use is still disputed.

    Background

    Upper Mount Bethel is a township in the Slate Belt region of Northampton County, in Pennsylvania’s Lehigh Valley along the Delaware River. The question of how its zoning ordinances apply to data centers dates back two and a half years, to an April 2024 request for an interpretation from the township’s zoning officer. Amazon’s interest in a data center deal, beginning in 2024, prompted a March 2025 draft ordinance that was never adopted.

    Hyperscale data centers are very large facilities built to serve cloud and large-scale computing customers. They need large tracts of land, substantial electricity and approval from local governments, which often lack rules written for this kind of use. Slate Belt Holdings, the Talen Energy and River Pointe Logistics partnership, has pursued both a legal track through zoning challenges and a negotiated track through a proposed $50 million community benefits agreement.

    Sources

    Source: ‘They’re on notice’: Upper Mount Bethel board gives supervisors 60 days to take action on data center regulation (The Morning Call): the zoning hearing board postpones Slate Belt Holdings’ validity challenge to Dec. 17 and gives supervisors 60 days to act on data center rules.

  • Bloom’s 2.8 GW Oracle Pact Is 1.2 GW Signed: Oracle Is Buying AI Power in Stages

    Bloom’s 2.8 GW Oracle Pact Is 1.2 GW Signed: Oracle Is Buying AI Power in Stages

    TL;DR · 30-second read

    The Short Version

    Oracle, a large business-software and cloud-computing company, is lining up its own power supply for its artificial intelligence data centers. It plans to use fuel cells from Bloom Energy: machines that make electricity on site through a chemical reaction instead of drawing it from the power grid.

    The plan covers up to 2.8 gigawatts. One gigawatt is roughly what a large power plant produces. But only 1.2 gigawatts is signed so far. The rest is an intention.

    Investors cheered anyway. Bloom’s share price jumped more than 14 percent in trading after the market closed, passing $200 for the first time.

    Stocktwits reported on April 13, 2026 that Bloom Energy (NYSE: BE) had expanded its partnership with Oracle (NYSE: ORCL) to supply on-site fuel cell power to Oracle’s AI and cloud data centers across the United States. Under the deal, Oracle intends to use up to 2.8 gigawatts (GW) of Bloom fuel cell systems. An initial 1.2 GW has been contracted, with deployment underway and continuing into 2027.

    Bloom shares rose more than 14% in after-hours trading, which is trading that happens after the regular market session ends, and crossed $200 for the first time. Oracle shares were up about 1% after hours. Bloom also confirmed that on April 9, 2026 it formally issued Oracle a warrant to buy more than 3.5 million Class A shares, under terms first agreed when the collaboration was announced in October 2025.

    Executive Summary

    Bloom Energy has turned an initial deployment with Oracle into a much larger framework. Oracle intends to source up to 2.8 GW of on-site fuel cell capacity for its AI data centers, and 1.2 GW of that is under contract today. Bloom presents the deal as evidence of a broader shift toward onsite, distributed energy, meaning power generated where it is used rather than delivered over the utility grid.

    The deal matters for two reasons. First, it is a hyperscale cloud provider, one of the handful of companies that operate data centers at enormous scale, choosing on-site generation as a primary way to energize AI capacity, at gigawatt scale. Second, the structure of the deal is revealing. A contracted tranche sits beneath a larger stated intention, and that tells buyers, suppliers and investors how a large AI operator is choosing to commit to power: in increments, as capacity is needed.

    The market read the headline number generously. Bloom’s stock moved more than 14% on the news, against roughly 1% for Oracle. That gap reflects how much more this agreement means to the supplier than to the customer.

    The Firm Number Is 1.2 GW, Not 2.8

    The 2.8 GW figure is a ceiling. According to the company, Oracle “intends to tap into up to” that amount. The contracted portion is 1.2 GW, a little over 40% of the headline total. The remaining 1.6 GW is described as intent, not obligation. For readers outside the power business, a gigawatt is a billion watts, which is roughly the output of a large power plant. A contract is what turns a gigawatt from an aspiration into revenue a supplier can plan around.

    That split is the mechanism behind staged buying. Bloom describes its systems as modular, built from repeatable units that can be added as needed. It says deployment of the contracted tranche is already underway and runs into 2027. A structure like this lets Oracle add generating capacity as new data halls come online instead of locking in the full 2.8 GW at the start. Oracle keeps flexibility on timing, sites and volume. Bloom gets firm visibility on 1.2 GW and a stated pathway to more.

    The people affected are concrete. Bloom’s manufacturing, supply chain and installation crews have to plan against the contracted tranche, while the stock appears to be pricing a good deal of the ceiling. Whether the other 1.6 GW goes to Bloom depends on future Oracle decisions that have not been announced. Grid utilities and other on-site power vendors serving the same sites therefore have something at stake in how those decisions land.

    What Oracle Is Really Buying Is Time

    Bloom’s central pitch is speed. The company says its modular systems deploy much faster than traditional power setups, which reduces risk and shortens project timelines. As evidence, it cites a fully operational system delivered to Oracle in 55 days last year, ahead of schedule. For a cloud provider, a finished data center without electricity is capital that earns nothing. Securing a new large grid connection can require substation and transmission upgrades that sit outside the operator’s control, so on-site generation offers a way to put the power supply on the developer’s own schedule.

    That argument has limits. Bloom has not said how large the 55-day system was. Repeating that pace across multiple U.S. sites at gigawatt scale is a different manufacturing, permitting and logistics task from a single delivery. On-site generation also swaps one dependency for another: grid interconnection risk becomes fuel-supply and equipment-maintenance risk, borne locally.

    A Customer With a Stake in the Supplier

    When the collaboration was announced in October 2025, Bloom agreed to grant Oracle a warrant to buy more than 3.5 million Class A shares, “to support broader adoption of its fuel cell technology.” A warrant is a right to buy stock at a set price in the future. Bloom formally issued it on April 9, 2026, four days before the expansion was announced, under the previously disclosed terms.

    Equity incentives for anchor customers are a recognized way for a supplier to win and keep a reference account. They also align incentives in a way buyers and investors should notice. If Bloom’s shares rise, Oracle’s warrant becomes more valuable, and an expanded Oracle commitment is one thing that can move those shares. The exercise price was not part of the April 13 announcement, so the value of Oracle’s position cannot be judged from what has been made public. Existing Bloom shareholders face some dilution if the warrant is exercised.

    Why Bloom Moved 14% and Oracle Moved 1%

    The stock reactions show where the deal is material. For Bloom, a 1.2 GW contracted tranche with a 2.8 GW ceiling is a defining customer relationship. For Oracle, power procurement is one input among many in a large AI buildout. Bloom shares have risen more than 886% over the past year, compared with about 15% for Oracle, so expectations for Bloom were already high before this announcement. Retail sentiment on Stocktwits moved from bearish to bullish over the day.

    A stock at those levels leaves little room for the uncontracted portion to slip. The useful test over the coming quarters is conversion: whether Bloom reports additional Oracle capacity moving from intent to contract, and whether the 55-day deployment pace holds as volumes grow.

    Background

    Bloom Energy makes stationary fuel cell systems that generate electricity at the customer’s site. It has positioned them as an alternative or complement to grid power for large, power-hungry facilities. Oracle, long known for database and enterprise software, operates Oracle Cloud Infrastructure and has been expanding its data center capacity to serve AI workloads. The two companies announced a collaboration in October 2025, which included Bloom’s agreement to grant Oracle a warrant for more than 3.5 million shares, and Bloom subsequently delivered a system that it says was fully operational in 55 days.

    AI data centers concentrate very large electricity demand in single locations, and connecting that demand to the grid can depend on utility upgrades outside a developer’s control. On-site generation such as fuel cells is one way operators try to bring capacity online on their own timetable, at the cost of taking on fuel, equipment and maintenance responsibilities locally.

    Sources

    Source: BE Stock Soars More Than 14% After-Hours: Expanded Oracle Deal Drives Boost, Stocktwits’ report on Bloom Energy’s expanded on-site fuel cell agreement with Oracle and the market reaction.

  • Bloom’s $139M LS Electric Deal Shows On-Site Data Center Power Can’t Skip the Gear Queue

    Bloom’s $139M LS Electric Deal Shows On-Site Data Center Power Can’t Skip the Gear Queue

    TL;DR · 30-second read

    The Short Version

    Bloom Energy, an American company that makes fuel cells, has placed a $139 million order with LS Electric, a South Korean maker of heavy electrical equipment. Fuel cells are machines that turn natural gas or hydrogen into electricity through a chemical reaction instead of burning it. The order is for a data center, a warehouse full of computers, in the United States.

    The surprise is that making your own power next to the computers does not remove the need for equipment that routes that power safely to them. How fast that equipment gets built now helps decide when the data center switches on.

    KED Global reported that LS Electric, the South Korean power-equipment manufacturer, has won a $139 million order from Bloom Energy, the US maker of solid oxide fuel cells, for a data center project in the United States. Neither the data center operator nor the site has been named publicly, and the companies have not itemized which equipment the order covers or how much generating capacity it supports.

    Executive Summary

    The order puts a dollar figure on something easy to overlook in discussions of on-site data center power. Bloom sells fuel cells that generate electricity at the customer’s premises, an approach often presented as a way to avoid waiting years for a utility grid connection. That electricity still has to be converted, protected and distributed before it reaches a server rack. LS Electric’s core business is the equipment that does this work: switchgear, which switches and protects electrical circuits; transformers, which change voltage levels; and power distribution systems.

    A $139 million commitment to one electrical supplier on a single data center project shows that on-site generation brings a substantial equipment supply chain with it. For data center developers, this means the speed of on-site power depends partly on electrical-equipment factories as well as on the fuel-cell maker. For investors, it means a meaningful share of the revenue from these projects goes to suppliers like LS Electric, and so does some of the risk of late delivery.

    On-Site Power Swaps One Queue for Another

    The main argument for generating electricity at a data center is speed. Connecting a large new facility to the utility grid usually means joining an interconnection queue, the waiting line utilities keep for projects that need new grid capacity, and data center developers have increasingly looked at on-site generation as a way around it. Bloom’s fuel cells are one of the best-known options. They can be installed near the load and produce electricity electrochemically from natural gas or hydrogen.

    A fuel cell produces direct current, though, and a data center’s electrical system expects conditioned, protected power at specific voltages. Getting from one to the other takes the same categories of equipment found in a utility substation or a conventional data center electrical room: switchgear to isolate faults and route power, transformers to step voltage up or down, and distribution gear to deliver it to the uninterruptible power supplies and server halls. LS Electric makes exactly these products. A $139 million order from Bloom to LS Electric for one US data center shows how much of this hardware an on-site power deployment can require.

    Those equipment categories are the same ones the industry has flagged as hard to get quickly, because utilities, renewable developers and data center builders all draw on the same factories. The developer that avoids the interconnection queue by going on-site still waits at the equipment maker’s production line. The utility’s timeline becomes less relevant, while the switchgear and transformer suppliers’ timelines become more relevant.

    Where the Money in On-Site Power Goes

    Coverage of fuel-cell deployments tends to focus on the fuel cells. This order shows the value of the overall project is shared more widely. Bloom is acting as a buyer and integrator here, purchasing electrical systems from a specialist instead of making everything itself. That is normal practice in power generation, but it changes how investors should read Bloom’s data center wins. Part of every project’s revenue passes through to suppliers, and Bloom’s margins on a project depend on what it pays for this equipment as well as on the economics of its fuel cells.

    For LS Electric, the order is a route into US data center demand through a power-generation partner rather than through direct contracts with data center operators. Equipment makers that sell to on-site generation providers, and not only to utilities and developers, gain a second channel into the same buildout. If on-site generation keeps growing as a share of new data center capacity, much of that growth will show up in the order books of electrical-equipment suppliers.

    Shared Revenue Means Shared Delivery Risk

    The supply-chain link also spreads exposure. If LS Electric’s equipment arrives late, Bloom’s installation slips, and the data center’s energization date (the day it can start drawing power) slips with it. The operator, which chose on-site power for speed, carries the consequences. When a project depends this heavily on a single supplier, that supplier’s factory capacity, shipping logistics and quality control become part of the project’s critical path.

    Because LS Electric is a Korean manufacturer serving a US project, international logistics and the treatment of imported electrical equipment under trade policy matter as well. None of this suggests the order is at risk. It describes the kind of exposure that comes with this model. Developers and investors evaluating on-site power should ask about the fuel-cell supplier’s capacity and also about the electrical-equipment suppliers behind it.

    Background

    Bloom Energy has spent years promoting its solid oxide fuel cells as on-site power for commercial and industrial customers. Data centers have become an increasingly important market as electricity demand, much of it driven by AI computing, has outpaced the ability of many utilities to connect new large loads quickly. On-site generation lets a developer bring power online near the facility, either as a bridge until a grid connection arrives or as a long-term primary source.

    LS Electric is a long-established South Korean maker of power transmission and distribution equipment. Its products include switchgear, transformers and power distribution systems that utilities and industrial customers use to move and control electricity. Demand for these product categories has risen worldwide alongside grid modernization, renewable energy buildouts and data center construction.

    Sources

    Source: LS Electric wins $139 million Bloom Energy order for US data center, KED Global’s report on LS Electric’s order from Bloom Energy for a US data center project.

  • Oracle’s 2.4 GW Bloom Reaffirmation Shows On-Site AI Power Trades on Delivery Risk

    Oracle’s 2.4 GW Bloom Reaffirmation Shows On-Site AI Power Trades on Delivery Risk

    TL;DR · 30-second read

    The Short Version

    Oracle, one of the giant companies that rents out computing power over the internet, has confirmed it still plans to buy a huge amount of power equipment from a company called Bloom Energy.

    The equipment is fuel cells: machines that turn natural gas into electricity on site without burning it. Together they would produce roughly as much power as two large nuclear reactors, to run artificial intelligence computer centers.

    A formal warning had said deliveries might be disrupted. When Oracle said the deal was still on, Bloom’s stock jumped about 8 percent. The deal didn’t get bigger. Investors were simply relieved it would happen.

    Bloom Energy shares climbed after Oracle reaffirmed its commitment to a 2.4 gigawatt (GW) fuel cell contract that had been put in question by a force majeure notice. A force majeure notice is a formal warning that events outside a party’s control may delay or prevent it from meeting its obligations. Yahoo Finance put Bloom’s gain at 8.7%, and 24/7 Wall St. reported a jump of about 8%.

    The contract covers on-site fuel cell generation intended to power AI data center capacity. Oracle’s reaffirmation left the headline size of the commitment unchanged at 2.4 GW.

    Executive Summary

    Oracle has publicly stood by a 2.4 GW fuel cell agreement with Bloom Energy after a force majeure notice raised the possibility that deliveries under the contract could be disrupted. Investors responded by bidding Bloom’s shares up roughly 8% to 8.7%.

    The notable part is what did not change. The contract was 2.4 GW before the notice and 2.4 GW after it. The share move therefore reflects restored confidence that the equipment will actually be delivered and used, not any new demand. That is a useful read on how the market values on-site power for AI. The megawatts on paper matter less than whether the delivery chain behind them holds.

    The episode also highlights concentration. A single customer’s statement about a single contract moved Bloom’s valuation this much, which suggests investors see this relationship as central to the company’s outlook. Bloom has not publicly quantified that dependence.

    A Reaffirmation, Not a New Order, Moved the Stock

    Bloom’s rally did not come from new business. The contracted capacity stayed at 2.4 GW, and no expanded scope was announced. What changed was the perceived probability that the contract would be performed as planned. A move of roughly 8% to 8.7% on that alone is effectively a market price on delivery risk. Before the reaffirmation, investors were discounting the chance that the notice would lead to delays, renegotiation or a smaller commitment. Oracle’s statement removed much of that discount.

    The mechanism runs through the force majeure clause. That clause lets a party suspend or delay its obligations when an event beyond its control intervenes. For on-site generation, performance depends on a long chain: fuel cell modules have to be manufactured, shipped, installed, connected to a gas supply, permitted and commissioned. A notice under that clause signals that at least one link is under strain. For the buyer, the stakes are high, because on-site fuel cells are the power plant for the data center. If the modules arrive late, the computing capacity switches on late.

    Three groups feel that directly. Bloom’s shareholders are exposed because revenue timing depends on delivery. Oracle’s capacity plans assume the power arrives on schedule. Contractors and lenders have built plans around the same timeline. For on-site AI power, the variable that moves value is delivery certainty, and this episode puts a number on how much.

    Why AI Builders Are Taking Power Into Their Own Hands

    Large AI data centers need enormous, steady amounts of electricity. Connecting that much new demand to the utility grid often means joining an interconnection queue, the line of projects waiting for grid operators to study and approve new connections, and that wait can stretch for years. On-site generation offers a way around it. Fuel cells can be installed next to the data center and added in modules as capacity grows.

    That speed comes with a trade-off. An operator that generates its own power exchanges dependence on a regulated utility for dependence on a vendor contract. That means exposure to the manufacturer’s factory output, to natural gas pipeline availability and to local air permits. Fuel cells avoid combustion, but they still emit carbon dioxide when running on natural gas, so permitting and emissions questions do not disappear.

    Scale sharpens the risk. A gigawatt is 1,000 megawatts, and 2.4 GW is roughly the output of two to three large nuclear reactors. At that size, on-site generation is no longer backup power. It is the primary supply for the campus, so any disruption to it matters.

    One Buyer, Outsized Weight

    The size of the market’s reaction suggests investors treat the Oracle contract as central to Bloom’s outlook. Bloom has not said what share of its backlog or expected revenue the contract represents. Concentration of this kind cuts both ways. An anchor customer from the hyperscale tier (the handful of companies running cloud platforms at global scale) validates the technology at a scale few buyers could offer, and it can justify factory expansion. The same concentration means one counterparty’s decisions on site priorities, schedules or procurement flow directly into Bloom’s revenue timing.

    In fairness to both companies, Oracle’s reaffirmation is a meaningful signal. A buyer looking to exit or reprice a deal could have used a force majeure notice as leverage, and Oracle chose to stand by the commitment publicly. Still, a public reaffirmation is not the same as disclosed terms. It is not known whether delivery schedules, pricing or remedies were adjusted along the way.

    What Buyers and Financiers Should Take From It

    For data center developers, the practical lesson is in the contract drafting. That includes how force majeure is defined, what remedies apply if delivery slips, whether deliveries are staged against milestones, and what backup supply exists if a vendor cannot perform. These terms carry weight well beyond boilerplate when on-site generation is the campus’s main power source.

    Lenders and equity investors underwriting AI campuses powered on site will likely want more visibility into the vendor’s manufacturing throughput and the fuel supply behind each project. This episode shows the market is already pricing that risk. Better disclosure is how the discount narrows.

    Background

    Bloom Energy builds solid oxide fuel cells, modular systems that convert natural gas or other fuels into electricity through an electrochemical reaction rather than combustion. The company has positioned its systems as a fast way for data center operators to secure large blocks of power without waiting for grid upgrades, a pitch that has gained traction as AI workloads push electricity demand well beyond what many local grids can supply quickly.

    Oracle runs Oracle Cloud Infrastructure and has become one of the most aggressive builders of AI computing capacity. Its proxy statement filed with the SEC on September 25 lists Clayton M. Magouyrk and Michael D. Sicilia, alongside Safra A. Catz, as principal executive officers for the fiscal year ended May 31, 2026. That is the leadership team now responsible for Oracle’s large power commitments, including the 2.4 GW Bloom contract.

    Sources

    Source: Bloom Energy (BE) Is Up 8.7% After Oracle Reaffirms 2.4 GW AI Fuel Cell Deal Commitment – Has The Bull Case Changed? (Yahoo Finance) and Bloom Energy Jumps 8% as Oracle Reaffirms 2.4 GW Fuel Cell Contract After Force Majeure Notice (24/7 Wall St.): coverage of Bloom Energy’s share gain after Oracle reaffirmed its 2.4 GW fuel cell contract.

    Primary sources: Oracle Corporation, Schedule 14A definitive proxy statement (DEF 14A), filed with the SEC on September 25, 2026.

  • Upper Merion’s 4.6M sq ft Denial Shows Why Self-Powered AI Campuses Still Need Town Hall

    Upper Merion’s 4.6M sq ft Denial Shows Why Self-Powered AI Campuses Still Need Town Hall

    TL;DR · 30-second read

    The Short Version

    A suburb outside Philadelphia voted down five giant buildings meant to hold the computers that run artificial intelligence. Together they would have covered roughly 80 football fields of floor space.

    The builder promised $55.4 million a year in local taxes, mostly for schools. Neighbors worried about noise, dirty air and home values, and more than 17,000 people signed a petition against it.

    The twist: to avoid raising local electric bills, the site would have made its own power with gas-burning generators. That fix became part of the fight. The builder says an appeal is coming.

    WHYY reported that Upper Merion Township’s Board of Supervisors on Thursday, August 13, 2026, rejected all five applications from developer Brian O’Neill for a hyperscale data center campus totaling more than 4.6 million square feet near King of Prussia, Pennsylvania. The campus would have been built at Discovery Labs, a health care and life sciences office property run by O’Neill’s company, and was designed to generate its own power using natural gas turbines with diesel backup generators.

    The vote followed a July planning commission recommendation against three of the five projects, a developer lawsuit seeking more review time that a Montgomery County judge cleared on the afternoon of the vote, and a resident petition with more than 17,000 signatures. O’Neill told 6ABC an appeal is planned.

    Executive Summary

    A suburban Philadelphia township has turned down one of the largest data center proposals in the region, despite the developer’s estimate of $55.4 million a year in local tax revenue, nearly $38 million of it for the local school district. The board’s stated grounds were procedural and specific: Chair Tina Garzillo said the township never received landscape, traffic impact, emergency management or sanitary sewer plans, and called the projects “at best speculative and conceptual in nature.”

    The case matters beyond Upper Merion because of how the campus was to be powered. Generating electricity on site is increasingly pitched as the responsible way to build AI-scale facilities without burdening the grid or ratepayers. Here, that design put gas turbines and diesel generators in front of a township board and its residents, and air pollution became a central objection. Self-generation did not remove the local approval gate; it added a power plant to what that gate had to judge.

    For developers, investors and utilities trying to forecast where AI capacity will actually get built, the vote is a reminder that a proposed campus is not capacity until it clears local land-use approval, and that the power strategy is now part of that approval.

    Self-Generation Swapped a Grid Question for a Local One

    The campus was pitched as a good grid neighbor: it would generate its own electricity so it would not strain the regional power system or push up household bills. That is an increasingly common design for AI-scale facilities, often called on-site or behind-the-meter power because the electricity is produced on the property rather than drawn from the utility. But in Upper Merion, the power plant came with the building. The proposal included natural gas turbines and diesel backup generators, and air pollution became one of residents’ leading objections, alongside noise and water use.

    That is the mechanism this vote exposes. A data center that buys power from the grid leaves most generation questions to utilities and grid planners. A data center that makes its own power brings combustion equipment, fuel supply and emissions controls onto a site that a township board must approve, in front of neighbors who will live beside it. O’Neill said scrubbers would control emissions and that discussions had been held with Bloom Energy about using fuel cells on site, but no deal had been finalized, which left the board and residents to evaluate the turbine-and-diesel version of the plan.

    Upper Merion is not the only place where on-site power has drawn local enforcement. In Vineland, New Jersey, where a data center under construction plans to use Bloom fuel cells, the city issued a stop-work order on parts of the 300-megawatt facility (a megawatt is a measure of electrical capacity), alleging that a liquefied natural gas tank and fuel cell units were being installed before city approval. Generating power on site may answer concerns about grid strain and electricity bills, but it adds a power plant to the list of things local officials get to review.

    The Board Ruled on Missing Plans, Not Just Public Anger

    The emotional scenes at the meeting can obscure what the board actually said. Garzillo cited the absence of landscape, traffic impact, emergency management and sanitary sewer plans. Those are ordinary engineering submissions for any large industrial project, and their absence gave the board a procedural basis for denial that stands apart from residents’ objections.

    The process had been heading this way for weeks. In July, the township planning commission voted not to recommend three of the five projects. The developer then sued, seeking an extension and arguing there had not been adequate time to address township concerns; a Montgomery County judge vacated the injunction on the afternoon of the vote, clearing the board to reject the extension. Vice chair William Jenaway said the developer had earlier opportunities to respond to review comments and did not do so in a timely fashion. The two sides’ accounts of the timeline differ, and an appeal would likely test them.

    For the industry, the lesson is about entitlement risk, meaning the risk that a site never receives the land-use approvals it needs. As AI demand pushes developers to repurpose existing properties, here a life sciences office campus, into data center sites, applications that arrive without complete plans risk being denied on the paperwork alone, whatever their scale or revenue promise.

    A $55.4 Million Tax Pitch Met a 17,000-Signature Petition

    The developer’s economic case was large on its face: an estimated $55.4 million a year in local tax revenue, with nearly $38 million going to the Upper Merion Area School District. For a suburban township, that is the kind of figure that usually carries weight. It did not carry the vote. Residents weighed it against quality-of-life costs that do not appear in a revenue projection, and one resident nearing retirement said a loss of home equity “would destroy our savings.”

    Both sides’ claims deserve the same scrutiny. The tax figure is the developer’s own estimate, and its value depends on assumptions about how much gets built and when. Residents’ fears about property values and health are sincere, but their magnitude depends on specifics that were never settled, such as how many turbines would run, how often, and with what controls. The water debate is similar: O’Neill said a closed-loop system would reuse cooling water, while hyperscale facilities in general can consume up to 5 million gallons of drinking water a day. Closed-loop designs reduce consumption substantially, but how much make-up water this campus would need was not specified.

    The backdrop is shifting against developers who lead with revenue. A University of Pennsylvania Annenberg Public Policy Center survey found a 12-point jump in widespread resistance to new data center construction over four months. In that climate, detailed answers on emissions, water and noise are likely to matter more to approval than headline tax estimates.

    What Comes Next for AI Capacity Around Philadelphia

    The denial is not necessarily final. The developer can appeal or submit new or revised plans, and has said an appeal is planned. Residents have said they will oppose any appeal or new application. O’Neill is also seeking approval for a data center in nearby Plymouth Township, which becomes the next test of whether a more complete application, and possibly a different power plan such as the fuel cells under discussion, changes the outcome.

    For operators, investors and utilities counting future AI capacity in the region, the practical point is that local land-use approval is a gate that neither scale, tax revenue nor a plan to stay off the grid can bypass. Capacity forecasts that treat proposed campuses as inevitable will miss decisions like this one, which are made parcel by parcel at township meetings.

    Background

    Upper Merion Township sits in Montgomery County, Pennsylvania, and includes much of the King of Prussia area, a major suburban commercial and office hub northwest of Philadelphia. Data centers, the buildings that house the servers behind internet services, have existed since the mid-20th century, but the rise of power-hungry artificial intelligence has pushed demand toward much larger hyperscale campuses that need substantial electricity and cooling.

    That demand has driven developers to look at existing commercial properties, such as office and life sciences campuses, as potential data center sites, and to consider generating their own power rather than waiting on the grid. It has also produced growing local resistance in the Philadelphia region and beyond, where township boards, not state or federal agencies, hold the zoning and land-use approvals that decide whether a campus can be built.

    Sources

    Source: Upper Merion board denies hyperscale data center campus near King of Prussia, WHYY’s report on the township supervisors’ vote rejecting a five-building data center campus.

  • Aalo’s 50 MW Pod Shows Small Nuclear’s AI Benchmark Is On-Site Gas

    Aalo’s 50 MW Pod Shows Small Nuclear’s AI Benchmark Is On-Site Gas

    TL;DR · 30-second read

    The Short Version

    Crusoe, a company that rents out computing power for artificial intelligence, will spend six months testing a small nuclear reactor from startup Aalo alongside one of its compact data centers.

    Aalo’s reactor is sized to match the gas-fired generators many of these sites already run on. The test is about whether it can keep the computers running reliably, even as their demand for electricity swings up and down.

    If it works, the two companies aim to start building for real in 2029, but first Aalo needs a green light from federal nuclear safety regulators.

    Crusoe, a cloud provider built specifically for AI workloads (often called a neocloud), and reactor developer Aalo Atomics will pair Aalo’s small modular reactor (SMR) technology with Crusoe’s modular Spark AI data center for a six-month trial, Fierce Network reported on August 11, 2026. The test will examine the reactor’s capacity factor, meaning how consistently it can run without downtime, along with its ability to follow the fluctuating power draw of AI workloads when combined with battery storage.

    Both companies are targeting commercial nuclear-powered AI data center deployments in 2029. Aalo’s first product is the Aalo One Reactor in a 50-megawatt Aalo Pod configuration. Commercial deployment in the United States depends on approval from the Nuclear Regulatory Commission.

    Executive Summary

    The Crusoe–Aalo trial is a small-scale engineering test, not a power purchase deal. The goal is to find out whether a factory-built reactor and a factory-built data center can be tightly coupled on one site. According to Aalo CEO Matt Loszak, this will be the first time the two types of infrastructure are so closely integrated. Crusoe co-founder Cully Cavness described the work as mostly electrical engineering: controls, monitoring, load profiles and fault conditions running in both directions.

    The strategic significance lies in how the reactor is sized. The large nuclear deals signed by the major hyperscalers mostly secure output from existing large plants. Aalo, by contrast, has built its 50-megawatt pod to fit the 10-to-50-megawatt range where on-site gas turbines currently power many data centers, and says it can make nuclear’s economics very competitive with natural gas. If that holds, small reactors would be competing for the on-site generator slot rather than against utility-scale nuclear. That claim is still a target and has not been demonstrated.

    The Benchmark Is a Gas Turbine, Not a Gigawatt Plant

    Hyperscaler nuclear deals have mostly involved large legacy plants that deliver power over the grid. Aalo is aiming at a different market. Loszak said many data centers are being powered by gas turbines and small modular gas generators in the 10-to-50-megawatt range, and that this is the reason Aalo chose a 50-megawatt pod. He described the design as a balance between economy of scale, which argues for larger units, and economy of number, which argues for mass-producing many identical ones. He added that a reactor as small as 1 megawatt would suffer on economics.

    This sizing decision identifies the real competitor. A 50-megawatt pod that can be stacked is meant to fill the same role in a site plan that on-site gas fills today: dedicated, behind-the-meter capacity, sized in blocks, added as a campus grows. It is not meant to replace a grid connection to a 1-gigawatt plant. The people affected are data center developers and neoclouds that currently use gas to get around grid interconnection delays. Crusoe is one of them, and Cavness said the company already pursues natural gas, solar and batteries as a portfolio.

    The comparison also sets the standard the technology must meet. On-site gas is a known quantity with established supply chains. Aalo’s statement that it can be very competitive with natural gas is a stated goal, and neither company has published cost-per-megawatt-hour figures to back it up. The benchmark is clear. Whether Aalo can hit it has not yet been shown.

    Why the Trial Is About Load Following, Not Reactor Physics

    Loszak said the first priority is capacity factor: can the reactor run for months with very limited downtime? The second priority is load following, meaning how well the power supply tracks demand as it changes. AI workloads are not flat. Loszak said Crusoe may simulate the different fluctuation patterns it sees across training and inference jobs. The setup will include battery storage so the companies can test how nuclear and batteries share the job of absorbing those swings.

    This is the practical challenge of tight coupling. When a data center draws from the grid, the grid absorbs its variability. When a reactor sits beside the load with no grid in between, any mismatch between supply and demand has to be handled on site, through controls, storage, or the reactor’s own response. Cavness’s emphasis on fault conditions in both directions points to the same problem: a disturbance on either the data center side or the reactor side has to be contained without taking down the other.

    This is also where the gas comparison is tested in practice. On-site generators are often chosen because operators can dispatch them to match load. A nuclear pod paired with batteries needs to show similar operational behavior before a buyer will treat it as a replacement. Six months of data on that pairing is exactly what a prospective buyer would want to see.

    Time Is the Cost Line Both Companies Are Trying to Cut

    Cavness described nuclear’s historic problem as time and money. Build timelines of a decade or more lead to cost overruns, which raise the price per kilowatt-hour. The proposed solution is modular, mass-manufactured units that bring down both the timeline and the cost. Aalo reached criticality, a sustained nuclear chain reaction, within three years of its founding, which is fast by the standards of the nuclear industry. Crusoe’s modular data center approach follows the same manufacturing-led thinking.

    The 2029 target is aggressive, and several dependencies stand in the way. Aalo needs Nuclear Regulatory Commission approval before any domestic commercial deployment. The companies also acknowledge they will need to win over local communities, which are increasingly skeptical of data centers even without a reactor involved. Their stated answer is careful site selection and public education. A successful trial would reduce engineering risk, but it would not settle licensing, siting or cost, and those are the factors that will determine whether small nuclear actually takes a share of the on-site power market from gas.

    Background

    Crusoe is a neocloud, a cloud provider focused on AI computing, that builds its own data centers, including its modular Spark design. It has pursued several energy sources, including natural gas, solar and batteries, to power those sites as grid capacity for data centers has become harder to secure. Aalo Atomics is a nuclear developer building factory-made reactor pods. Its first product is the Aalo One Reactor in a 50-megawatt Aalo Pod configuration, and the company reached criticality within three years of its founding.

    Nuclear power has become a major topic in data center planning. All three major hyperscalers have signed nuclear supply agreements, mostly with large existing plants. Small modular reactors are a newer, largely unproven category. Proponents argue they can avoid the decade-plus build times and cost overruns of traditional nuclear, but they have yet to demonstrate that at commercial scale.

    Sources

    Source: Neocloud Crusoe puts Aalo nuclear power to the AI data center test (Fierce Network), covering Crusoe and Aalo’s six-month trial pairing a small modular reactor with Crusoe’s Spark AI data center.

  • Vantage’s 95-Ton NOx Permit Shows AI Self-Power Is Judged by Ceiling, Not Output

    Vantage’s 95-Ton NOx Permit Shows AI Self-Power Is Judged by Ceiling, Not Output

    TL;DR · 30-second read

    The Short Version

    A data center in Virginia makes its own electricity with eight gas-burning turbines, each about the size of a bus, close to people’s homes.

    A Harvard researcher estimated its air pollution could cause tens of millions of dollars a year in health harm. Virginia’s environmental regulator moved quickly to challenge that estimate, even though one of its own experts said judging health effects was outside his training.

    Why it matters: more of the buildings that run artificial intelligence want their own power plants, and fights like this one will decide where they can be built.

    Politico reported on July 25 that internal emails, obtained through a public records request, show Virginia’s Department of Environmental Quality (DEQ) organized a rebuttal to a health analysis of Vantage Data Centers’ self-powered facility in Sterling, Loudoun County, before that analysis was published. DEQ Director Michael Rolband forwarded the Piedmont Environmental Council’s report to seven air-permitting staff within five minutes of receiving it in early March. He asked for “a defensible critique that will stand the test of public and critical review.”

    The Piedmont-commissioned study estimated that soot from the facility’s eight natural gas turbines could cause $53 million to $99 million a year in health costs, and 3.4 to 6.5 premature deaths, if the site emitted the maximum its permit allows. DEQ’s own report, released in late April, said actual emissions were 28 to 69 percent of permitted levels. The agency has said its staff “did not collaborate with PEC, Vantage or any other external participants” in its assessment.

    Executive Summary

    This dispute centers on the only data center in Virginia that generates its own electricity. Its on-site power system combines eight bus-sized natural gas turbines with dozens of backup diesel generators, and it sits in a residential area. DEQ’s 2023 permits allow up to 95 tons a year of nitrogen oxides (gases that help form smog) and more than 56 tons of soot, meaning fine particles that lodge in the lungs. Both pollutants are linked to asthma and heart attacks. Those limits are typical for a minor pollution source. The dispute arose because of where the facility sits, not because of how large the limits are.

    It matters beyond one building because self-generation is becoming the industry’s answer to grid congestion. Dominion Energy disclosed in February that it receives about 10 new connection requests a month from large-load customers. It also disclosed that dozens of projects requesting a combined 70 gigawatts are waiting to connect, three times the utility’s all-time peak load. Vantage itself described its turbines to DEQ as “a temporary solution given the current power constraints of the grid.”

    The episode shows how the health case against on-site generation gets built. Critics model the legal ceiling in the permit, not the meter reading. Even when actual emissions come in well below that ceiling, the permit keeps the higher figure in play.

    The Permit Ceiling Is the Number That Gets Modeled

    Michael Cork, a biostatistics research fellow at Harvard’s T.H. Chan School of Public Health, did not have Vantage’s operating data when Piedmont hired him. He used the only public figures available, which were the permitted maximums. That produced the headline range of $53 million to $99 million in annual health costs. Vantage objected that this was a “worst case scenario,” and DEQ’s report made the same point: measured emissions of soot, volatile organic compounds and nitrogen oxides ran at 28 to 69 percent of what the permit allows.

    The correction lowered the estimate but did not end the argument. Cork reran his model with DEQ’s actual-emissions figures and arrived at $33 million to $63 million a year and two to four premature deaths. He also noted that nothing in the permit prevents the facility from increasing output back toward its ceiling. That is the mechanism the headline refers to. A permit is a legal right to emit, so outside analysts can reasonably treat it as the facility’s potential footprint, whatever the current meter shows.

    This matters for developers planning on-site generation. Operators often seek permit headroom so they can run harder when the grid is strained or when load grows. That headroom has operational value. It is also, in effect, a published number that critics can model. A project that wants its real-world performance to count in public debate may need to make that performance binding, through enforceable limits closer to expected output or regular public reporting of actual emissions, rather than relying on a regulator to vouch for it after the fact.

    Why Self-Power Keeps Spreading Anyway

    The economics pushing data centers toward their own power plants are not easing. A 70-gigawatt connection backlog at a single utility means many projects face long waits for grid power. On-site gas turbines offer a way to start operating sooner. They also answer the political charge that data centers raise household electricity bills. The federal government has encouraged the model too: President Donald Trump promoted a wave of power plants for new data centers at a White House meeting with technology companies in early March.

    The Vantage case shows the trade-off. Self-generation moves the power question off the utility’s books and onto the local air-permit docket. The cost stops appearing in rate cases and starts appearing in health-damage estimates for the neighborhood downwind. The more facilities follow this path, the more often permitting agencies will be asked to rule on disputes that go beyond their traditional remit.

    A Regulator’s Rebuttal Is Not a Health Finding

    The emails show disagreement inside DEQ about how far the agency should go. Michael Kiss, a manager in the Office of Air Quality Assessments, wrote that the magnitude of impacts was “relatively low” and not surprising. He also said he had “absolutely no comments on the health study” because “we aren’t qualified to do that review.” Robert Burnley, who led DEQ from 2002 to 2006, said the conventional defense of a permit is narrower: that the emissions do not cause a violation of air quality standards.

    Several claims in DEQ’s final report are open to challenge on their own terms. The report said health impacts were “unlikely” because Loudoun County soot levels are below EPA standards. Health experts, including former EPA air quality official Michael Koerber, said soot can harm health even at EPA-approved levels, and the World Health Organization’s guideline is nearly half the EPA figure. The report also cited 2023 state health data, collected before Vantage began operating, to describe Loudoun as one of Virginia’s healthiest localities. That data says little about the facility’s effect.

    For the industry, the lesson is practical rather than partisan. The Data Center Coalition circulated DEQ’s report to reporters and lawmakers as evidence that the Piedmont analysis was wrong. A rebuttal whose health reasoning is contested gives less protection than it appears to. The most useful thing DEQ published was arguably the actual-emissions data. It narrowed the dispute to a real number, even though that number still produced a substantial estimate.

    Siting Near Homes Is Where the Costs Land

    DEQ’s permits for Vantage are in line with other minor-source permits in the state. The concern stems from the facility’s location in a residential area. Neighbors report constant noise as well as air pollution. Vantage says it has taken steps to reduce the noise. Health-cost models of the kind Cork used scale with the number of people in a plume’s path, so the same turbines sited away from housing would generate a very different estimate.

    Virginia has more than 600 data centers, and many future self-powered sites will compete for land in areas that are already developed. Where generation sits relative to homes is likely to be as consequential for permitting risk as how many tons it is allowed to emit.

    Background

    Northern Virginia, and Loudoun County in particular, hosts the world’s largest concentration of data centers, and Virginia is home to more than 600 of them. Dominion Energy, the region’s largest utility, has struggled to keep pace with demand. Large projects now wait in a queue that is far larger than anything the utility has ever served. In response, developers have begun planning facilities that generate their own electricity on site, usually with natural gas turbines.

    Vantage Data Centers’ Sterling facility came before that trend. DEQ issued its air permits in 2023. The Piedmont Environmental Council, a Virginia conservation group, has been active on data center land-use and environmental impacts. The Data Center Coalition is the industry’s trade group.

    Sources

    Source: Emails show how Virginia regulators downplayed data center health concerns — Politico’s report on internal DEQ emails concerning Vantage’s self-powered Loudoun County data center.