Bloom Energy and global investment firm Brookfield expanded their strategic partnership fivefold on June 30, 2026 — from $5 billion to $25 billion — in a deal that signals something larger than a single corporate announcement: the AI industry is not waiting for the electricity grid anymore. Hyperscalers and data center developers are increasingly bypassing utility interconnection queues by generating power on-site with solid oxide fuel cells, a shift that has turned Bloom Energy into one of the fastest-growing companies across any sector in 2026.

The $25 billion commitment is part of Brookfield's dedicated AI Infrastructure Fund, launched in November 2025 with a target of deploying $100 billion across AI factories, power solutions, and compute infrastructure. Brookfield's Head of AI Infrastructure, Sikander Rashid, described the expanded collaboration as strengthening the firm's ability to provide end-to-end AI infrastructure solutions — in his words, "from electrons to tokens, for some of the world's most sophisticated customers."

Bloom Energy shares closed up 10% on June 30 on the news, at $302.70. The stock has surged more than 1,200% over the past 12 months as institutional capital bets that the AI buildout's defining bottleneck is not chips, compute, or software — it is power.

AI's Power Problem Is Not a Chip Problem

The urgency behind the deal comes from a mismatch between the pace of AI ambition and the pace of grid construction. A single modern AI training campus can require more than a gigawatt of power. New utility connections, however, now require navigating interconnection queues that stretch five to seven years in many U.S. markets. The U.S. Department of Energy projects that U.S. data center electricity consumption will reach between 325 and 580 terawatt-hours by 2028, up from 176 terawatt-hours in 2023 — growth the existing grid was never designed to accommodate.

The result is that roughly one-third of all planned new data center power capacity in the United States is now designed to skip the grid entirely, generating electricity on-site, according to research cited by Axios. Data center operators are making a deliberate choice to route around the regulated electricity system rather than queue for connections that may not arrive for years.

Bloom's fuel cell systems have become the primary vehicle for that routing. The company closed approximately $7.65 billion in binding data center contracts in a roughly 90-day period in early 2026 — more than the entire fuel cell industry's cumulative data center revenue from the prior decade.

How Bloom Energy's Solid Oxide Fuel Cells Work

Understanding why Bloom Energy has won this market requires understanding what a solid oxide fuel cell (SOFC) actually is — and why it differs fundamentally from the gas turbines and diesel generators it is replacing.

An SOFC is an electrochemical conversion device. It produces electricity by oxidizing natural gas directly through a solid ceramic (oxide) electrolyte, conducting negative oxygen ions from cathode to anode at temperatures between 600 and 1,000 degrees Celsius. No combustion takes place. That distinction matters for two engineering reasons.

First, because combustion is absent, the Carnot cycle efficiency limit that constrains conventional heat engines does not apply in the same way. Bloom's solid oxide fuel cells achieve 54% fuel efficiency at the point of use — 15 to 20 percentage points higher than most open-cycle gas turbines. For a median-sized 175 megawatt data center, that efficiency advantage translates to estimated fuel cost savings of $70 million to $100 million over five years, based on assumptions about gas prices. Combined with a heat recovery system, total efficiency climbs from 54% to more than 90%.

Second, fuel cells have no rotating parts. That design choice makes them uniquely suited to AI workloads, which are notoriously volatile: testing has shown that AI power loads can swing from 20% to more than 150% of provisioned power in milliseconds, with dozens of such swings occurring every minute. Conventional combustion generators, constrained by mechanical inertia, cannot respond at that speed. Bloom's systems respond at least twice as fast as rotating generators when stepping up, and instantly when stepping down. Combined with supercapacitors, they reach full response within milliseconds.

Bloom's systems are also modular. Individual 325-kilowatt building blocks stack into larger configurations, scaling from single megawatts to hundreds — and Bloom's stacked design delivers up to 100 megawatts per acre, roughly double the density of natural gas turbines. That footprint advantage matters in land-constrained suburban markets where most data centers are built.

The most consequential engineering characteristic for AI infrastructure is deployment speed. Bloom's Energy Servers arrive prefabricated and can be commissioned in months rather than years. Oracle selected Bloom Energy to provide onsite power at one of its AI data center campuses and received its first system in just 55 days — 35 days ahead of the 90-day target — a result that convinced Oracle to expand its agreement to up to 2.8 gigawatts of Bloom's systems across its operations.

A $25 Billion Bet, and What It Signals for the Grid

The fivefold expansion of Brookfield's commitment, announced June 30, builds on the October 2025 initial $5 billion framework, which was one of Brookfield's first seed investments in its AI Infrastructure Fund. The expanded deal covers global deployment of Bloom's fuel cell systems at AI data centers and hyperscale AI factories, combining Brookfield's capital and infrastructure development scale with Bloom's deployable power platform.

Aman Joshi, Bloom's Chief Commercial Officer, said the enlarged commitment reflects market momentum following several recently announced large-scale deals. "When we formed this partnership, we said it was the first phase of a much larger vision," Joshi said. "Bloom is uniquely positioned to address the urgent need for clean, reliable power to support the rapid growth of AI."

Brookfield, which manages more than $1 trillion in assets under management, has estimated that total spending on AI-related infrastructure will exceed $1 trillion this decade.

The structural implication that neither company's announcement addressed directly is what this buildout means for the electricity grid itself. At gigawatt scale, AI data centers generating their own power on-site are not supplementing the regulated grid — they are strategically bypassing it. That shift hollows out the load growth that utilities depend on to justify planned transmission and generation investment, raises questions about stranded-asset risk for utility capital expenditure, and presents regulators with a novel governance problem: large-scale natural gas generation operating behind the meter, outside traditional utility rate structures and oversight. Investors will need to monitor how utility planning assumptions and regulatory frameworks adapt to a model in which the grid's most power-hungry customers are no longer its customers at all.

Risks Investors and Data Center Operators Should Weigh

The deal's headline number should not obscure the execution challenges ahead. Bloom Energy's stock, while up more than 1,200% over the past year, is trading at more than 130 times forward earnings as of early July. BMO Capital maintains a Hold rating with a price target of $279, citing early-stage execution risk. The BMO analysis notes that a financing commitment and a commissioned project are different things: the $25 billion framework is a ceiling on available funding, not a guarantee of revenue.

Competitive pressure is also intensifying. Chevron and Microsoft have deployed natural gas turbines as alternative on-site generation solutions. The U.S. Department of Energy is financing next-generation nuclear reactors targeting the same data center power market, and FuelCell Energy and Plug Power are competing in the fuel cell segment. Any of these alternatives could limit Bloom's pricing power on new contracts or compress margins as the market matures.

There is also a carbon caveat that the "clean power" framing can obscure. Bloom's fuel cells still emit carbon dioxide when running on natural gas — they are meaningfully cleaner than combustion alternatives, producing virtually no nitrogen oxides, sulfur oxides, or particulate matter, but they are not zero-carbon. Bloom has flagged a pathway to lower-emission operation using biogas or hydrogen, but both fuels remain limited and expensive at the scale data centers require.

What a $25 Billion Bet Signals for the AI Infrastructure Race

For the broader AI industry, the deal's significance may reach beyond its headline number. It represents one of the clearest institutional signals that the defining constraint in the race to build AI infrastructure is not accelerators or interconnects — it is electrons. Whoever can deliver reliable power fast enough to match the pace of AI ambition will shape which data center projects get built and which stay in queue.

Bloom Energy's 2026 financial guidance — raised to $3.4 billion to $3.8 billion in revenue, implying roughly 80% year-over-year growth at the midpoint — suggests the market has already decided that fuel cells are the near-term answer. Brookfield's $25 billion commitment suggests institutional capital agrees. The more open question is whether the regulatory and grid planning apparatus that governs U.S. electricity infrastructure is prepared for the pace and scale of what is now happening entirely outside its oversight.

Frequently Asked Questions

How do Bloom Energy's solid oxide fuel cells generate electricity without combustion?

Bloom Energy's solid oxide fuel cells convert natural gas to electricity through an electrochemical process rather than burning fuel. At operating temperatures between 600 and 1,000 degrees Celsius, a ceramic (solid oxide) electrolyte conducts oxygen ions from a cathode to an anode, where they oxidize the natural gas. This direct chemical-to-electrical conversion bypasses the mechanical turbine step entirely, which is why fuel cells achieve 54% electrical efficiency at point of use — well above the 35-to-40% range typical of open-cycle gas turbines — while producing virtually no nitrogen oxides, sulfur oxides, or particulate pollution.

Why is AI creating such a severe electricity shortage for data centers?

AI training and inference workloads require continuous, high-density power at a scale that is unprecedented for the size of their physical footprint. A single large AI training campus can require a gigawatt or more of power. U.S. grid interconnection queues to accommodate demand of that scale now stretch five to seven years in many markets, according to utility and industry data. The Lawrence Berkeley National Laboratory's 2024 report, commissioned by the U.S. Department of Energy, projects U.S. data center electricity demand will reach between 325 and 580 terawatt-hours by 2028, up from 176 terawatt-hours in 2023 — growth that no combination of near-term grid upgrades can fully accommodate. Data center operators are choosing on-site generation specifically to escape that bottleneck.

What happens to the electricity grid if AI data centers generate their own power at scale?

At the scale that the Brookfield-Bloom partnership is targeting — potentially gigawatts of behind-the-meter generation globally — AI data centers are not merely supplementing the grid; they are opting out of it. That structural shift carries consequences that go beyond AI infrastructure. Utilities rely on large industrial loads to justify transmission and generation investment. When those loads disappear behind the meter, utilities face stranded-asset risk on infrastructure they built in anticipation of demand that never materialized. Regulators also face a novel challenge: large-scale natural gas generation operating outside traditional utility rate structures, oversight, and emissions permitting requirements. How U.S. energy policy adapts to behind-the-meter generation at AI scale is an open question that neither the companies nor regulators have fully answered.

Is Bloom Energy's fuel cell technology carbon-free?

No. Bloom's fuel cells are substantially cleaner than combustion-based alternatives — they produce virtually no nitrogen oxides, sulfur oxides, or particulate matter — but they still emit carbon dioxide when running on natural gas. Bloom has described pathways to lower-emission operation using biogas or hydrogen, but those fuels remain expensive and limited in supply at the scale data centers require. Carbon capture integration is another option the company has flagged; Bloom's mid-year 2026 data center report notes that roughly one-third of on-site-powered data center sites plan to incorporate carbon capture by 2030. For now, fuel cells represent a bridge technology: faster and cleaner than the combustion alternatives currently in use, but not a zero-carbon solution.