Fuel cells are gaining traction among US hyperscalers racing to build data centers for the AI boom but struggling with grid constraints to secure quick and reliable power. Energy research firm Rystad Energy projects investments in fuel cells for data centers to rise to about $30 billion by 2030 from $2.8 billion in 2025, a trajectory that reflects both the urgency of the AI buildout and the stubborn limits of the American electricity grid. A fuel cell generates electricity through a chemical reaction between a fuel, such as natural gas or hydrogen, and oxygen, converting the fuel’s energy directly into electricity rather than burning it. That distinction matters for data center operators who need power quickly, cleanly and without waiting years for a grid interconnection.
The scale of the problem is easy to underestimate. Data centers require massive amounts of reliable electricity to run computing equipment and cooling systems around the clock, and linking large electricity loads to the US grid can take three to six years, according to Rystad estimates. That timeline is incompatible with the pace at which AI companies are trying to deploy capacity. The result is a forcing function: developers are exploring on-site power sources that can be permitted, installed and switched on in months rather than years. Rystad predicts that about 40% of projected US data-center capacity in 2030 could pursue dedicated on-site power rather than rely entirely on the grid, a shift that would fundamentally change how the industry thinks about energy procurement.
Bloom Energy is the most visible beneficiary of this shift. The company operates solid-oxide technology ready for 100% pure hydrogen, while FuelCell Energy uses molten carbonate technology capped at 40% hydrogen blends, even as both systems are fuel-flexible. Bloom CEO KR Sridhar offered a concrete illustration of the speed advantage on a post-earnings call on Thursday. “We announced our first project with Oracle and delivered power to their Utah site in less than two months,” he said. “We are achieving the same thing with Nebius in New Jersey. We were able to get planning board approvals for our fuel cells within three months of signing the contract.” Those timelines are striking when set against the multi-year wait for grid interconnection, and they explain why operators under pressure to bring AI capacity online are increasingly willing to pay a premium for on-site generation.
The financial results tell a similar story. Bloom’s third-quarter revenue surged 138% to $1.24 billion, and the company posted net profit of $250.3 million, compared to a year-earlier loss of $23.1 million. It raised its 2026 revenue forecast to $4.4 billion to $4.6 billion. Oracle intends to procure up to 2.8 gigawatts of Bloom fuel-cell systems, of which 1.2 gigawatts has been contracted. Those figures suggest that fuel cells have moved from pilot projects to serious infrastructure commitments, and that at least one major cloud provider has decided the technology is reliable enough to anchor a significant portion of its power strategy.
There is an environmental argument here that deserves careful scrutiny. Fuel cells produce lower local emissions than conventional gas-fired generation because they avoid combustion within the stack. That is a genuine advantage for air quality and for permitting in populated areas, where opposition to conventional generation can delay projects indefinitely. But the climate picture depends heavily on the fuel source. Natural gas still produces carbon dioxide, and hydrogen supply remains limited and expensive in most US markets. Bloom’s ability to run on 100% pure hydrogen is a promising capability, yet the infrastructure to deliver that hydrogen at scale is not yet in place. For now, most fuel cell deployments will run on natural gas, which means they are cleaner than the alternatives at the point of generation but not emissions-free.
The grid itself is another consideration. On-site generation solves a timing problem for individual operators, but it does not address the broader question of whether the US electricity system can support the AI buildout over the long term. If a large share of new data-center capacity bypasses the grid entirely, utilities may find themselves with less revenue to fund the transmission and distribution upgrades that other customers depend on. There is also the risk that on-site generation becomes a way for large technology companies to insulate themselves from grid reliability problems while smaller customers bear the costs of an aging system. Those equity questions are not yet central to the fuel cell conversation, but they will become harder to ignore as the technology scales.
For anyone who has watched the energy industry cycle through hype and disappointment, the current enthusiasm for fuel cells carries a familiar risk. The technology has been promising for decades, and previous waves of interest faded when subsidies expired or natural gas prices shifted. What is different this time is the demand signal. AI data centers are not a speculative market; they are being built at extraordinary speed by companies with enormous balance sheets, and those companies need power now. That urgency gives fuel cells a commercial foothold that earlier booms lacked. Whether that foothold becomes a permanent position or a temporary bridge to grid-scale renewables and storage will depend on hydrogen economics, permitting reform and the willingness of hyperscalers to keep paying a premium for speed. The market has clearly made a bet, and the next few years will reveal whether it was the right one.
