The stock began soaring (again) in the middle of last year as anything and everything plausibly connected to artificial intelligence was going vertical. Today, Bloom Energy is trading at more than $111 a share, with a market cap north of $26 billion — and that’s after a dramatic fall from its all-time high price of over $135 per share, reached in November. By contrast, Southwest Airlines is worth around $22 billion; Edison International, the parent company of Southern California Edison, is worth about $22.5 billion.
This is all despite Bloom recording regular losses according to generally accepted accounting principles, although its quarterly revenue has risen by over 50%, and its reported non-GAAP and adjusted margins and profits have grown considerably. The company has signed deals or deployed its fuel cells with Oracle, the utility AEP, Amazon Web Services, gas providers, the network infrastructure company Equinix, the real estate developer Brookfield, and the artificial intelligence infrastructure company CoreWeave, Bloom’s chief executive and founder, KR Sridhar, said in its October earnings call.
While fuel cells have been pitched for decades as a way to safely use hydrogen for energy, fuel cells can also run on natural gas or biogas, which the company has seized on as a way to ride the data center boom. Bloom leadership has said that the company will double its manufacturing capacity by the end of this year, which it says will “support” a projected four-fold annual revenue increase. “The AI build-outs and their power demands are making on-site power generated by natural gas a necessity,” Sridhar said during the earnings call.
To get a sense of how euphoric perception of Bloom Energy has been, Morgan Stanley bumped its price target from $44 dollars a share to $85 on September 16 — then just over a month later, bumped it again to $155, calling the company “one of our favorite ‘time to power’ stocks given its available capacity and near-term expansion plans.”
Bloom has also won plaudits from semiconductor and data center industry analysts. The research firm SemiAnalysis described Bloom’s fuel cells as a “a fairly niche solution [that] is now taking an increasingly large share of the pie.”
It’s been a long journey from green tech darling to AI infrastructure for Bloom Energy — and fuel cells as a technology.
Bloom was founded in 2001, originally as Ion America, and quickly attracted high profile Silicon Valley investors. By 2010, fuel cells (and Bloom) were still being pitched as the generation source of the future, with The New York Times reporting in 2010 that Bloom had “spent nearly a decade developing a new variety of solid oxide fuel cell, considered the most efficient but most technologically challenging fuel-cell technology.” That product launch followed some $400 million in funding, and Bloom would hit an almost $3 billion valuation in 2011.
By 2016, however, when the company first filed with the Securities and Exchange Commission to sell shares to the public, it was being described by the Wall Street Journal as “a once-ballyhooed alternative energy startup,” in an article that said the fuel cell industry had been an “elusive target for decades, with a succession of companies unable to realize its business potential.” The company finally went public in 2018 at a valuation of $1.6 billion.
Then came the AI boom.
Fuel cells don’t use combustion to generate power, instead combining oxygen ions with hydrogen from natural gas and generating emissions of carbon dioxide and water, albeit without the particulate pollution of other forms of fossil-fuel-based electricity generation. This makes the process of getting permits from the Environmental Protection Agency “significantly smoother and easier than that of combustion generators,” SemiAnalysis wrote in a report.
In today’s context, Bloom’s fuel cells are yet another on-site, behind-the-meter natural gas power solution for data centers. “The rapid expansion of AI data centers in the U.S. is colliding with grid bottlenecks, driving operators to adopt BTM generation for speed-to-power and resilience to their modularity, fast deployment, and ability to handle volatile AI workloads,” Jefferies analyst Dushyant Ailani wrote in a note to clients. “Natural gas reciprocating engines, Batteries, and Bloom fuel cells are emerging as a preferred solution due to their modularity, fast deployment, and ability to handle volatile AI workloads.”
SemiAnalysis estimates that capital expenditure for Bloom fuel cells are substantially higher than those for gas turbines on a kilowatt-hour basis — $3,000 to $4,000 for fuel cells, compared to between $1,500 and $2,500 for turbines. But where the company excels is in speed. “The big turbines are sold out for four or five years,” Maheep Mandloi, an analyst at Mizuho Securities, told me. “The smaller ones for behind the meter for one to two years. These guys can deliver, if needed, within 90 days.”
Like other data center-related companies, Bloom has faced some local opposition, though not a debilitating amount. In Hilliard, Ohio, the state siting board overrode concerns about the deployment of more than 200 fuel cells at an AWS facility.
Bloom is also far from the only company that has realigned itself to ride the AI wave. Caterpillar, which makes simple turbine systems largely for the oil and gas industry, has become a data center darling, while the major turbine manufacturers Mitsubishi, Siemens Energy, and GE Vernova have all seen dramatic increases in their stock price in the last year. Korean industrial conglomerate Doosan is now developing a new large-scale turbine. Even the supersonic jet startup Boom is developing a gas turbine for data centers.
While artificial intelligence — or at least artificial intelligence companies — promises unforeseen technological and scientific advancements, so far it’s being powered by the technological and scientific advancements of the past.
Homes in Greenland's snowbound capital of Nuuk. Leon Neal/Getty Images