After Bloom’s Monday share pop, KR Sridhar (shown here in 2020) founded the company in 2001.
Tim Pannell for Forbes
Would you pay 18 times revenues for shares of a 24-year-old fuel cell maker that has never turned a profit?
Bloom Energy makes solid oxide fuel cells — space age looking boxes that generate electricity from natural gas and emit carbon dioxide. KR Sridhar founded the company in 2001. As the legend goes, Bloom’s tech began with Sridhar’s work devising oxygen-producing machines for NASA; he later figured he could turn the device around to produce electricity. In the years since, Bloom has deployed thousands of its systems, which today generate some 1.4 gigawatts of electricity, enough to power about a million homes.
And yet Bloom has generated only losses for its shareholders. Bloom’s cumulative historic deficit, per SEC filings, is $4 billion. Net losses the past three years have averaged more than $200 million annually.
But who needs profits when the market is obsessed with AI datacenters? On Monday real estate giant Brookfield announced a $5 billion, multi-year deal with Bloom to deploy their boxes to ensure reliable power at Brookfield’s wave of hyperscale datacenters, especially in Europe. Bloom shares rocketed 27% on the news. And no wonder; the headline dollar figure equates to more than three years of Bloom revenues.
It was a big day for CEO Sridhar, 64, by far the biggest Bloom shareholder with 2.7%. Though he’s been selling in recent months, Sridhar’s 4.6 million shares are now worth $490 million. Bloom shares (at $106 Tuesday morning) are up more than 900% in the past 12 months to a market cap of $25 billion.
Could Brookfield be Bloom’s big break? After decades of trying, Bloom boxes just haven’t caught on. They’re best for off-grid applications, “distributed generation.” They are not an alternative to standby backup generators because they take hours to warm up and are designed to work continuously. Nor do Bloom boxes help much in reducing carbon dioxide, with emissions on par with advanced, utility-scale gas turbines (about 900 pounds of CO2 per megawatthour).
Aside from CO2, Bloom boxes’ emission of noxious compounds is low – but that’s because elaborate filtration mechanisms capture the bad stuff. These filters full of hazmat (like benzene and sulfur compounds) get replaced every year or so. As a result, every single one of Bloom’s thousands of units producing 1.4 gigawatts is under a service and maintenance contract. Analysts at TD Cowen in a fuel cell report over the summer figured that power from fuel cells cost at least 1.5 times as much deploying a solar field plus batteries.
Analyst Maheep Mandloi at Mizuho sees Bloom providing one-sixth of Brookfield’s datacenter power generation over 5 years, equating to about 200 megawatts per year of Bloom box orders. But given Bloom’s “constrained” manufacturing capacity, much of that will be backend loaded.
The most likely destinations for Bloom are Italy and France because those countries already rely on gas for a large portion of power generation, figures analyst Dushyant Ailani at Jefferies. He notes that Bloom “is benefitting from a clear investor thirst to invest in the thematic of additionality and speed-to-power with less regard to the fundamentals and duration of the cycle.” Indeed $5 billion in fuel cells over years sounds like a lot until you consider that Brookfield affiliates are working on a $23 billion, 1 gigawatt datacenter in France and a $10 billion, 750 mw center in Sweden.
Bloom has plenty of datacenter experience with the likes of Google, Amazon, Intel and Oracle. But there have been some durability concerns, with units needing to be retrofitted or “repowered” after six years or so as components break down under 1,000-degree operating temperatures.
It’s not a good look when a manufacturer has to buy back its equipment. Bloom in recent SEC filings disclosed several situations in recent SEC filings where it has had to buy back more than $100 million worth of its boxes. One involved “failed sale and lease-back transactions termination accounting” associated with a 2015 project (PPA V) backed by Constellation and Intel. Bloom ended up buying back 37 megawatts of systems operating under long-term contract, found a new financier to underwrite replacing the systems, and took a $124 million impairment charge.
Hopefully Bloom’s growing pains and big charges are finally behind it. Net losses were $66 million in the first half of 2025 (on $730 million revenue), versus a $119 million loss in the same period a year ago. Priced at nearly 18 times revenues, shares in Bloom are levitated by the AI bubble, for now.