Bloom Energy reported second-quarter revenue of $1.065 billion after the close on Tuesday, up 165.5% from $401.2 million a year ago. It is the first billion-dollar quarter in the company’s history. Product revenue — the fuel-cell hardware itself, which is the line that tracks data center demand — rose 215.4% to $935.4 million from $296.6 million. Shares climbed roughly 11% after hours, to about $186.
The margin story is the part that will get read twice. Gross margin came in at 33.4%, up 668 basis points from 26.7% a year earlier. Operating income was $182.2 million, against a $3.5 million operating loss in the same quarter last year. Adjusted earnings landed at $0.78 a share versus roughly $0.41 expected.
Management then raised the year. Full-year revenue guidance moved to $3.9–$4.2 billion from $3.4–$3.8 billion, a midpoint implying about 100% growth. Adjusted EPS guidance went to $2.55–$2.85 from $1.85–$2.25.
The math the bears were pointing at
We wrote about this print before it landed. Two things were hanging over it: a pair of short-seller reports questioning the “$20 billion contracted backlog” Bloom cites in commercial materials against the roughly $492 million of remaining performance obligations disclosed in its SEC filings, and a second-half revenue ramp the old guidance required.
The second problem just got smaller. First-half revenue is $1.82 billion — $751.1 million in Q1 plus $1.065 billion in Q2. Hitting the new $4.05 billion midpoint requires about $2.23 billion in the back half, or roughly $1.12 billion a quarter. That is only about 5% above what Bloom just printed.
Our take: This is a raise anchored to a run-rate that already exists, not to a hockey stick. In April the company was defending a guide that needed acceleration it hadn’t yet demonstrated; tonight it raised into one it has. That is the single most important change in the story, and it is why the stock moved. The backlog gap is a separate argument and it stays open. A figure quoted in commercial materials that doesn’t reconcile to the audited line in the filings is a disclosure problem, not automatically a demand problem — and both can be true at once. The demand is now visibly landing in the P&L. The $20 billion still needs a definition anyone can check.
Why an AI story wears a fuel-cell costume
Bloom doesn’t sell cheap power. It sells fast power. Its pitch to hyperscalers and colocation operators is on-site generation that arrives on a construction schedule instead of an interconnection-queue schedule — the constraint that has quietly become the gating factor on AI capacity in several US markets. Every gigawatt announcement that lands with a multi-year utility timeline attached is, indirectly, a sales lead for someone selling behind-the-meter generation.
That is also the risk. The same demand that produced a 215% product-revenue quarter is concentrated, lumpy, and dependent on customers whose capex plans get re-cut every earnings season. Bloom is now a levered bet on the AI build continuing at pace — a bet that has been rewarding revenue growth and punishing anything that smells like an unfunded obligation.
What to watch
- The 10-Q. Whether remaining performance obligations move meaningfully toward the $20 billion figure or stay near the $492 million previously disclosed. That is the number that settles the argument.
- Q3 gross margin. 33.4% arrived with volume behind it. The question is whether it holds as the customer mix shifts and installation costs scale.
- Named customers. Bloom’s disclosure has leaned on aggregate demand. Signed, named data center contracts would do more for the story than another guidance raise.
- Another raise in October. The current midpoint asks for about 5% quarterly growth from here. If Q3 comes in hot, the guide is conservative by construction.
- Financing terms across the AI power complex. Who carries the obligation matters as much as who signs the contract — as Meta’s data center structure showed this week.