FuelCell Energy: The Data Center Story Is Priced Ahead of the Math

A fuel cell company that spent the past year being recast as an AI data center play just reported a quarter where revenue fell and gross losses widened sharply, and the stock dropped 16% the day the news came out. The surprise is not the drop. It is that the market had spent the preceding year paying up for a story the company's own numbers still do not support.

That company is FuelCell Energy (FCEL), a manufacturer of large carbonate fuel cells that has been trying to pivot from utility and hydrogen projects to powering data centers. The pivot gave the stock its run: it is up roughly 247% over the past year, hit a peak near $38, and still trades around $14 even after this pullback, giving it a market value near $1.1 billion. The catalyst was a string of data center deals, led by a June agreement with FIT Energy to supply up to 380 megawatts of power for data centers, and the promise that this would turn a lumpy project business into a growth engine.

The numbers behind that promise, reported on September 2 for the fiscal third quarter ended July 31, run the other way. Revenue fell 29% year over year to $33 million. The company recorded a gross loss of $24.5 million, roughly five times the year-ago loss, and it booked a $17 million inventory charge because the price on the initial FIT Energy deal came in below what it costs to build the systems. Adjusted EBITDA losses more than doubled to $36.7 million. The net loss improved only because the prior-year period was burdened by one-time impairment and restructuring charges; the quarter was not actually better.

This is the part that matters for anyone tempted to treat the attention as the thesis. The flagship data center product is, at current scale, being sold below cost, and that scale is small: the company ran at an annualized production rate of about 37 megawatts during the quarter, targeting 100 megawatts by October and 500 megawatts by 2028. Unit margins stay negative until that ramp happens. Management's own profit target is positive adjusted EBITDA in the fiscal fourth quarter of 2027 — more than a year away, and conditional on converting options into orders, hitting delivery schedules, and cutting costs.

The disconnect between the story and the math is most visible in the backlog headline everyone cheered. FCEL reported total "committed and awarded" backlog of $3.65 billion. But only $1.3 billion of that is committed — firm orders the company can count as revenue — and that figure grew just 4.1% from a year earlier. The other $2.35 billion was added in a single quarter and is "awarded capacity backlog," which the company itself describes as non-binding option capacity, not a firm order or a guarantee of future revenue. The 380-megawatt FIT deal, for instance, contains a firm 30-megawatt commitment and options for the rest, with delivery of that first phase only expected in the current fiscal fourth quarter.

The other way the story outruns the math is dilution. FCEL's $737 million cash pile looks reassuring on its face, but it was funded by selling shares, not by making money. The company did a 1-for-30 reverse split in November 2024 and has issued heavily since; shares outstanding roughly doubled from about 46 million at fiscal 2025 year-end to near 80 million by July 2026, including a July offering of 12.3 million shares that raised about $245 million. Free cash flow is still around negative $130 million over the trailing twelve months. For a value lens, this is the mirror image of the mispricings this strategy normally hunts for: instead of a cheap multiple below a real growth rate, FCEL trades near 6.7x trailing sales with no earnings to anchor a price-to-earnings multiple at all.

None of this has stopped AInvest's aggregate signal from labeling the stock a Buy, and the sales pipeline has jumped to about 10 gigawatts with data centers driving most of it. Sentiment is not the problem here; the economics are.

The honest read is that the market has priced the AI data center story as though the firm orders already existed, while the firm order book barely moved and the current product still loses money on every megawatt it ships. The check that would turn the story into math is specific and observable: whether the $2.35 billion of awarded options convert into firm committed backlog, whether gross margin turns positive as production scales toward 100 megawatts, and whether the positive-EBITDA target for fiscal 2027 can be reached without another dilutive share sale. Until that happens, a $1.1 billion market cap is paying for a narrative, and the numbers say the earnings are not there yet.