FuelCell Energy Swings Toward Data Center Power With $3.6 Billion Backlog, But Legacy Cost Structure Still Bleeding Cash
Q3 fiscal 2026 earnings call, September 2, 2026
FuelCell Energy used its fiscal third-quarter call to lay out a pivot toward AI data center power that is starting to show up in contracted numbers, even as the financials made clear the company is still absorbing painful costs from its legacy manufacturing base. Total committed and awarded capacity backlog jumped to $3.6 billion as of July 31, up from a much smaller base, with $1.3 billion of that classified as committed backlog and a newly disclosed $2.4 billion bucket called awarded capacity backlog. CEO Jason Few framed the quarter as the moment the company converted a "growing pipeline" into "tangible commercial commitments," but the underlying earnings print shows how expensive that transition still is.
Fit Energy Deal Is the Commercial Proof Point, and the Cost Culprit
The centerpiece of the quarter is a capital equipment purchase agreement with a company called Fit Energy covering up to 380 megawatts across four phases, structured so that only the initial 30-megawatt Phase 0 is currently committed, with Phases 1 through 3 sitting in the new awarded capacity category and subject entirely to Fit Energy's discretion. Few called the deal "commercial proof that FuelCell Energy Block system can be the solution to some of the current public perception challenges facing data centers by providing scalable, clean, quiet, behind-the-meter power." Management was explicit that phases are not sequential and Fit Energy can elect Phase 1 without waiting for Phase 0 to be operational.
The Fit Energy order is also directly responsible for the quarter's ugly gross margin performance. FuelCell Energy is currently producing at an annualized rate of just 37 megawatts, well below the volume needed to make its cost structure match the contractual pricing embedded in the Fit Energy agreement. That mismatch drove roughly $17 million of charges in the quarter, split between $4 million to write down inventory to net realizable value and $13 million for losses on firm purchase commitments. CFO Michael Bishop was careful to frame these as tied to identified inventory and Phase 0 commitments rather than a read on the deal's long-term economics, but the charges were the single biggest driver of a $24.5 million gross loss for the quarter, compared with a $5.1 million gross loss a year earlier. Revenue fell 29% year-over-year to $33 million, with declines across every category as the company wound down its Gyeonggi Green Energy repowering work in South Korea and Groton generation output went offline for a planned upgrade.
New Backlog Taxonomy Signals a Business Model Shift Toward Capacity Reservations
Management's decision to split backlog into committed and awarded capacity categories is itself a notable disclosure change, reflecting what Few described as customers increasingly wanting to reserve manufacturing capacity while they finalize data center designs and lock in their own offtake agreements. Subsequent to quarter-end, the company closed a 75-megawatt capacity reservation agreement with a major co-location data center operator in Texas, which management said comes with expectations of follow-on opportunities with the same customer. Bishop indicated future reservation deals, including the Texas agreement, will flow into the awarded capacity bucket going forward, describing the approach as following "industry practice these days where folks are disclosing reservations whether they're called CRAs or something else." Investors should treat the awarded capacity figure with caution, however; Bishop was explicit that "awarded capacity backlog is not contracted firm order backlog or a guarantee of future revenue," and amounts may never convert, or convert on a materially different timeline than currently modeled.
Manufacturing Ramp Is the Real Gating Factor for Profitability
FuelCell Energy is targeting an annualized production rate of 100 megawatts at its Torrington, Connecticut facility by October 2026, nearly tripling the current 37-megawatt rate, en route to a longer-term goal of 500 megawatts of annualized capacity by June 2028. The company says this expansion, estimated to cost between $200 million and $275 million, is fully funded and is being executed in step with committed backlog rather than ahead of demand. A new high-volume tape caster now being installed is described as the key bottleneck-breaker for future capacity. Management pushed back its full-year capital expenditure guidance to $10 million to $20 million from a prior $20 million to $30 million range, attributing the change to equipment delivery timing rather than execution slippage.
Perhaps the most consequential new disclosure was a revised profitability target: FuelCell Energy now expects to reach positive adjusted EBITDA in the fourth quarter of fiscal 2027, a full year out. Bishop tied that target to at least 100 megawatts of production volume, continued conversion of awarded capacity into committed contracts, and cost reductions from higher procurement volumes, while acknowledging "there can be no assurance" the company hits these production rates or conversion timelines. Adjusted EBITDA for the quarter came in at negative $36.7 million, worse than the negative $16.4 million posted a year ago, though management attributed the deterioration entirely to the Fit Energy-related charges rather than any structural change in the underlying business.
Balance Sheet Strength Buys Time, But Dilution Has Been Heavy
The company ended the quarter with $737.3 million in total cash, cash equivalents, and restricted cash, up sharply from $440.9 million at the end of the prior quarter, with $658.1 million of that unrestricted. FuelCell Energy carries no corporate convertible or high-yield debt, with its $153.6 million of total debt consisting mostly of project-level financing and Export-Import Bank facilities tied to Korean deliveries. That cash cushion was built substantially through equity issuance: the company raised roughly $298 million in net proceeds during the quarter, including $245.5 million from a July underwritten offering and $52.9 million through its open-market sale program. The weighted average share count has climbed to 70.4 million shares, a reminder that the improved per-share loss figures partly reflect dilution rather than pure operational improvement. Net loss for the quarter was $45.3 million, or $0.64 per share, versus $92.5 million, or $3.78 per share, a year earlier, though the prior-year period included $68.5 million of asset impairment and restructuring charges that inflate the year-over-year comparison.
Carbon Capture With ExxonMobil Moves From Lab to Refinery
Beyond the data center narrative, FuelCell Energy delivered and installed the first two carbonate fuel cell carbon capture modules at ExxonMobil's Rotterdam complex under a joint development agreement, which Few called "the world's first industrial-scale demonstration" of the jointly developed technology. The Rotterdam units are designed to capture more than 90% of carbon directly from a low-concentration emissions stream, which management argues is both technically harder and commercially more valuable than capturing from concentrated sources, since it widens the addressable market to industries ranging from bottling plants to refineries. Few noted that federal support via 45Q tax credit provisions under the OBBBA has actually improved, alongside what he described as strong policy tailwinds in Europe and Asia. Every energy block the company currently ships is "carbon capture ready," according to Few, though the Rotterdam installation itself is optimized specifically for capture rather than power generation.
Siemens Partnership Targets the Electrical Complexity Behind AI Workloads
A memorandum of understanding signed with Siemens during the quarter is aimed at the electrical balance-of-plant systems needed for large fuel cell installations exceeding 100 megawatts, combining FuelCell Energy's generation technology with battery storage, microgrid controls, and medium-voltage equipment. Few described the goal as managing "the full spectrum of power variability from minutes down to microseconds," a capability he argued is necessary to support high-density AI workloads reliably. On the supply chain side, management emphasized that its carbonate fuel cell platform is scandium-free and does not depend on rare earth minerals, drawing instead on globally abundant nickel and steel, with more than 90% of its supply chain sourced domestically and roughly 93% of energy block components reusable or recyclable through a take-back program. In a call dominated by AI power narratives, this was pitched as a differentiator against geopolitical supply risk facing other clean energy hardware makers.