Ceres Power Says Solid Oxide Fuel Cells Are Now Cost-Competitive With Gas Turbines, With Weichai Scaling to 200 Megawatts in Under Two Years
Interim results call, September 23, 2026
Ceres Power used its interim results presentation to make a claim that would have sounded aspirational just a year or two ago: its solid oxide fuel cell technology is now cost-competitive with conventional power generation, not merely faster to deploy. CEO Phil Caldwell told investors the company sits "there or thereabouts" on cost versus gas turbines and reciprocating engines today, a milestone that changes the investment case from a niche backup-power story to a genuine alternative for behind-the-meter power at scale.
The timing argument remains central to the pitch. Caldwell noted that gas turbine lead times now stretch five to seven years, nuclear capacity is "into the next decade," and grid connections take five to ten years in many markets, including the U.K. Ceres believes this creates what Caldwell called a "five-year window" to establish solid oxide as a permanent fixture in the power mix rather than a stopgap. The company's own execution has compressed dramatically in that time: the first factory built with Doosan took roughly four years around the Covid period; the most recent license signed with China's Weichai is on track for full production in two years or less, a function of standardizing around the newly launched Endura stack platform.
Weichai's speed and Doosan's first export order
Among the existing licensee base, Weichai's scale-up stood out as the most aggressive. The company is targeting first production later this year or early next, with a stated goal of 200 megawatts of capacity within two years of signing, a pace management repeatedly flagged as unusually fast even by Ceres' own accelerating benchmarks. Doosan in South Korea, an earlier-generation licensee still running Ceres' original, smaller-footprint stack rather than Endura, signed its first export contract, a roughly GBP 60 million stack supply deal with Germany's Reverion. Reverion's system is notable for running in both directions, generating power when needed and synthetic methane when it isn't, which Caldwell described as a way to "arbitrage" both sides of the energy transition. Royalties on that contract fall within the company's standard range of $50 to $100 per kilowatt depending on scope.
Delta in Taiwan is running initial production at its Tainan facility and has announced a second, larger plant in Guanyin for mass-market manufacturing, reinforcing Delta's position as one of the ecosystem's best-capitalized partners given its existing scale in data center hardware. In Japan, DENSO has secured GBP 165 million in government support to continue developing SOEC hydrogen technology, though Caldwell was clear that hydrogen remains a secondary priority behind power generation in the near term.
The 800-volt DC angle investors haven't fully priced
The most novel technical argument in the presentation centered on data center power architecture. Caldwell pointed to Nvidia's push toward 800-volt DC infrastructure for AI data centers as a structural tailwind for fuel cells specifically, because they are "DC native," generating direct current rather than the alternating current produced by turbines and engines. Nvidia's own estimate, cited by Caldwell, is roughly 5% efficiency gains from going DC-direct in a data center, alongside reduced copper usage and elimination of switchgear and transformers, and the ability to push more power into each server rack. Management expects this requirement to firm up around 2028-2029, layering a forward-looking technical moat on top of the near-term cost and speed arguments.
Permitting emerged as the third leg of the thesis, and arguably the most durable one. Caldwell described instances where power generation contracts originally awarded to conventional turbines or reciprocating engines have been "flipped" to solid oxide, driven by near-zero water usage, absence of combustion byproducts like SOx and NOx, and near-silent operation that allows siting closer to urban centers and inside buildings. "Just because you can build power doesn't mean you should build power or you have the right to build power," Caldwell said, framing permitting friction, not generation capacity, as the binding constraint many data center developers now face.
Guidance and financials hold steady
Ceres reiterated full-year guidance of GBP 45 million in contracted revenue, having delivered roughly half of that in the first half, a pace CFO Stuart Paynter characterized as being on track. Gross margin remains high given the licensing-heavy revenue mix. R&D spending fell to GBP 18 million in the first half from GBP 25 million a year earlier, which management attributed to Endura moving from development into commercialization rather than a pullback in innovation capacity. The company raised just over GBP 100 million in an oversubscribed equity issuance earlier in the year, a move Caldwell framed less as a funding necessity and more as a signal to partners making multi-decade factory investments that Ceres will be around as their technology provider "for the next 20 years." Management reiterated confidence in signing at least one new manufacturing licensee this year, describing the pipeline as growing across the U.S., Asia and Europe, though Paynter cautioned that contract negotiation and revenue recognition timing remain inherently unpredictable, noting the Weichai agreement's revenue recognition schedule was complicated by its status as an update to an older 2018 contract. The company said it is now templating new contracts to make recognition timing clearer.
Royalties still years from mattering
Perhaps the most sobering disclosure for investors focused on near-term profitability: royalties, the revenue line that ultimately justifies the licensing model, remain immaterial today and are not expected to become significant until partners scale cumulative deployments toward roughly one gigawatt, which management pegs to the late part of this decade. Until then, the business remains dependent on upfront license fees and engineering services revenue tied to new signings, reinforcing why the cadence of new manufacturing license agreements, targeted at roughly one per year, matters so much to the equity story. Caldwell noted the company has not fundamentally altered its licensing fee structure despite stronger negotiating leverage from surging demand, wary of front-loading fees in a way that could distort partners' investment decisions before they reach payback.
Ceres sizes the addressable power generation market at 22 gigawatts by the end of the decade, with data centers representing about half of that opportunity, and a separate 38-gigawatt hydrogen market opportunity emerging by 2035 built on the same underlying stack technology. The dual-market positioning, power now, hydrogen later, was presented as a deliberate hedge rather than an either-or bet, though it also means investors are being asked to underwrite a technology transition whose economics won't be fully visible for several more years.