Nuclear Hydrogen: Make Fuel—or Sell the Power?

Start with the competing receipt

In January 2026, Vistra announced 20-year power-purchase agreements with Meta covering operating generation from Davis-Besse and Perry, plus uprates across three nuclear sites. Purchases begin in late 2026, and the electricity remains on the grid. The price is not public, but the commercial signal is clear: a nuclear megawatt-hour can attract a long-duration buyer without the owner building an electrolyser, storage system and hydrogen delivery chain.

That changes the test for nuclear hydrogen. The comparison is no longer an electrolyser against an abstract hour of surplus electricity. It is hydrogen netback against the best grid, capacity-market or bilateral power value available at that reactor, for that term, with that buyer credit. If the power cheque is larger and simpler, selling electricity is not a failed hydrogen project. It is the control case working as intended.

The scarce asset is not water-splitting chemistry. It is the reactor megawatt-hour.

Nine Mile Point proves a narrow, useful case

The strongest operating receipt in the project set is Nine Mile Point in New York. In 2023, a 1.25 MW low-temperature electrolyser began producing hydrogen for the nuclear station's own cooling system. The system replaces hydrogen previously made from fossil fuel off site and delivered to the plant. A 2023 DOE program review estimated more than US$800,000 per year of avoided external hydrogen purchases.

Its narrowness is the lesson. The reactor, interconnection and site already existed. The electrolyser can run at high utilization. The customer sits behind the fence. Compression, merchant delivery and speculative offtake are largely removed. The relevant counterfactual is not merely wholesale power; it is an expensive industrial product delivered by truck.

That does not establish the economics of a new reactor built to sell commodity hydrogen. It establishes a procurement-substitution niche: existing nuclear plus captive or adjacent high-value demand.

High capacity factor does not erase opportunity cost

A light-water reactor can feed an alkaline or PEM electrolyser using familiar low-temperature electrolysis. In the historical INL/NREL comparison used for this analysis, the electrical requirement is 50.2 kWh per kilogram of hydrogen. One electrical MWh therefore produces about 19.9 kg H₂ before compression, storage, delivery, degradation and auxiliary loads.

Nuclear-like utilization helps spread electrolyser capital over more kilograms. It also means consuming electricity during hours when the reactor could have sold power. The correct model therefore carries both sides of the allocation: realized hydrogen value and policy support, less conversion capital, variable cost and delivery, compared with the best electricity value displaced.

High-temperature electrolysis adds heat—and another risk register

Solid-oxide electrolysis supplies steam at roughly 700–800°C so heat provides part of the water-splitting energy. A historical NREL model used 35.1 kWh of electricity plus 11.15 kWh of thermal energy per kilogram. On that basis, one electrical MWh produces about 28.5 kg H₂ while drawing roughly 0.318 MWhth.

The electrical yield is attractive, but the heat is not free. Extraction steam can reduce turbine output. Heat exchangers, steam loops, controls and the safety case add equipment and licensing work. Electrolysers must demonstrate durability under high-temperature operation, and the reactor and hydrogen plant need a credible operating agreement about which asset follows the grid.

China's HTR-PM is an important reactor-platform proof point, but it is not a commercial hydrogen receipt. An April 2026 Tsinghua update reports progress on the distinct iodine-sulfur thermochemical route, including pilot demonstration R&D and commercial-scale demonstration work anticipated around 2027. Reactor operation, hydrogen R&D and a bankable hydrogen project are three different evidence rungs.

Better electrical yield is not the same as free heat, durable equipment or a contracted product.

The demonstration ladder still has missing rungs

DOE previously published a 2025 target for low-temperature hydrogen work at Davis-Besse and a 2026 target for a high-temperature demonstration at Prairie Island. The latest detailed DOE status snapshot located for this episode is from 2024: Nine Mile Point was producing, Davis-Besse was working through electrical and switchgear scale-up, and Prairie Island was a 150 kW high-temperature system tied to the plant's thermal line.

No newer primary hydrogen-commissioning notice was located for Davis-Besse or Prairie Island in the August 16 refresh. That is a source gap, not evidence of cancellation. It does, however, show why models, targets, site-integration demonstrations and operating customer receipts must stay on separate rungs. The top rung would disclose utilization, degradation, hydrogen delivery cost and customer payments from an integrated commercial plant.

Powering AI makes the control case financeable

Long-duration nuclear-power commitments now span multiple structures: Meta and Constellation announced an agreement for the full 1,121 MW output of Clinton beginning in 2027; Talen's expanded Amazon relationship supports up to 1,920 MW at Susquehanna; Google and Kairos are targeting up to 500 MW through 2035; and Amazon backed an Energy Northwest/X-energy plan beginning with 320 MW and expandable to 960 MW.

These announcements do not disclose one universal electricity price, and development agreements are not operating projects. They do establish a credible competing use with large-credit buyers. A hydrogen developer must therefore show why financing a second plant and a second product market produces a stronger risk-adjusted receipt.

Darlington is the honest small modular reactor anchor

Ontario Power Generation is building a 300 MW BWRX-300 at Darlington, targeting grid service by the end of 2030. OPG disclosed a C$7.7 billion budget for the first unit plus shared infrastructure, while the four-unit plan could total 1,200 MW. This is valuable evidence because the project has moved from presentation to construction and because its default product is electricity.

The BWRX-300 is a boiling-water reactor, not the high-temperature gas-cooled reactor required by the most aggressive heat-assisted concepts. Its disclosed budget also includes shared infrastructure intended for more than one unit, so dividing C$7.7 billion by 300 MW would not produce a clean overnight-cost comparison. A new modular-reactor hydrogen project stacks reactor licensing and schedule, first-unit capital, electrolyser performance, integration, hydrogen delivery, offtake and policy timing. Existing components do not make the combined project de-risked.

Run both contracts through the same calculator

The free editable Nuclear MWh Allocation Calculator keeps the reactor common to both output choices. It calculates kilograms per reactor MWh, realized hydrogen and policy value, conversion and integration cost, delivery, thermal opportunity cost and the resulting allocation advantage over electricity. Reactor construction cost is deliberately outside this incremental comparison; a new-build sponsor must first choose the higher-value product, then separately prove the reactor's cost and schedule.

The illustrative sensitivity shows the burden clearly. With the other new-modular defaults held constant, the hydrogen case turns slightly positive around US$5/kg when the power alternative is only US$30/MWh. At US$75/MWh power, it needs close to US$8/kg. Those are not forecasts. They are a way to identify which contract term actually flips the decision.

Policy can move the spread, but schedules can miss the window

U.S. 45V rules include a constrained route for qualifying existing nuclear electricity rather than an automatic credit for all nuclear hydrogen, and current IRS instructions make construction timing material. Canada's Clean Hydrogen investment tax credit is explicitly tied to lifecycle carbon intensity, with published capital-credit tiers of 40%, 25%, 15% and zero as emissions rise. The European Union's final 2025/2359 methodology keeps renewable fuels and low-carbon hydrogen distinct: nuclear hydrogen is not automatically renewable, but it can seek the low-carbon route subject to lifecycle and electricity-accounting rules.

Policy can alter hydrogen netback. It cannot erase a reactor schedule, a weak buyer or an incomplete delivery chain. A project whose incentive expires before commissioning has not solved the allocation problem.

Four cases, four calls

Existing reactor plus captive or adjacent hydrogen: real niche. Underwrite the avoided delivered product and keep the customer close. New small modular reactor plus conventional electrolysis: prove the premium. Require a firm buyer and a disclosed margin over power. Advanced reactor plus high-temperature conversion: watch. Demand integrated durability and customer receipts, not only efficiency. Sell electricity: control case. Underwrite the actual contract, interconnection and buyer credit; if that cheque is larger and simpler, take it.

The verdict changes when the evidence changes: a durable take-or-pay hydrogen contract, independently verified integrated operation, disclosed degradation and delivery cost, policy that survives the build schedule, or a power market weak enough that conversion clearly wins. Until then, nuclear hydrogen is not one colour with one score. It is four different capital-allocation decisions.

Research cut-off: August 16, 2026. Technology and project-economics commentary only; not investment advice. Primary sources include the Nine Mile Point operating notice, Vistra/Meta nuclear-power agreement, OPG Darlington project record, DOE high-temperature-electrolysis targets, final U.S. 45V rules, EU low-carbon-fuels methodology and Tsinghua's April 2026 nuclear-hydrogen update.