Small Modular Reactors in 2026: Order Books, AI Off-take, and the Capex Curve
The first wave of SMR designs has moved from licensing slides to concrete pours and signed power purchase agreements, but the NuScale UAMPS cancellation still anchors investor memory and the load is now hyperscaler, not municipal.
Advanced nuclear is no longer a slide deck category. TerraPower has broken non-nuclear ground at Kemmerer, X-energy has signed a definitive engineering contract with Dow Seadrift, Holtec is preparing the SMR-300 around the Palisades restart, Kairos has the first Hermes test reactor under construction with a Google off-take for 500 megawatts, and the United States Nuclear Regulatory Commission is finalizing 10 CFR Part 53. The NuScale UAMPS cancellation in November 2023 remains the cautionary tale, with a quoted price drift from 58 dollars per megawatt hour in 2021 toward 89 dollars and ultimately 119 dollars by mid 2023. Capex is still binding, levelized cost runs 80 to 180 dollars per megawatt hour depending on financing, and the Inflation Reduction Act Section 45U production tax credit is now load bearing for project finance.
The setup entering 2026 #
The advanced nuclear market in 2026 is structured by three forces that did not coexist a year earlier. First, the hyperscaler off-take bid has compressed the demand-side risk that killed the NuScale Carbon Free Power Project. Microsoft signed a 20-year power purchase agreement around the Three Mile Island Unit 1 restart now branded the Crane Clean Energy Center, Amazon Web Services anchored data center load adjacent to the Talen Susquehanna site under a 960 megawatt arrangement, and Meta signed a memorandum of understanding with Constellation in late 2024. Second, the Nuclear Regulatory Commission is finalizing 10 CFR Part 53, the technology inclusive risk informed framework mandated by the Nuclear Energy Innovation and Modernization Act, with a final rule expected during 2025. Third, the Inflation Reduction Act Section 45U production tax credit, set at a base 3 dollars per megawatt hour and rising to 15 dollars with prevailing wage and apprenticeship compliance, has changed the merit order for any reactor that can clear a 2032 commercial operation date.
The NuScale cancellation still conditions every conversation. The UAMPS Carbon Free Power Project was sized at 462 megawatts across a six module VOYGR-6 configuration at Idaho National Laboratory, with target commercial operation in 2030. The contracted price escalated from 58 dollars per megawatt hour in the 2020 cost estimate to 89 dollars in early 2023 and reached 119 dollars per megawatt hour by termination on 8 November 2023, after subscription levels fell below the threshold needed to advance. Capex per kilowatt at termination implied roughly 9,300 dollars on an overnight basis. The lessons priced into 2026 deals are unambiguous: anchor off-take must be investment grade, supply chain inflation needs explicit pass-through, and the construction package must absorb first of a kind risk through fixed price elements rather than spreading it across municipal subscribers.
The active project map #
TerraPower broke non-nuclear ground on the Natrium demonstration plant at Kemmerer, Wyoming on 10 June 2024, the first sodium fast reactor construction start in the United States in decades. Natrium pairs a 345 megawatt sodium-cooled reactor with a molten salt thermal storage system that can dispatch up to 500 megawatts for several hours, sized to firm intermittent renewables on the PacifiCorp system. Total project cost is in the 4 billion dollar range with a Department of Energy Advanced Reactor Demonstration Program cost share, target commercial operation has slipped from 2028 toward 2030 to 2031 because of HALEU supply timing, and the construction permit application is under NRC review.
X-energy signed a definitive engineering, procurement, and construction services agreement with Dow in March 2024 covering the Long Mott Generating Station at the Dow Seadrift industrial complex on the Texas Gulf Coast. The configuration is four Xe-100 high temperature gas-cooled reactor modules at 80 megawatts electric each, generating both electricity and high temperature process steam to displace natural gas combustion at the chemicals site. Combined output is approximately 320 megawatts electric plus 800 megawatts thermal, the construction permit application was submitted to the NRC in 2025, and Amazon Web Services committed an additional 500 million dollars in the X-energy Series C-1 in October 2024.
Holtec is moving on two fronts. The SMR-300, a 300 megawatt pressurized water reactor design, is targeted for first deployment at the Palisades site in Covert, Michigan, where Holtec is restarting the existing 800 megawatt unit with a 1.52 billion dollar Department of Energy loan guarantee announced in March 2024. The SMR-300 construction permit application is targeted for 2026 with the first twin-unit deployment intended at Palisades. Kairos Power has Hermes, a 35 megawatt thermal non-power demonstration reactor, under construction at Oak Ridge after the NRC issued a construction permit in December 2023, the first non light water reactor permit in the United States in over 50 years. Google signed a master agreement with Kairos in October 2024 for up to 500 megawatts across multiple plants.
| Project | Design | Capacity | Site | Off-taker | Target COD |
|---|---|---|---|---|---|
| TerraPower Natrium | Sodium fast plus storage | 345 MW plus 500 MW peak | Kemmerer, Wyoming | PacifiCorp utility load | 2030 to 2031 |
| X-energy at Dow | HTGR pebble bed | 320 MWe plus 800 MWth | Seadrift, Texas | Dow chemicals onsite | 2030 |
| Holtec SMR-300 | Light water PWR SMR | 2 x 300 MW | Covert, Michigan | Michigan utility load | 2030 to 2031 |
| Kairos Hermes plus commercial | Fluoride salt cooled HTR | 35 MWth demo, then 50 MW | Oak Ridge, then TVA region | Google up to 500 MW | 2027 demo, 2030 commercial |
| Oklo Aurora | Compact sodium fast | 15 MW initial, scaling | Idaho National Laboratory | Equinix, Diamondback | 2027 to 2028 target |
| Microsoft Three Mile Island | Existing PWR restart | 835 MW (Unit 1) | Londonderry, Pennsylvania | Microsoft 20-year PPA | 2028 |
| Last Energy 20 MW pilot | Compact PWR | 20 MW per unit | United Kingdom and Texas | Industrial off-takers | 2027 target |
The data center off-take stack #
Hyperscaler off-take has moved from optional ESG framing to load-bearing project finance. Microsoft signed a 20-year power purchase agreement with Constellation Energy on 20 September 2024 to support the restart of Three Mile Island Unit 1, an 835 megawatt pressurized water reactor that ceased operation in 2019 for economic reasons, with site investment guided at 1.6 billion dollars. Amazon Web Services announced in March 2024 the acquisition of the Cumulus Data campus adjacent to the Talen Susquehanna nuclear plant for 650 million dollars, with rights to scale data center load up to 960 megawatts behind the meter. The Federal Energy Regulatory Commission rejected the initial interconnection service amendment in November 2024, and a revised arrangement is now structured around front of the meter delivery. Meta signed a request for proposal in December 2024 for up to 4 gigawatts of new nuclear capacity in the early 2030s.
Promethean models 2026 to 2030 hyperscaler nuclear off-take at 15 to 25 gigawatts of firm contracted volume, dominated by existing reactor uprates and restarts because new build SMR capacity will not be online in time. The 2030 to 2035 window is where SMR deployment begins to matter for AI load.
Part 53, HALEU, and the regulatory bottlenecks #
The Nuclear Regulatory Commission published the proposed 10 CFR Part 53 rule in March 2024 and is targeting a final rule in 2025 with implementation guidance through 2026. Part 53 is a technology inclusive, risk informed, performance based licensing framework designed to handle non light water reactors without forcing applicants through the prescriptive 10 CFR Part 50 and Part 52 paths originally written for large pressurized water reactors. Industry comments on the proposed rule have flagged residual prescriptiveness as still tighter than international and Department of Energy reference frameworks, but the rule is expected to materially shorten review cycles for designs that map cleanly to it.
The harder constraint is fuel. Most advanced reactor designs, including Natrium, Xe-100, Hermes, and Aurora, use high assay low enriched uranium, enriched between 5 and 19.75 percent uranium 235, against the 3 to 5 percent of conventional light water reactors. Russian state enterprise Rosatom historically supplied the only commercial volumes, and the Prohibiting Russian Uranium Imports Act signed in May 2024 closed that route subject to limited Department of Energy waivers through 2027. Centrus Energy began operating its American Centrifuge Plant in Piketon, Ohio in October 2023, producing the first 20 kilograms of HALEU and reaching the 90 kilogram contract milestone with the Department of Energy in early 2024. Capacity remains far below the 40 metric tons per year required by the late 2020s under base case advanced reactor schedules. Argus reads the HALEU constraint as the single most likely cause of slippage on Natrium, Xe-100, and Hermes commercial operation dates.
The capex curve and 45U economics #
Reported and modeled levelized cost for first of a kind SMR units sits in the 80 to 180 dollar per megawatt hour band, driven by financing structure and capital cost overrun assumptions. At 9,000 dollars per kilowatt overnight, a 30-year debt tenor, a 7 percent weighted average cost of capital, and a 90 percent capacity factor, the levelized cost lands near 110 to 130 dollars per megawatt hour before tax credits. Push the overnight cost to 12,000 dollars per kilowatt with a two-year overrun and the model produces 160 to 180 dollars. Pull it to 6,500 dollars on the second of a kind and stretch debt to 40 years under a Department of Energy loan guarantee, and the levelized cost falls into the 70 to 85 dollar band.
The Inflation Reduction Act Section 45U production tax credit is the largest structural change in nuclear project economics in a generation. It applies to existing zero emission nuclear units from 2024 through 2032, with a base of 0.3 cents per kilowatt hour rising to 1.5 cents with prevailing wage compliance, equivalent to 3 to 15 dollars per megawatt hour. For new build advanced reactors, Section 45Y carries the same 3 to 15 dollar structure for facilities placed in service after 2024 with a longer 10-year window. Both credits have transferability provisions that have created a functioning secondary market.
| Scenario | Overnight cost (USD per kW) | WACC | Tax credit treatment | LCOE band (USD per MWh) |
|---|---|---|---|---|
| FOAK pessimistic | 12,000 with 24 month overrun | 8.5 percent | 45Y at 5 USD per MWh effective | 150 to 180 |
| FOAK base | 9,500 to 11,000 on schedule | 7.0 percent | 45Y at 11 USD per MWh effective | 100 to 130 |
| SOAK with DOE loan | 7,500 to 9,000 | 5.5 percent blended | 45Y at 13 USD per MWh effective | 80 to 100 |
| NOAK target | 6,000 to 7,000 | 5.0 percent blended | 45Y monetized via transfer | 65 to 85 |
| Existing fleet uplift | Sunk | Owner cost of capital | 45U up to 15 USD per MWh | 30 to 50 cash cost |
International parallels and the export contest #
Outside the United States, the SMR contest is sharpening. Rolls-Royce SMR cleared United Kingdom Generic Design Assessment Step 2 in 2024 and was selected in February 2026 as preferred bidder under the Great British Nuclear competition, with target deployment in the early 2030s. The Rolls-Royce design is a 470 megawatt close-coupled pressurized water reactor positioned as a factory-built solution. Rosatom is marketing the RITM-200N, a 55 megawatt land-based adaptation of its icebreaker reactor, to Egypt, Bangladesh, and Uzbekistan, where Russian export finance moves faster than United States Export Import Bank facilities. The first onshore RITM-200N at Ust Kuyga in Yakutia is targeted for 2028.
United States and South Korea cooperation crystallized in 2024 around the Korea Hydro and Nuclear Power and Holtec memorandum of understanding for SMR-300 deployment in Korea and joint export pursuit. KHNP brings deep engineering, procurement, and construction track record from APR1400 builds in Korea and the United Arab Emirates Barakah project. The export contest matters because SMR economics improve materially with serial deployment, and the United States cannot reach a viable nth of a kind cost without combining domestic data center load, allied utility orders, and a third country export pipeline.
What to track and three scenarios for 2030 #
The base case Argus assigns roughly 50 percent probability has three to five SMR or advanced reactor units in late stage construction or near commercial operation by 2030, led by Natrium, the Dow Seadrift Xe-100 cluster, and the first Holtec SMR-300 unit at Palisades, with Kairos Hermes already operational in test mode and the first commercial Kairos plant under construction. Levelized cost on first of a kind units lands in the 100 to 130 dollar per megawatt hour band, hyperscaler off-take is fully subscribed for the first wave, and Part 53 is operational. The upside case at 25 percent probability adds two more designs to the operational fleet by 2032, sees HALEU supply scale to 30 to 40 metric tons per year by 2030, and produces a learning curve that pulls second of a kind capex below 8,000 dollars per kilowatt. The downside case at 25 percent probability has TerraPower or Holtec slip to 2033, HALEU constrained, the 45Y credit substantially modified after a 2026 election, and the data center off-take pipeline rotating to gas with carbon capture.
The actionable variables for 2026 are the Part 53 final rule text, the Centrus and Orano HALEU production runs, the NRC construction permit decisions on Xe-100 at Seadrift and Holtec SMR-300, the Three Mile Island Unit 1 license renewal docket, the Oklo Aurora resubmittal after the 2022 denial, and the FERC posture on hyperscaler co-location arrangements. Promethean recommends investors and corporate buyers underwrite SMR exposure by anchoring on 2030 to 2032 commercial operation dates rather than 2028 marketing dates, treating any 45Y monetization above 11 dollars per megawatt hour as upside, and reserving 15 to 20 percent contingency on first of a kind capex even after fixed price contract structures.
Sources #
- United States Nuclear Regulatory Commission Part 53 rulemaking
- United States Department of Energy Office of Nuclear Energy
- IAEA Advanced Reactors Information System
- IEA Nuclear Power and Secure Energy Transitions
- Nuclear Energy Institute resource library
- TerraPower Natrium project updates
- X-energy news and disclosures
- NuScale Power investor relations
- World Nuclear News
- Centrus Energy HALEU operations
Adjacent reading.
Johor's Data Center Boom: Singapore's Spillover and Malaysia's Grid Bet
Singapore's moratorium pushed roughly 1.6 GW of latent demand across the causeway, and Johor is now Southeast Asia's most concentrated hyperscaler buildout. The...
Read brief → Energy and transition economicsSmall modular reactors meet the hyperscaler load curve
Eighteen months after the Google Kairos and Amazon X-energy announcements, the SMR thesis has moved from PowerPoint to procurement. The binding constraints are ...
Read brief → AI compute and energyTexas, ERCOT, and the AI Siting Reset
Senate Bill 6, the ERCOT large-load study, and a 30 to 40 GW interconnection queue are forcing hyperscalers to rethink West Texas, behind-the-meter gas, and the...
Read brief →