Nuclear SMRs in 2026: NuScale Lessons, TerraPower Wyoming, and the AI Datacentre PPAs
Where small modular reactors sit in 2026 — NuScale cancellation lessons, TerraPower Wyoming, X-energy + Dow, Oklo, Rolls-Royce SMR, GE Hitachi BWRX-300, and the AI datacentre demand surge.
The small modular reactor (SMR) industry spent 2023 absorbing the worst headline it had ever received — NuScale’s cancellation of the Carbon Free Power Project in Idaho, the only fully NRC-licensed SMR design in the US, after the projected cost-per-megawatt-hour ballooned past what the buying utilities could justify. The cancellation, formally announced in November 2023, raised serious questions about whether the SMR thesis would ever survive contact with actual construction costs and utility procurement.
By 2026, the answer turns out to be more nuanced than the obituary writers expected. NuScale’s cancellation was a real warning about specific project economics rather than the death of the category. TerraPower has broken ground in Wyoming, X-energy has signed real contracts, Oklo is moving toward NRC approval, Rolls-Royce SMR is in active competition in the UK, GE Hitachi’s BWRX-300 is moving toward construction at Ontario Power Generation, and the AI datacentre boom has driven utility PPAs that have completely changed the demand picture for nuclear of every flavour.
This post walks through where SMRs actually sit in 2026, what the NuScale story taught the industry, and how the AI datacentre demand surge has rewritten the economics.
The NuScale cancellation, and what it actually meant#
The NuScale Carbon Free Power Project at the Idaho National Laboratory was supposed to be the first commercial SMR deployment in the US, with the Utah Associated Municipal Power Systems consortium as the off-taker. NuScale held the only NRC design certification for an SMR in the country (received in 2020) and the project had been in development since 2015. The November 2023 cancellation came after projected LCOE estimates rose from around 58 dollars per megawatt-hour in 2021 to roughly 89 dollars in mid-2023 — too expensive for the consortium to justify against the alternatives, particularly with the Inflation Reduction Act tax-credit changes that made renewables-plus-storage more competitive.
The honest reading of the NuScale cancellation is that the problem was project-specific (a first-of-a-kind build on an inflation-disrupted supply chain, with limited project-finance precedent) and design-specific (the six-module configuration the customer wanted required a specific power-island design that scaled poorly versus the smaller demonstration configurations). It was not a finding that SMRs as a category do not work. NuScale the company continues to operate, has pivoted toward international projects (with sites in Poland and Romania), and is positioning around the SMR-300 design as the commercial path forward.
The wider lesson the industry absorbed was that first-of-a-kind nuclear builds are expensive in unpredictable ways, that utility off-takers are unwilling to bear the cost-overrun risk alone, and that the SMR business model needs strategic anchor customers — sovereign governments, hyperscale corporations, or vertically-integrated utilities — that can absorb FOAK cost variance.
TerraPower and the Wyoming construction#
TerraPower, the Bill Gates-backed company founded in 2008, broke ground on its Natrium plant near Kemmerer, Wyoming in June 2024. The plant pairs a 345-megawatt sodium-cooled fast reactor with a molten-salt energy storage system, allowing the plant to run at constant reactor output while ramping the grid-facing power up and down to follow demand. The construction site, on the location of a retiring PacifiCorp coal plant, is a tangible signal that the SMR thesis is moving past the design-and-licensing phase into actual concrete and steel.
The project has hit real delays — primarily on the HALEU (high-assay low-enriched uranium) fuel supply, which depends on US enrichment capacity that is still being rebuilt after a long period of Russian fuel imports — but the underlying construction has continued, and the targeted operating date now sits in the early 2030s.
X-energy and the Dow Texas plant#
X-energy, the high-temperature gas-cooled reactor company, signed a 2023 agreement with Dow Chemical for a four-pack of Xe-100 reactors at Dow’s Seadrift, Texas petrochemical site, providing both electricity and process heat for the manufacturing complex. The 2024 follow-on filings advanced the licensing work, and the project is moving toward construction in the late 2020s. The Dow-X-energy deal is the canonical example of an SMR sold to a large industrial customer for process-heat applications that conventional gigawatt-scale nuclear does not address. Amazon Web Services led a billion-dollar funding round for X-energy in October 2024, positioning the company to serve the AWS-Talen and adjacent datacentre opportunities.
Oklo and the microreactor track#
Oklo, the Silicon Valley-backed microreactor company that went public via SPAC in May 2024, takes a different approach to the SMR thesis. The Aurora reactor is much smaller — 15 to 50 megawatts versus 300-plus for NuScale or BWRX-300 — and aims at remote-industrial and behind-the-meter datacentre applications where the alternative is diesel generation or grid extension. Oklo received DOE site approval at Idaho National Laboratory in 2019, had its initial NRC licence application rejected in 2022, and resubmitted in late 2025 after a redesign and a build-out of the licensing team. The 2024 Sam Altman-led SPAC raise gave the company the capital runway to push through the second-attempt licensing process. The first Aurora deployment is targeted for the late 2020s.
Rolls-Royce SMR and the UK competition#
Rolls-Royce SMR is the UK-government-backed entrant designed around a 470-megawatt pressurised water reactor — bigger than the NuScale 77-megawatt module but smaller than a conventional gigawatt plant. The company is part of the UK’s Great British Nuclear competition launched in 2023, which down-selected to a small group of finalists in 2024 and is expected to award contracts in 2025 and 2026 for an initial deployment of two to four SMRs in the UK. International deployments are in active discussion with Czech Republic, Sweden, and the Netherlands.
GE Hitachi BWRX-300 and Ontario Power Generation#
GE Hitachi’s BWRX-300, a 300-megawatt boiling water reactor design derived from the company’s larger ESBWR, has the most advanced active construction in North America. Ontario Power Generation broke ground on the Darlington New Nuclear Project site in 2024 with the first BWRX-300 unit targeted for commercial operation in 2029. The site is licensed for up to four units. The OPG project is the first western SMR build to clear all the major regulatory milestones and get to actual construction, and the early-2026 status is that the project is on schedule. Tennessee Valley Authority and Saskatchewan’s SaskPower have both indicated they will follow with their own BWRX-300 deployments.
The AI datacentre demand surge#
The most important single change in the nuclear story since 2023 is the AI datacentre demand surge. Hyperscalers building out frontier-AI training and inference capacity have run into the limits of where the existing grid can deliver multi-hundred-megawatt loads on short timelines, and the procurement teams have moved decisively toward direct nuclear PPAs as a solution.
The Constellation-Microsoft deal announced in September 2024 restarts Three Mile Island Unit 1, which had been shut in 2019, with all output going to Microsoft datacentres under a twenty-year PPA. The AWS-Talen deal announced in March 2024 sells a 960-megawatt portion of the Susquehanna nuclear plant’s output to a co-located AWS datacentre campus. The Google-Kairos deal announced in October 2024 commits Google to buy power from a fleet of Kairos SMRs targeted for operation in the late 2020s and early 2030s.
These deals have transformed the SMR demand picture. The buyers are large, the timelines are long, and the willingness to underwrite first-of-a-kind cost variance is qualitatively different from what utility off-takers like UAMPS could provide on the NuScale project. The hyperscaler PPAs are now the most credible commercial path for SMR companies to get FOAK projects funded and built.
What this means in 2026#
The realistic 2026 read on nuclear SMRs is that the category has survived its near-death moment, the build pipeline is genuinely materialising on the BWRX-300, TerraPower, and X-energy projects, and the hyperscaler PPA market has rewritten the commercial economics in ways that make the next round of FOAK projects more financeable than the NuScale-UAMPS project ever was. None of this means SMRs will deliver the grid-scale clean-firm power the 2030s need — that depends on construction execution that the industry has not yet demonstrated at scale — but the trajectory is decisively more constructive than it looked in late 2023.
Where pdpspectra fits#
Our data-engineering practice and cloud-infrastructure practice build the energy-data and grid-modelling platforms that nuclear developers, datacentre operators, and utilities need for PPA structuring, load-forecast modelling, and operational-data integration.
Related reading: the fusion energy 2026 post, the AI energy utilities post, and the energy grid optimization post.
SMRs survived 2023; the hyperscaler PPAs are what makes the rest of the decade interesting. Talk to our team about your energy-data program.