Nuclear Renaissance: Why 10 Countries Are Betting on Small Modular Reactors
Nuclear power spent the decade following Fukushima in managed retreat. Germany shut down its last reactor in 2023. Belgium extended retirements repeatedly before ultimately closing. In the United States, merchant nuclear plants in deregulated electricity markets closed despite producing low-carbon electricity at competitive cost, because capacity markets did not value their 24/7 reliability premium and their capital structures could not absorb the earnings volatility of spot power prices. The narrative arc seemed settled: existing nuclear as a stranded legacy asset, new nuclear as economically unviable without French-style state subsidies that democratic governments could not maintain across electoral cycles. That narrative is being revised with unusual speed in 2025, and the revision is driven not by ideology but by three converging pressures that have changed the economic calculus for nuclear power: data centre electricity demand, grid reliability requirements at high renewable penetration, and a genuine step-change in reactor design — the small modular reactor — that addresses nuclear's historical weakness of gigawatt-scale capital commitment before a single kilowatt-hour is generated.
What Makes SMRs Structurally Different from Conventional Nuclear
Conventional nuclear power plants — the GW-scale pressurised water reactors that provide 10% of global electricity — have three structural problems that SMRs are designed to solve. First, capital scale: a single AP1000 reactor costs USD 6–10 billion and takes 10–15 years from groundbreaking to first power, creating a capital commitment that overwhelms the balance sheets of private utilities and requires either government ownership or contract-for-difference guarantees that are politically fragile over 15-year development horizons. An SMR — defined as a reactor below 300 MWe — costs USD 1–3 billion per unit, a commitment that private capital can underwrite through project finance structures with revenue certainty from corporate power purchase agreements rather than government price guarantees. Second, construction risk: large nuclear plants are built on-site using custom-fabricated components that accumulate cost overruns from weather, labour, regulatory interpretation changes, and supply chain delays — Vogtle Units 3 and 4 in Georgia, the most recent US nuclear construction, came in at USD 36 billion against an original estimate of USD 14 billion. SMRs are designed for factory fabrication and modular site assembly — the reactor pressure vessel, steam generators, and primary coolant loop components are manufactured in controlled industrial settings and shipped to site, dramatically reducing field construction scope and the rework cost that drives large nuclear overruns. Third, siting flexibility: GW-scale plants require access to large water bodies for cooling and are incompatible with most industrial and data centre applications. SMRs with passive cooling systems can be sited at industrial facilities, mine sites, remote communities, and data centre campuses — expanding the addressable market from utility-scale grid supply to distributed industrial power.
The 10 Countries Building Real SMR Programmes
The United States leads on commercial deployment momentum. NuScale Power's VOYGR design — the first SMR to receive NRC design certification, in 2022 — was the global reference until its anchor project (Carbon Free Power Project, Utah) was cancelled in November 2023 due to subscriber withdrawal. The cancellation was widely misread as an SMR death knell; it was actually a specific project finance failure in a specific electricity market context. The underlying NRC-certified design remains the global regulatory gold standard. TerraPower's Natrium (sodium fast reactor, 345 MW), backed by Bill Gates and Wyoming state investment, is under construction at the Kemmerer site targeting 2030 operation. X-energy's Xe-100 (pebble bed HTGR, 80 MW per module) has DOE Advanced Reactor Demonstration Programme funding and a deployment agreement with Dow Chemical for industrial process heat — the first commercial nuclear-to-industrial heat supply arrangement in the United States. Canada has the most advanced government SMR action plan — four reactor designs (X-energy, Terrestrial Energy IMSR, Westinghouse eVinci, Ultra Safe Nuclear) have received Phase 1 vendor design review completion from the Canadian Nuclear Safety Commission, and Ontario Power Generation has selected GE-Hitachi's BWRX-300 for deployment at Darlington Nuclear Generating Station, targeting 2029 grid connection. The UK has committed GBP 2.5 billion to small modular reactor development through Great British Nuclear, with Rolls-Royce SMR (470 MW, a consortium including BNF Resources) as the primary domestic contender alongside Holtec's SMR-300. Poland, which has no operating nuclear capacity and is completing its coal exit, has selected Westinghouse AP300 (a scaled-down AP1000 derivative) for its first nuclear programme — six units at two sites targeting 2033 first power. South Korea's SMART reactor (100 MW, pressurised water design) achieved design approval from the Korean Institute of Nuclear Safety in 2012 and has export agreements with Saudi Arabia under evaluation. Romania, Czech Republic, and Sweden all have active SMR procurement or evaluation processes underway, driven by the EU taxonomy's inclusion of nuclear as a sustainable finance category — a critical capital markets enabler that brought European institutional investors back to nuclear project finance in 2023–2024.
The Data Centre Catalyst That Changed the Timeline
The proximate cause of accelerated SMR commercial timelines is electricity demand from AI data centres, and the speed at which major technology companies have moved from policy advocacy to direct investment is without precedent in energy project development. Microsoft's 20-year power purchase agreement with Constellation Energy for the restart of Three Mile Island Unit 1 (835 MW, operational 2028) was the announcement that signalled the seriousness of tech company nuclear demand. Google's agreement with Kairos Power for 500 MW of SMR capacity (HTGR design, multiple sites targeting 2030–2035) was the first direct SMR corporate PPA. Amazon's investment in X-energy and its announcement of SMR agreements with Energy Northwest and Dominion Energy collectively represent 5 GW of SMR demand commitments from a single buyer — demand certainty at a scale that transforms project finance viability for every SMR developer in the pipeline. These commitments matter not only for the specific projects they fund but for the demonstration effect on capital markets: when Microsoft, Google, and Amazon sign 20-year power purchase agreements for nuclear output, institutional investors can model nuclear project cash flows against investment-grade counterparty credit rather than merchant electricity price risk, reducing the cost of capital by 200–400 basis points and making projects viable that were previously unfinanceable.
The Risks Honest Analysis Must Include
SMR optimism requires three honest constraints. First, no SMR design outside of China's HTR-PM and Russia's RITM-200 (both state-owned, non-exportable technology) has operated commercially. The BWRX-300, AP300, and Natrium designs are extensions of proven reactor physics but have not demonstrated the factory fabrication cost reductions that make their economic projections realistic — NuScale's cancellation was partly driven by projected overnight costs rising from USD 4,200/kW to USD 9,000/kW as engineering progressed, a pattern consistent with every previous nuclear "learning curve" projection that proved optimistic. Second, nuclear waste disposal remains politically unresolved in every country on the SMR deployment list except Finland, which has built the world's first geological repository at Onkalo. SMR deployment at scale generates the same spent fuel volumes per MWh as conventional nuclear, and the absence of national waste disposal solutions will eventually constrain deployment permits in jurisdictions without geological storage. Third, the regulatory timelines for novel designs — sodium fast reactors, molten salt, pebble bed — are genuinely uncertain because no regulatory body has previously licensed these architectures in commercial service. The 2030–2035 deployment targets for advanced designs assume regulatory review timelines that have never been achieved for first-of-a-kind designs.
The Decade Ahead
The SMR renaissance is real but the timeline is compressed by optimism. The realistic scenario is 5–10 GW of SMR capacity operational globally by 2035, concentrated in Canada, the United States, and the UK, with a further 20–30 GW under construction targeting 2040 completion. This is commercially significant — it represents a USD 150–250 billion market for reactor vendors, fuel suppliers, and balance-of-plant contractors over 15 years — but it is not the grid transformation story that data centre demand proponents are projecting on 2030 timelines. The technologies that will power AI data centres through 2030 are the ones already being built: offshore wind, utility solar, and grid-scale batteries. SMRs are the power source that will matter for the 2035–2050 period, and the investment decisions being made now — in design certification, supply chain development, workforce training, and waste policy — will determine whether that potential is realised or deferred once again.