Nuclear Energy Is Being Rehabilitated by AI Power Demand — and This Time the Commercial Pull Is More Durable Than Policy Support Alone

AI Data Center Power Demand Is the Catalyst That Nuclear Energy's Proponents Did Not Forecast but Cannot Afford to Miss

The commercial logic connecting AI data center power demand to nuclear energy is straightforward and has compressed the timeline for nuclear investment decisions that would normally require decades of policy, financing, and regulatory work. Data center operators need power that is reliable — guaranteed to be available 24 hours a day, 365 days a year, without the weather-dependent variability that makes wind and solar generation inadequate as the sole power source for computational workloads that cannot be curtailed without service disruption. They need power that is carbon-free — because the hyperscalers operating the majority of US and European AI infrastructure have corporate sustainability commitments that make purchasing carbon-intensive power for AI workloads incompatible with their stated climate goals. And they need power in quantities large enough to justify long-term power purchase agreements rather than spot market purchases, because the capital intensity of data center infrastructure requires long-term revenue certainty that spot market power pricing cannot provide. Nuclear energy satisfies all three requirements simultaneously — reliable, carbon-free, and scalable to the quantities that hyperscale data center campuses require — in a way that no other generation technology currently matches. Wind and solar satisfy the carbon requirement but not the reliability requirement without storage at a scale that is not currently commercially available. Natural gas satisfies the reliability requirement but not the carbon requirement. Nuclear satisfies both, which is why it has moved from the periphery to the centre of AI infrastructure power strategy in the US and Europe simultaneously.

Constellation Energy's Three Mile Island Unit 1 restart is the most commercially significant nuclear energy development in the US since the Energy Policy Act of 2005 created the production tax credit that enabled the last wave of US nuclear capacity additions. The Microsoft power purchase agreement that justified the restart — structured to provide Microsoft with a long-term supply of carbon-free baseload power for its data center operations in the PJM Interconnection territory — has become the template that other data center operators and nuclear plant owners are actively replicating. Vistra Energy's Comanche Peak plant in Texas and Dominion Energy's North Anna nuclear facilities are both in discussions with hyperscalers about similar arrangements, and the pattern of nuclear-to-data-center power purchase agreements is creating a financial basis for nuclear plant life extension investments that had been commercially marginal before the AI power demand surge changed the power market economics for always-on carbon-free generation.

European Nuclear Policy Is Reversing — and the Commercial Consequences Are Material

The European nuclear policy reversal is proceeding at different speeds across EU member states but is now sufficiently broad to constitute a genuine continental shift rather than a country-specific policy reorientation. Germany's nuclear phase-out — completed in April 2023 when the country's last three commercial reactors were shut down — is now the policy that stands out as anomalous in a European context where France, the UK, the Netherlands, Poland, Sweden, Finland, and the Czech Republic are all in various stages of nuclear capacity expansion or life extension. The EU taxonomy's formal inclusion of nuclear power as a sustainable energy source, which unlocked European green bond financing for nuclear projects, is the most consequential policy change because it removes the financing cost disadvantage that had made new European nuclear project economics difficult to justify against the rapidly declining cost of renewable alternatives. French state utility EDF's ambitious EPR2 programme, the UK's commitment to six new nuclear reactors including Hinkley Point C and Sizewell C, and Poland's agreement with Westinghouse for AP1000 reactor construction are all expressions of the same underlying commercial logic: that reliable, carbon-free baseload power is worth a premium that the AI economy and corporate carbon commitments together make sustainable.

The small modular reactor market represents the most commercially dynamic segment of nuclear energy's rehabilitation, because SMRs offer a capital cost profile and construction timeline that is materially more compatible with private financing and utility balance sheet constraints than the gigawatt-scale EPR and AP1000 reactor designs that dominate European expansion plans. NuScale's VOYGR received NRC design certification in 2022, establishing the US regulatory precedent for SMR deployment. Rolls-Royce's SMR programme in the UK, TerraPower's Natrium reactor development in the US with Wyoming state support, and X-energy's TRISO fuel pebble bed reactor design licensed to utilities in both the US and Canada are all advancing on development timelines that anticipate first commercial operation in the 2030-2033 range. For the AI infrastructure operators who need power supply certainty over the same timeframe, SMR commitments that align with their data center buildout schedules are increasingly appearing in long-range power procurement planning — even though the actual power delivery date is several years away — because securing the power supply commitment now is commercially valuable even at the cost of a premium over near-term alternatives.