Major technology companies including Microsoft and Meta have locked in 10 gigawatts of nuclear power through long-term agreements to fuel their AI data centers, yet a comprehensive global analysis reveals that nuclear expansion will grow only incrementally over the next five years, not dramatically reshape the energy landscape. While these corporate commitments signal serious intent to secure clean baseload electricity for compute-intensive workloads, the broader nuclear pipeline faces persistent delays, cost overruns, and cancellations that will likely keep nuclear's contribution to global power growth modest compared to wind and solar.

Where Is Big Tech's Nuclear Power Coming From?

The technology sector's nuclear strategy is concentrated in the United States, where all 10 gigawatts of data center-linked nuclear capacity currently under agreement is located. Three states dominate this emerging market. Pennsylvania leads with 5.5 gigawatts, followed by Ohio with 2.4 gigawatts, and Illinois with 1.1 gigawatt. These aren't hypothetical plans; they represent binding power purchase agreements (PPAs) between tech companies and nuclear operators.

Microsoft's deal with Constellation Energy stands out as the most visible commitment. The company signed a 20-year agreement to help reopen a retired reactor at Three Mile Island in Pennsylvania, which will be renamed the Crane Clean Energy Center with a restart targeted for the second half of 2027. The facility is designed in part to power Microsoft's data centers and AI workloads, though transmission constraints mean full output may not be available until after 2030. Meta has pursued a dual-track approach, securing a 20-year PPA with Constellation for the Clinton nuclear plant in central Illinois to support continued operation after state support programs expire in mid-2027, while also partnering with Oklo to develop a nuclear campus in Pike County, Ohio, intended to directly support large-scale AI infrastructure like its Prometheus supercluster computing system.

Why Isn't Nuclear Scaling Faster to Meet AI Demand?

Despite corporate enthusiasm, the global nuclear outlook reveals structural constraints that will keep nuclear's growth incremental rather than transformative through 2030. The Global Nuclear Power Tracker, which monitors 1,825 nuclear units across 62 countries, shows that while 91 gigawatts of capacity has already broken ground worldwide, only about 60 percent (55 gigawatts) has a stated in-service date of 2030 or earlier. Even projects already under construction face significant execution risk.

Small modular reactors (SMRs), which tech companies and policymakers often cite as a flexible solution for distributed data center power needs, remain at early stages with limited operating experience, making meaningful scale-up before the 2030s unlikely. This reality undercuts the narrative that SMRs will rapidly solve AI's power challenges. Instead, the near-term nuclear pipeline is dominated by large conventional reactors, many of which are experiencing delays. China's Xudapu plant, at 5.1 gigawatts, is the largest single project expected this decade, while Türkiye's Akkuyu and Egypt's El Dabaa projects, each planned at 4.8 gigawatts and both backed by Russian financing, represent each nation's inaugural nuclear deployment.

The historical record illustrates why caution is warranted. Globally, more nuclear capacity has been cancelled (562 gigawatts) than has ever come online (544 gigawatts). China and the United States, projected to become the top two countries by operating nuclear capacity in the coming years, are also the world's leading sources of cancelled nuclear capacity, with 381 gigawatts of nuclear power cancelled across these two geographies alone.

How to Assess Nuclear's Role in AI Infrastructure Planning

• Realistic Timeline Expectations: Projects already under construction represent the most plausible near-term additions by 2030; pre-construction and announced projects with 2030 targets may not materialize until after that date due to nuclear's historically long and uncertain development cycles.
• Comparison to Renewable Alternatives: Wind and solar continue to scale rapidly with shorter development timelines and decreasing costs, making them the primary drivers of near-term power sector decarbonization, while nuclear remains a slower-moving option despite corporate interest.
• Geographic Concentration Risk: Tech companies' nuclear PPAs are heavily concentrated in three U.S. states; diversification across regions and fuel sources reduces exposure to transmission constraints and regulatory delays that could delay full capacity utilization.
• Lifetime Extension Strategy: Existing nuclear plants may provide more reliable near-term capacity retention than new builds; lifetime extensions are becoming increasingly common in geographies with established nuclear infrastructure.

If all or most of the current pipeline to 2030 materializes, nuclear power could expand its role in the world's electricity mix by the end of the decade. However, the Global Energy Monitor's analysis suggests that realized outcomes are likely to fall short of current 2030 targets. Nuclear construction timelines remain long and highly uncertain, especially compared with those of wind or solar projects. Globally, nuclear capacity could grow by about 12 percent in the next five years if most of today's project pipeline materializes, with 55.3 gigawatts potentially coming online by the end of 2030 from reactors that have already broken ground. After accounting for roughly 5 gigawatts of planned retirements, net additions by the end of this decade could reach 12 percent growth relative to today's 408 gigawatt operating fleet.

This represents a practical upper bound for near-term additions. An average net growth of 10 gigawatts per year would match 2016 for the highest net annual growth within the last three decades, marking a notable expansion for a mature technology. Yet risks and constraints remain defining features of the nuclear outlook. Project cancellations and delays are not isolated incidents but a persistent pattern across regions and decades, suggesting that tech companies betting on nuclear for AI infrastructure should maintain contingency plans and diversified energy portfolios.