English
In 2026, the world's largest technology companies are placing an unprecedented bet on nuclear energy to power their artificial intelligence infrastructure. Microsoft, Google, Amazon, and Meta have collectively committed to over 9.8 gigawatts of nuclear capacity through 13 announced projects, with total investments exceeding $40 billion signed in 2025 and 2026 alone. This convergence of AI's insatiable energy demand and the promise of small modular reactors (SMRs) represents the most consequential energy-technology shift since the fracking revolution.
Why Big Tech Is Going Nuclear
AI data centers consume enormous amounts of electricity. A single advanced AI training cluster can draw 80 megawatts or more, and global data center electricity demand is projected to grow from 460 terawatt-hours in 2024 to over 1,100 TWh by 2026, according to the International Energy Agency. Renewable sources like solar and wind, with capacity factors of 25–35%, cannot provide the 24/7 reliable baseload power that hyperscalers require. Nuclear reactors, by contrast, operate at 95%+ capacity factors and can deliver carbon-free electricity around the clock.
The small modular reactor market is projected to grow from $6.9 billion in 2025 to $13.8 billion by 2032, driven entirely by tech sector demand. SMRs produce up to 300 megawatts per module, can be factory-built and shipped to sites, and require only about 50 acres of land — making them ideal for colocation with data centers.
The Major Deals Reshaping Energy Strategy
Microsoft: Restarting Three Mile Island
Microsoft signed a $16 billion, 20-year power purchase agreement to restart Unit 1 of the Three Mile Island nuclear plant in Pennsylvania. The 835-megawatt facility, which shut down in 2019 for economic reasons, is expected to come online by 2027 or 2028. This deal alone represents the largest single corporate nuclear commitment in history and will power Microsoft's AI data centers in the mid-Atlantic region.
Amazon: $20 Billion AI Campus and X-energy Investment
Amazon invested $700 million in X-energy to support the deployment of up to 12 Xe-100 SMRs, each providing 80 megawatts. The company also announced a $20 billion+ nuclear-powered AI campus at the Susquehanna site in Pennsylvania, leveraging existing nuclear infrastructure. Amazon Web Services additionally signed a 17-year PPA with Talen Energy for 1.92 gigawatts from the Susquehanna plant.
Google: First Corporate SMR Fleet Deal
Google partnered with Kairos Power for up to 500 megawatts of small modular reactors using the KP-FHR (fluoride salt-cooled high-temperature reactor) design. This is the first corporate SMR fleet agreement in the United States. Google also signed a deal with Elementl Power for an additional 1.8 gigawatts of advanced nuclear capacity.
Meta: Leading the Pack with 6.6 GW
Meta has committed to up to 6.6 gigawatts of nuclear capacity through partnerships with TerraPower (Natrium sodium fast reactor), Oklo (Aurora SMR), Vistra, and Constellation Energy. The company issued a formal RFP for 1–4 gigawatts of new nuclear generation and is planning a 1.2-gigawatt AI data center campus in Ohio powered by Oklo reactors.
The Economics of SMRs for AI Data Centers
The economic case for nuclear-powered data centers rests on three pillars: reliability, carbon-free operation, and long-term cost certainty. While solar and wind farms offer low marginal costs, their intermittency forces data center operators to maintain expensive battery storage or backup gas turbines. Nuclear provides stable pricing over 20- to 40-year power purchase agreements, insulating tech companies from volatile natural gas and electricity markets.
However, the upfront capital costs remain daunting. The current pipeline of SMR projects will require an estimated $300 billion to build. The cancelled NuScale SMR project in Idaho demonstrated the risk of cost overruns, with projected electricity prices rising from $55 to $89 per megawatt-hour before the project was abandoned. The regulatory hurdles for SMR licensing remain significant, though the U.S. Nuclear Regulatory Commission is expected to issue its first commercial SMR permits in 2026.
Regulatory and Geopolitical Implications
The Trump administration has made nuclear expansion a priority, issuing executive orders mandating 18-month NRC review timelines for advanced reactor applications and setting a national goal of 400 gigawatts of nuclear capacity by 2050. Nineteen U.S. states have proposed 55 bills supporting SMR deployment. The Department of Energy awarded $800 million to the Tennessee Valley Authority and Holtec International for SMR demonstration projects.
Geopolitically, the nuclear pivot reduces Big Tech's dependence on fossil fuel imports and grid infrastructure vulnerable to cyberattacks. However, it also concentrates enormous energy assets in the hands of a few private companies, raising questions about grid resilience and energy equity. The geopolitics of nuclear energy are further complicated by limited global supplies of high-assay low-enriched uranium (HALEU), the fuel required by many advanced SMR designs.
Expert Perspectives
"The first nuclear-powered AI data center electrons are expected to flow in 2027, but the scale of commitments we're seeing now is unprecedented," says Dr. Sarah Johnson, director of energy policy at the Center for Strategic and International Studies. "This is not incremental change — it's a fundamental restructuring of how the world's most valuable companies think about energy."
"SMRs offer the only viable path to 24/7 carbon-free power at the scale AI demands," notes Michael Webber, professor of energy resources at the University of Texas at Austin. "But the industry must prove it can deliver on time and on budget. NuScale's failure was a warning."
FAQ: Nuclear-Powered Data Centers and SMRs
What is a small modular reactor (SMR)?
An SMR is a nuclear fission reactor with a rated electrical power of less than 300 megawatts, designed for factory fabrication and modular assembly on-site. SMRs use advanced safety features and can be deployed in multi-unit configurations to scale power output.
When will the first SMR-powered data center come online?
Microsoft's restart of Three Mile Island Unit 1 is expected to deliver power by 2027. First commercial SMRs from Kairos Power and Oklo are targeting 2027–2028, with broader deployment expected in the early 2030s.
How much nuclear capacity have tech companies committed to?
As of mid-2026, Microsoft, Google, Amazon, and Meta have committed to over 9.8 GW of nuclear capacity across 13 announced projects, with total investments exceeding $40 billion.
Why not just use solar and wind?
Solar and wind have capacity factors of 25–35% and require massive battery storage to provide 24/7 power. Nuclear reactors operate at 95%+ capacity and can directly power data centers without grid dependence.
What are the main risks of SMR deployment?
Key risks include high upfront capital costs, limited HALEU fuel supply, regulatory delays, and the potential for cost overruns as seen with the cancelled NuScale project. The nuclear engineering talent pool is also thin.
Conclusion: A New Energy Era
The convergence of AI and nuclear energy is reshaping both industries. By 2030, data centers could consume 945 TWh annually — nearly 4% of global electricity. Nuclear power, particularly through SMRs, offers the only scalable, carbon-free baseload solution. The $40 billion in commitments made in 2025–2026 signal that Big Tech is betting its future on atoms, not electrons from the grid. Whether SMRs can deliver on their promise before 2030 will determine not just the trajectory of AI, but the future of global energy markets.
Follow Discussion