Big tech nuclear power purchase agreements (PPAs) are long-term contracts through which Microsoft, Google, Amazon, and Meta are collectively securing more than 10 gigawatts of nuclear-generated electricity to power AI data centers. These deals are transforming the U.S. nuclear sector from a financially distressed industry into a strategically critical infrastructure supplier for the digital economy.
Key Takeaways
- Microsoft, Google, Amazon, and Meta have signed over 10 GW of nuclear PPAs in the past year, with Meta alone contracting up to 6.6 GW of nuclear capacity.
- The nuclear energy data center PPA market was valued at $6.2 billion in 2025 and is projected to reach $47.8 billion by 2034.
- The global small modular reactor pipeline reached 47 GW at the end of Q1 2026, with more than half located in the U.S., driven primarily by tech company demand.
The Scale of the Commitment
The numbers are difficult to overstate. According to research compiled by Trellis and SoftwareSeni, the four largest U.S. technology companies have collectively contracted nuclear capacity at a scale that would have seemed implausible five years ago, when nuclear operators were shuttering plants and writing off stranded assets. Meta leads the group with up to 6.6 GW of contracted nuclear capacity. Amazon follows at 1.9 GW. Microsoft’s landmark agreement covers 835 MW tied to the restart of Three Mile Island. Google has contracted 500 MW through its partnership with Kairos Power. Together, these four companies signed nearly half of all clean energy deals in 2025, a concentration of purchasing power that has no modern precedent in the energy sector.
The financial consequences have been immediate and visible. Constellation Energy, the primary operator benefiting from these agreements, saw its stock surge 475% over three years. The nuclear energy data center PPA market, valued at $6.2 billion in 2025, is projected to reach $47.8 billion by 2034, according to market analysis cited by Investing.com. Uranium producers and nuclear engineering firms are experiencing their strongest growth cycle in decades.
Four Companies, Four Distinct Strategies
What makes this moment analytically interesting is that Microsoft, Google, Amazon, and Meta are not pursuing identical approaches. Each company’s nuclear strategy reflects a different tolerance for technology risk, timeline pressure, and capital deployment.
| Company | Contracted Capacity | Technology Approach | Key Partner / Asset | Target Timeline |
|---|---|---|---|---|
| Microsoft | 835 MW | Conventional reactor restart | Three Mile Island (Constellation) | 2028 |
| Amazon | 1.9 GW | Small modular reactors (SMRs) | X-Energy; Cascade facility (up to 12 SMRs) | Early 2030s |
| 500 MW | Experimental molten salt SMRs | Kairos Power | 2030–2035 | |
| Meta | Up to 6.6 GW | Mixed / broad nuclear procurement | Multiple operators | 2030s |
Microsoft’s approach prioritizes speed. By committing $16 billion to restart Three Mile Island by 2028, the company is betting on proven reactor technology and an existing licensed site rather than waiting for next-generation designs to clear regulatory hurdles. It is the fastest credible path to large-scale firm nuclear power on the grid.
Amazon is betting on scale and modularity. Its $500 million investment in X-Energy and plans for up to 12 small modular reactors at its Cascade facility represent a longer-horizon wager that SMR manufacturing economics will improve sufficiently to make fleet deployment cost-competitive. The early 2030s target acknowledges that SMR construction timelines are not yet proven at commercial scale.
Google’s partnership with Kairos Power is the most technically speculative of the four. Molten salt reactor technology has never been commercially deployed. The 2030–2035 timeline is contingent on Kairos successfully demonstrating the technology at scale, obtaining NRC licensing, and executing construction without material delays. Google appears to be treating this as a strategic technology hedge rather than a near-term capacity solution.
Meta’s strategy is the broadest. Contracting up to 6.6 GW across multiple operators positions the company to capture nuclear capacity wherever it becomes available, rather than concentrating risk in a single technology or operator relationship.
Why it matters for builders: Nuclear PPAs are now setting the baseline for firm power procurement in AI infrastructure planning. Any data center project targeting 100 MW or more at a single site should model nuclear availability as a primary supply scenario, not an alternative.
Why Renewables Alone Are Not Sufficient
The underlying driver of these agreements is not ideological preference for nuclear power. It is the physical reality of AI workload characteristics. Large language model training and inference are continuous, high-density electrical loads. Solar and wind generation are intermittent by nature. Battery storage at the scale required to firm gigawatt-class renewable portfolios remains economically and logistically prohibitive for most data center operators today.
Nuclear generation provides firm, dispatchable, carbon-free baseload power at high capacity factors — typically above 90%. For a hyperscaler operating a 500 MW campus that cannot tolerate generation gaps, a nuclear PPA is structurally more compatible with operational requirements than an equivalent renewable portfolio requiring storage backup. The global SMR pipeline reaching 47 GW at the end of Q1 2026, with more than half in the U.S., reflects this demand signal propagating through the supply chain.
Risks That Cannot Be Discounted
The optimism embedded in these announcements carries real technical and regulatory risk. Google’s molten salt reactor technology has no commercial deployment precedent. All four companies’ timelines assume no material regulatory delays, no significant construction cost overruns, and no supply chain disruptions in specialized nuclear components — assumptions that the history of nuclear construction in Western markets does not fully support.
The Three Mile Island restart is the most de-risked of the four strategies, but it still requires NRC approval, workforce reconstitution, and component refurbishment on a compressed schedule. SMR projects face the additional challenge that no Western SMR design has yet completed a full commercial construction cycle, meaning cost and schedule estimates carry wide uncertainty bands.
Why this matters for builders, developers, and investors
For anyone planning, financing, or building energy infrastructure for AI data centers, these agreements establish a new procurement benchmark. Nuclear PPAs are now being signed at 835 MW to 6.6 GW scale, with 15- to 20-year contract structures that lock in capacity well ahead of commercial operation. Developers evaluating site selection, grid interconnection, and power supply strategy for large campuses need to assess nuclear availability as a first-order variable, not an afterthought. The $47.8 billion projected market size by 2034 signals that this is a durable structural shift, not a short-term procurement trend.
Tools & Resources
- Seeking Alpha — Track nuclear operator financials, Constellation Energy coverage, and uranium sector equity data relevant to PPA market developments.
- The Motley Fool — In-depth investment research on the energy transition, SMR developers, and the infrastructure companies supplying the AI power buildout.
FAQ
What is a nuclear power purchase agreement?
A nuclear power purchase agreement is a long-term contract between a nuclear plant operator and an electricity buyer — in this case, a technology company — that guarantees the buyer a fixed volume of nuclear-generated electricity at a negotiated price over a defined period, typically 15 to 25 years. These agreements provide revenue certainty for operators and firm power supply for buyers.
How much nuclear capacity have big tech companies contracted?
Microsoft, Google, Amazon, and Meta have collectively contracted over 10 GW of nuclear capacity. Meta leads with up to 6.6 GW, Amazon has contracted 1.9 GW, Microsoft’s Three Mile Island agreement covers 835 MW, and Google has contracted 500 MW through Kairos Power.
When will Microsoft’s Three Mile Island restart deliver power?
Microsoft’s agreement with Constellation Energy targets a Three Mile Island restart by 2028. The project involves a $16 billion commitment and requires NRC regulatory approval, workforce reconstitution, and component refurbishment before the plant can return to commercial operation.
What are small modular reactors and why are tech companies interested in them?
Small modular reactors are nuclear reactors with an electrical output typically below 300 MW, designed for factory fabrication and modular deployment. Tech companies are interested because SMRs can potentially be sited closer to data center campuses, deployed in fleets to match load growth, and built on shorter construction schedules than conventional large reactors — though no Western SMR design has yet completed a full commercial construction cycle.
What is the projected size of the nuclear data center PPA market?
The nuclear energy data center PPA market was valued at $6.2 billion in 2025 and is projected to reach $47.8 billion by 2034, reflecting the sustained demand from hyperscalers and the long contract durations typical of nuclear procurement agreements.
What are the main risks in big tech nuclear power agreements?
The primary risks include regulatory delays in NRC licensing, construction cost overruns, technology immaturity for unproven designs such as molten salt reactors, and supply chain constraints in specialized nuclear components. All announced timelines assume smooth regulatory and construction execution, which the historical record of Western nuclear projects does not guarantee.
Conclusion
Big tech nuclear power purchase agreements represent a structural realignment of the U.S. energy market, not a temporary procurement trend. Microsoft, Google, Amazon, and Meta have collectively committed to over 10 GW of nuclear capacity through a range of strategies — from proven reactor restarts to experimental molten salt technology — because the physics of AI workloads demand firm, dispatchable, carbon-free power at a scale that intermittent renewables cannot reliably deliver alone. The financial consequences are already visible in operator valuations and uranium markets. The risks are real: SMR timelines are unproven, regulatory processes are slow, and construction cost histories in nuclear are rarely favorable. But the demand signal is now large enough, and the contract structures long enough, that the nuclear industry has a credible commercial foundation it has not had in decades. For anyone building or financing the physical infrastructure of the AI economy, nuclear power has moved from a marginal consideration to a central planning variable.
Sources
- Trellis — Overview of Amazon, Google, Meta, and Microsoft nuclear procurement strategies and capacity figures.
- SoftwareSeni — Detailed breakdown of individual company nuclear investment strategies and technology partnerships.
- Investing.com — Nuclear energy market valuation data, Constellation Energy stock performance, and PPA market projections through 2034.