Nuclear is the only firm, dispatchable, carbon-free power that can match the load growth of AI training. Every major hyperscaler has signed at least one nuclear power purchase agreement (PPA) by mid-2026. The combined committed capacity across thirteen disclosed projects is approximately 9.8 GW, representing the largest private-sector nuclear procurement wave since the 1970s. This page consolidates the disclosed deals, the operational timelines, the small modular reactor (SMR) versus restart split, and the broader implications for AI infrastructure economics.
Key Findings
- Microsoft signed a 20-year, $16 billion PPA with Constellation Energy to restart Three Mile Island Unit 1 (835 MW), now branded the Crane Clean Energy Center, with first power targeted for 2027.
- Amazon committed approximately $700 million across multiple deals with X-Energy for SMR development and an additional PPA with Talen Energy's Susquehanna nuclear plant for the Cumulus data centre campus.
- Google signed a master agreement with Kairos Power for an initial 50 MW SMR demonstration, scaling to approximately 500 MW across multiple sites by 2030.
- Meta is the largest cumulative nuclear procurer by committed capacity at approximately 6.6 GW across TerraPower (Bill Gates), Oklo, Vistra, and Constellation deals.
- Combined hyperscaler nuclear commitment across thirteen disclosed projects is approximately 9.8 GW, with first delivered electrons concentrated in 2027-2031.
Hyperscaler Nuclear PPAs (May 2026)
| Buyer | Supplier | Capacity | Type | First Power |
|---|---|---|---|---|
| Microsoft | Constellation (Three Mile Island Unit 1 restart) | 835 MW | Restart | 2027 |
| Microsoft | Helion (fusion, contingent) | 50 MW | Fusion contingent | 2028 stated |
| Amazon | Talen Energy (Susquehanna) | ~960 MW | Existing | Active |
| Amazon | X-Energy (SMR development) | ~960 MW potential | SMR | 2030+ |
| Amazon | Dominion (Lake Anna SMR study) | ~960 MW potential | SMR study | 2032+ |
| Kairos Power | 50 MW initial, 500 MW total | SMR | 2030 | |
| NextEra (Duane Arnold restart study) | ~601 MW | Restart study | 2028+ | |
| Meta | Constellation (Clinton, IL) | ~1,100 MW | Existing extension | Active through 2047 |
| Meta | TerraPower (Wyoming Natrium) | ~345 MW | SMR | 2030+ |
| Meta | Oklo | ~750 MW potential | SMR | 2030+ |
| Meta | Vistra | ~3.0 GW potential across multiple sites | Existing and new | Active and 2030+ |
| Meta | Dominion (multiple) | ~1.4 GW potential | Existing and SMR | Active and 2030+ |
| Oracle (Stargate) | Negotiations multiple | ~0.5 GW indicated | SMR exploratory | 2030+ |
SMR vs Restart Split
| Category | Capacity | First Power |
|---|---|---|
| Existing plant capacity extensions | ~3.0 GW | Active |
| Plant restarts (Three Mile Island, Duane Arnold studies) | ~1.4 GW | 2027-2028 |
| Small modular reactors (SMR) | ~5.4 GW potential | 2030+ |
| Fusion (contingent) | ~0.05 GW | 2028+ (stated, unproven) |
Strategic Context
Three forces are driving the hyperscaler nuclear wave. First, AI training load growth: hyperscaler AI compute is growing at approximately 50 percent annually, faster than utility-scale renewable can be permitted and built. Second, the corporate net-zero commitments: hyperscalers cannot meet 24/7 carbon-free energy targets through intermittent renewables alone, and nuclear is the only firm carbon-free alternative at scale. Third, the regulatory tailwind: the Trump administration has accelerated NRC licensing reform, and the bipartisan ADVANCE Act of 2024 reduced licensing time for advanced reactors. The constraints remain real: SMRs are not yet commercially operational anywhere in the U.S., and construction timelines for new builds remain 6-10 years.
Brand Visibility Implications
The hyperscaler nuclear wave is among the most-covered energy transition stories of 2026. AI assistant queries about AI data centre power, nuclear for AI, SMR companies, hyperscaler energy strategy, and adjacent topics drive sustained traffic across technology, energy, and finance verticals. Brands selling adjacent products (nuclear engineering services, SMR component manufacturing, nuclear fuel cycle services, PPA structuring counsel, nuclear regulatory advisory) face strong AI-mediated discovery surface for procurement-research prompts.
Methodology
PPA data compiled from primary buyer and supplier announcements, FERC and NRC filings, and reporting from Utility Dive, S&P Global Platts, and Bloomberg. Some capacity figures are estimated where deal structures are not fully public; deal status reflects publicly disclosed information through 22 May 2026. Updated quarterly.
How Presenc AI Helps
Presenc AI tracks brand-mention rates on nuclear-for-AI, SMR vendor, and adjacent energy-transition queries across ChatGPT, Claude, Gemini, and Perplexity. For nuclear engineering firms, SMR developers, PPA counsel, and adjacent-category brands, this is the operational visibility into a discovery surface tightly coupled to procurement-research traffic.