AI's Nuclear Hunger: How Data Centers Reshape Global Energy

The artificial intelligence revolution is driving an unprecedented surge in electricity demand, with global data center power consumption projected to reach 1,100 TWh by 2026 — equivalent to the entire electricity use of Germany. This step-change in baseload demand, which neither solar nor wind can reliably meet around the clock, has triggered a historic wave of corporate nuclear commitments. From Microsoft restarting Three Mile Island to Google and Amazon signing small modular reactor (SMR) deals, and Oracle planning gigawatt-scale nuclear-powered data centers, the AI-energy nexus is reshaping the global power order.

The Scale of AI's Power Appetite

According to the International Energy Agency (IEA), data centers consumed approximately 460 TWh in 2022, representing 2% of global electricity. By 2026, that figure is expected to more than double, driven overwhelmingly by power-intensive AI training and inference workloads. Goldman Sachs forecasts U.S. data center demand alone will rise 165% to 8% of total U.S. power by 2030. Northern Virginia data centers already consume 25% of PJM Interconnection capacity, straining local grids and triggering ratepayer revolts as electricity costs have risen 42% since 2019.

The global energy transition timeline is being rewritten as hyperscalers confront a fundamental reality: renewables alone cannot provide the 24/7 carbon-free baseload power that AI infrastructure requires. This has made nuclear energy — the only dispatchable zero-carbon source — the centerpiece of Big Tech's long-term energy strategy.

Corporate Nuclear Commitments Reach Record Levels

Microsoft and Three Mile Island's Historic Restart

In September 2024, Constellation Energy signed a 20-year power purchase agreement with Microsoft to restart Three Mile Island Unit 1 — the undamaged reactor at the site of America's most famous nuclear accident. Renamed the Crane Clean Energy Center, the 835 MW plant received a $1 billion federal loan from the Department of Energy in November 2025, with operations expected to resume in 2027. The deal, valued at approximately $16 billion over its term, marks the first time a major tech company will receive dedicated nuclear electrons for AI workloads.

Google, Amazon, and the SMR Wave

Google committed to 500 MW from Kairos Power's KP-FHR reactors, with first delivery targeted by 2030. Amazon invested $700 million in X-energy for up to 12 Xe-100 high-temperature gas-cooled reactors, and has also converted Pennsylvania's Susquehanna site into a nuclear-powered AI campus with over $20 billion in planned investment. Meta leads the pack with up to 6.6 GW of committed nuclear capacity across TerraPower's Natrium reactor, Oklo's Aurora design, and partnerships with Vistra and Constellation.

As of May 2026, every major hyperscaler — Microsoft, Google, Amazon, Meta, and Equinix — has signed nuclear power deals, totaling 13 announced projects and over 9.8 GW of committed capacity. The corporate nuclear energy procurement trend is accelerating, with long-term PPAs from creditworthy tech companies creating a bankable model for financing first-of-a-kind SMRs.

Oracle's Gigawatt-Scale Nuclear Vision

Oracle founder Larry Ellison announced plans for a data center requiring more than a gigawatt of electricity, to be powered by three small nuclear reactors. The company has already selected a location and secured building permits for the SMRs, positioning Oracle at the forefront of behind-the-meter nuclear generation.

2026: A Watershed Year for SMR Regulation

The Nuclear Regulatory Commission (NRC) is treating 2026 as a pivotal year for advanced reactor licensing. In March 2026, the NRC issued its first construction permit for a commercial advanced nuclear plant — TerraPower's Natrium reactor at Kemmerer, Wyoming. The 345 MW sodium-cooled fast reactor features a patented molten salt energy storage system that can boost output to 500 MW during peak demand. This milestone, achieved through streamlined 18-month review, signals a new regulatory era for SMR deployment.

The NRC is also developing modernized regulations under 10 CFR Part 53, shifting to a technology-inclusive framework that accommodates diverse reactor designs. The NRC advanced reactor licensing framework is expected to accelerate approvals for subsequent applications from NuScale, Kairos Power, X-energy, and others currently in the pipeline.

Globally, the European Commission released its SMR strategy in March 2026, including a €200 million guarantee to support first-of-a-kind deployments. Canada's CNSC licensed GE Hitachi's BWRX-300 for construction in April 2025, while China and Russia continue to operate the world's only commercially deployed SMRs.

Geopolitical Implications: The New Uranium Supply Chain

The nuclear renaissance creates new geopolitical dependencies on uranium supply chains. According to industry data, 70% of utility uranium demand beyond 2027 remains uncontracted — the highest level in 30 years. The global uranium market faces a structural deficit, with annual consumption of ~180 million pounds exceeding primary production of ~130 million pounds by roughly 28%. This gap is projected to widen through 2045 as nuclear power expands.

Kazakhstan now supplies 43% of global uranium output, creating concentration risks. Spot uranium prices have surged from $29/lb in early 2021 to peaks above $100/lb in 2024. New mines take 10–15 years to develop, and current prices remain below the $120–150/lb needed to incentivize new production. The uranium supply chain geopolitics are becoming a strategic concern for energy security.

Expert Perspectives

This is the year the AI-energy nexus becomes a defining strategic reality, said Victoria Gonzalez, energy analyst and author of this report. The convergence of record corporate commitments, pending SMR licensing decisions, and accelerating AI power demand means 2026 will be remembered as the moment nuclear energy re-entered the mainstream — not as a relic of the past, but as the foundation of the AI future.

Energy Secretary Chris Wright noted that the Department of Energy aims to use loan programs to support nuclear expansion amid rising electricity demand from AI data centers. The DOE's Advanced Reactor Demonstration Program has been instrumental in advancing TerraPower's Natrium project toward commercial operation by 2030.

Frequently Asked Questions

How much electricity will AI data centers consume in 2026?

AI data centers are projected to consume approximately 1,100 TWh globally in 2026, more than double the 460 TWh consumed in 2022, according to IEA forecasts. This represents about 3% of total global electricity use.

Why can't renewables meet AI data center power demand?

AI data centers require 24/7 baseload power that is reliable and carbon-free. Solar and wind are intermittent, generating power only when the sun shines or wind blows. Nuclear provides continuous, dispatchable electricity that matches the always-on nature of AI infrastructure.

What is a small modular reactor (SMR)?

An SMR is a nuclear fission reactor with electrical output under 300 MWe, designed for factory fabrication and modular construction. SMRs offer passive safety features, lower upfront costs, and scalability through multi-unit configurations, making them attractive for data center power.

When will the first commercial SMRs be operational in the U.S.?

The NRC issued the first commercial advanced reactor construction permit for TerraPower's Natrium in March 2026, with completion targeted by 2030. Other designs from NuScale, Kairos Power, and X-energy are in various stages of licensing, with commercial operations expected in the early 2030s.

What are the risks of the uranium supply shortage?

With 70% of post-2027 demand uncontracted and primary production meeting only 72% of current consumption, the uranium market faces a structural deficit. This could lead to price spikes, supply constraints, and increased geopolitical dependence on Kazakhstan, which supplies 43% of global uranium.

Conclusion: A New Energy Paradigm

The AI-driven nuclear renaissance represents a fundamental shift in global energy strategy. As data center power demands continue to accelerate, the marriage of artificial intelligence and atomic energy is no longer a theoretical possibility — it is an operational reality. The decisions made in 2026 regarding SMR licensing, corporate offtake agreements, and uranium supply chain investments will shape the energy landscape for decades to come. The message from Big Tech is clear: the future of AI runs on nuclear power.

Sources

• International Energy Agency, Electricity 2026 Report
• U.S. Nuclear Regulatory Commission, News Release No. 26-028 (March 2026)
• TerraPower, NRC Approves Natrium Reactor Construction Permit (March 4, 2026)
• Bloomberg, Microsoft AI Deal Secures Three Mile Island Nuclear Restart (May 2026)
• CNBC, Trump Administration Backs Three Mile Island Restart with $1B Loan (November 2025)
• World Nuclear Association, Uranium Market Data
• Goldman Sachs, Data Center Power Demand Forecast