Google’s Massive Nuclear Power Deal Shows How Much Energy AI Really Needs

The biggest constraint on artificial intelligence may not be chips. It may be electricity. Google has signed one of the largest corporate power agreements in U.S. grid history, securing 3,590 megawatts from Constellation Energy as demand from AI data centers pushes the power system into a new era.

Google is locking up gigawatts of electricity

The agreement includes a 20-year power purchase deal for 890 megawatts from Constellation-owned nuclear plants. Constellation plans to invest more than $4.3 billion to upgrade existing reactors in Illinois, Pennsylvania and New Jersey so they can produce additional electricity.

Google also signed a separate 15-year agreement for another 2,700 megawatts within the PJM Interconnection, the country's largest regional power market. That portion is not tied to one specific source, but it gives Constellation long-term revenue certainty while giving Google access to a huge block of power.

AI is changing the economics of the grid

Data centers have moved from being a niche industrial load to one of the most important sources of electricity growth in the United States. The Energy Information Administration expects U.S. power consumption to reach record highs in both 2026 and 2027, with AI and cryptocurrency data centers among the biggest drivers.

The pressure is especially visible in PJM, where capacity prices have increased sharply since 2024. Grid operators are now considering rules that would require large data-center customers to bring their own power or accept curtailment during periods of peak demand.

Nuclear power is getting a second life from AI

For years, nuclear power struggled to attract new investment in the United States because projects were expensive and electricity demand was relatively flat. AI has changed the calculation. Technology companies now need reliable, around-the-clock power at a scale that intermittent sources alone may struggle to provide.

Google is not alone. Microsoft has backed the restart of a reactor at Three Mile Island, and Google previously contracted for power tied to the restart of a nuclear plant in Iowa. The pattern suggests that data-center growth is becoming one of the most important commercial forces behind the nuclear industry's revival.

The AI boom is becoming an energy story

Every new generation of AI model requires more computing power, and more computing power means more electricity. That creates a direct link between the pace of AI development and the pace at which the energy industry can build generation, transmission, storage and cooling infrastructure.

This also means the AI competition will increasingly spill into energy markets. Companies that secure long-term power may gain an advantage over rivals that have the chips but cannot connect enough capacity to the grid.

What to watch next

The Google-Constellation deal is worth watching because it may become a template for other hyperscalers. If AI demand keeps rising, the companies building the world's largest models may also become some of the most influential players in nuclear energy, power markets and grid investment.