Big Tech Turns to Nuclear Energy as AI Power Demands Surge

The rapid expansion of artificial intelligence has triggered a profound shift in global energy markets, as soaring AI electricity demand forces major cloud computing providers to secure nuclear power. Within a single week, two of the world's top three cloud corporations finalised 20-year nuclear electricity agreements. Amazon contracted 690 megawatts from Constellation Energy's Calvert Cliffs facility in Maryland, which includes 190 megawatts of extra capacity anticipated between 2030 and 2032. Meanwhile, Google completed a deal expected to inject 890 megawatts of additional nuclear power into the United States grid.

Rather than constructing new reactors from scratch, tech firms are funding output expansions at operational facilities. Constellation is allocating more than $4.3 billion to enhance 11 reactors across Illinois, Pennsylvania, and New Jersey. Elsewhere, an undamaged unit at Three Mile Island is returning to service under a two-decade agreement with Microsoft. To support such reactivations, the US Department of Energy is providing up to $4.4 billion in loans to help revive three decommissioned plants and restore approximately 2.25 gigawatts of power, including Michigan's Palisades site scheduled to resume output by March 2027.

Rising Costs for Consumers

Decades after safety disasters at Three Mile Island, Chornobyl, and Fukushima caused nuclear power to fall out of favour, the technology is experiencing a resurgence. Efficiency advances—such as more powerful chips or models like DeepSeek that cut training costs—have failed to reduce overall consumption. Instead, consistent with the economic principle observed by William Jevons in 1865, efficiency gains have expanded compute workloads. The International Energy Agency projects global data centre power use will more than double to roughly 945 terawatt-hours by 2030.

AI power contracts are locking in long-term supply while passing billions in grid costs onto households.

Because small modular reactors remain years away from commercial scale, tech firms are securing existing capacity within regional grids like PJM, which covers roughly 67 million people across 13 states and Washington, DC. Capacity auction prices in PJM have hit regulatory ceilings for three consecutive auctions. Independent monitors report data centre demand accounted for roughly $29.4 billion in capacity costs across the last four auctions, including $6.3 billion in the most recent event. The Natural Resources Defense Council projects these pressures could add $70 monthly to average household utility bills by 2028.

In response, Virginia created a distinct utility rate category for data centres, Ohio instituted minimum billing, and Illinois is deliberating new measures. Although Google's agreement voluntarily contributes new generation, federal regulators in October 2026 rejected proposals to mandate that large users supply their own power.

Global Implications

The escalating friction provides critical lessons for energy-rich Gulf nations mounting major infrastructure initiatives. Abu Dhabi is pursuing up to five gigawatts of AI capacity, Saudi Arabia aims for nearly two gigawatts through HUMAIN, and Qatar has formed a $20 billion joint venture with Brookfield that could expand to 1.5 gigawatts—representing about 12 percent of Qatar's generating capacity. As electricity availability dictates AI placement, establishing clear commercial terms before contracts lock in remains essential.