As U.S. electricity demand rises and technology companies seek to build more and larger data centers to drive artificial intelligence systems, the main question arising is how to generate all that power.

According to the International Energy Agency, large-scale data centers around the world used about 460 terawatt-hours of electricity in 2022, a figure that analysts expect to continue rising years into the future.

One potential solution being proposed is nuclear energy – produced by existing large-scale nuclear power plants, reactivated old ones, new ones that might be constructed with government subsidies, and other, smaller types of nuclear plants that are in development and not yet available.

The discussion around powering AI data centers, in particular, has involved a type of nuclear power plant called a small modular reactor. According to the International Atomic Energy Agency, there are about 70 different designs being researched and developed around the world, including reactors that could one day serve small or remote communities, military applications and even ships at sea or spacecraft.

Proponents say these reactor designs provide consistent power without climate-changing carbon emissions. They can also be located close to places that need their energy, reducing dependence on the electricity grid. They are still years from being commercially available: Demonstration projects may begin construction before 2030, with commercial ones reaching operation perhaps by the mid-2030s. And there is not yet a plan from the U.S. Department of Energy to

© The Conversation