Universities and China test nuclear reactors to power AI data centers amid soaring energy demands

University of Utah Tests Nuclear Reactors for AI Data Centers

The University of Utah is preparing to demonstrate whether nuclear reactors can address the mounting energy demands of AI data centers. This summer, in collaboration with Elemental Nuclear, researchers will repurpose the university's TRIGA reactor to power a mini AI data center in what they describe as a proof-of-concept experiment 1

1

. A novel power generator will capture thermal energy from the reactor to generate around 2 to 3 kilowatts of electricity—enough to power a live AI workload on a high-performance GPU node managed by the university's computational experts.

"This will be, to our knowledge, the first time any university reactor has produced electricity, not just our own," reactor manager Ted Gooding said. "It's a milestone for our students, but it also shows that small, safe reactors could live at data centers, rather than in labs" 1

1

. While the output is far less than the hundreds of megawatts full-scale facilities require, the experiment aims to illuminate whether microreactors for data centers could simplify nuclear energy generation for AI operations.

China Develops Truck-Mounted Miniature Nuclear Reactor

China is testing a more ambitious approach with a truck-mounted 10MW vehicle-mounted nuclear power unit designed to sustain massive data centers for decades without refueling . Wu Yican, chief scientific adviser to the Institute of Nuclear Energy Safety Technology, described the system as a "nuclear power bank" that exemplifies the new generation of nuclear energy systems. The miniature nuclear reactor can be transported to areas where grid connections are unavailable or unreliable, offering a decentralized energy source for remote regions, emergency backup, and AI computing facilities.

The mobility of this portable system addresses a critical gap as AI computing continues expanding beyond what existing grid infrastructure was built to handle. China currently operates 59 commercial nuclear units producing 467.7 billion kilowatt-hours of electricity, accounting for 4.82 percent of national demand and placing it second globally behind the United States .

Tech Companies Embrace Nuclear Fission for High Electricity Demands

Major tech companies are increasingly turning to nuclear solutions as they confront the high electricity demands of AI infrastructure. Amazon plans to develop 12 small modular nuclear reactors with nearly 1GW of combined output, while Google is building a small reactor in Tennessee to supply its facilities . Amazon, Meta, Google, and Microsoft have all joined the World Nuclear Association, signaling industry-wide commitment to nuclear power. Microsoft has pushed into legacy infrastructure with involvement in restarting the Three Mile Island nuclear plant.

Elemental Nuclear founder Mike Luther emphasized the broader principle: "The energy produced through nuclear fission can ultimately power the computational systems driving artificial intelligence" 1

1

. Smaller systems such as Westinghouse's eVinci microreactor are being developed to deliver about 5MW for up to 100 months without refueling, demonstrating the industry's focus on compact, long-lasting solutions.

What This Means for AI Infrastructure

These experiments arrive as U.S. grids face mounting stress from AI data centers that threaten to overwhelm existing capacity. The University of Utah project leverages TRIGA research reactors—designed by General Atomics for academic research and essentially "neutron factories" according to the World Nuclear Association—to test whether tiny nuclear reactors for AI data centers can be simplified 1

1

. The power generator developed by Elemental Nuclear captures thermal energy normally dissipated by cooling systems and converts it into electricity.

While stakeholders still have much to work out before microreactors can be deployed at commercial scales, these initiatives could alleviate some impending problems. The short-term implications focus on validating technical feasibility, while long-term prospects involve creating portable, reliable energy sources that reduce dependence on traditional grids. Watch for further developments as these proof-of-concept experiments progress and tech companies continue investing in nuclear infrastructure to meet AI's insatiable appetite for power.