This summer, a 50-year-old research reactor at the University of Utah will flip a switch. For the first time. It will churn out electricity—not neutrons for experiments, but actual kilowatts to feed a miniature AI data center.
Ted Goodell, the reactor’s manager, calls it a milestone. Gizmodo reports his words: “This will be, to our knowledge, the first time any university reactor has produced electricity, not just our own. It’s a milestone for our students, but it also shows that small, safe reactors could live at data centers, rather than in labs.”
The setup repurposes the TRIGA reactor, built by General Atomics for academic neutron work. Normally, its heat dissipates into cooling water. Now, Elemental Nuclear’s novel generator captures that thermal output. Expect 2 to 3 kilowatts. Enough for a high-performance GPU node running live AI workloads, managed by the university’s computational team.
Mike Luther, Elemental Nuclear’s founder, frames the stakes. “This project is intended to demonstrate a powerful principle,” he says in the University of Utah announcement. “The energy produced through nuclear fission can ultimately power the computational systems driving artificial intelligence.”
Scale matters. A full AI data center gulps hundreds of megawatts. This test outputs thousandths of that. Proof-of-concept, though. Microreactors—portable, factory-built fission plants—aim for remote sites, military bases, or exactly this: hyperscale computing clusters.
And the pressure builds. AI’s electricity hunger strains U.S. grids. Goldman Sachs pegs Big Tech’s need at 85 to 90 gigawatts of new nuclear by decade’s end, per PC Mag. Data centers alone could triple power draw by 2030.
Tech giants aren’t waiting. Microsoft revived Three Mile Island’s Unit 1 under a 20-year deal, targeting 835 megawatts for AI by 2028. Meta locked in 6.6 gigawatts across projects, including expansions at Vistra plants and SMR development with Oklo and TerraPower, as detailed in NucNet.
Google partnered with Kairos Power for 500 megawatts of molten salt SMRs by 2035, starting with a 50-megawatt TVA grid tie-in for Tennessee and Alabama data centers. Amazon pledged over 5 gigawatts from X-energy’s high-temperature gas reactors by 2039, after leading a $500 million round; X-energy’s recent $1 billion IPO underscores the frenzy, via TechCrunch.
Startups swarm. Last Energy snagged $100 million in December for mini-reactors, eyes on AI loads (Tech Funding News). Aalo Atomics broke ground on a data-center pod prototype, aiming for criticality by July 4, 2026. Oklo, backed by OpenAI’s Sam Altman, went public and pipelines 14 gigawatts, including Equinix deals.
China moves faster. Twenty-six reactors, including 12 SMRs for 960 megawatts, target AI training—straight from a LinkedIn post citing The New Yorker. No cities. Just algorithms.
TRIGA’s trick? A compact Brayton cycle turbine grabs waste heat, spins it into power. Safe—TRIGAs pulse safely on uranium-zirconium hydride fuel. But commercialization lags. Microreactors need fuel certification, regulatory nods, grid hooks. Idaho National Lab’s DOME test bed just opened for Radiant’s Kaleidos unit, per DOE.
Challenges stack. Westinghouse’s Vogtle plants ran seven years late, $18 billion over. SMRs promise factory assembly, two-year builds. Yet NuScale canceled its first U.S. project on costs. Investors bet anyway: Blue Energy raised $380 million for shipyard-built grid-scale units (TechCrunch).
Nvidia aids. PhysicsNeMo simulates reactor designs with GPUs. Oklo and Los Alamos test plutonium fuels via AI digital twins for Genesis Mission.
Back in Utah. Success here validates co-location: reactors beside racks, no transmission losses. Emissions-free baseload. AI demands it—intermittents like solar falter without massive storage.
But hurdles. Public fear lingers from Three Mile Island’s 1979 mishap. Trump eased reviews, yet licensing drags. Supply chains strain on enriched uranium, components.
Elemental’s demo runs this summer. If the GPU hums on fission heat. A spark. For an industry chasing terawatt-hours.
Grids creak. Reactors stir. AI waits for no one.
WebProNews is an iEntry Publication