Overview
- University of Utah researchers and Elemental Nuclear plan a summer test that repurposes the campus TRIGA research reactor to run a high‑performance GPU on a live AI task.
- Engineers will use a reverse Brayton cycle, which compresses helium gas, heats it with reactor pool water, spins a small turbine, and then cools the gas in a cryogenic heat exchanger.
- Project targets call for about 50 kW of heat from the reactor, roughly 13 kW at the turbine, and a net electrical output of 2–3 kW to feed a single GPU node.
- TRIGA units are research reactors built for training and neutron work, not power production, so this is a small demonstration and not a solution for data centers that draw megawatts.
- Elemental Nuclear frames the trial as groundwork for testing across the TRIGA network and for a commercial microreactor push by 2030–2031, with scaling and regulatory hurdles still ahead.