What is a microreactor

The US Department of Energy defines a microreactor by its thermal output: between 1 and 20 megawatts thermal, used directly as heat or converted into electricity. That is small even next to an SMR, which runs up to 300 MWe per module, and the difference in size carries a difference in kind. An SMR is still a construction project: concrete poured on site, a vessel erected on site, years of building. A microreactor is assembled whole in a factory, travels by truck, ship, rail or plane, and runs for up to ten years without refuelling, on passive safety and without the crew of licensed operators a power station needs. What you buy is not a small plant: it is equipment that gets delivered.

The furthest along is Project Pele, the microreactor BWXT is building for the Strategic Capabilities Office of the US Department of War. It is a high-temperature gas-cooled reactor that puts out at least 1.5 MW electric and fits, systems included, into four 20-foot shipping containers. It runs on TRISO fuel, uranium kernels wrapped in layers of carbon and silicon carbide that hold the fission products inside each particle: fuel fabrication finished in November 2025, core stacking was completed in June 2026, and the demonstration will run at Idaho National Laboratory. Westinghouse's eVinci takes a different route: 15 MWt and 5 MWe with no pumped coolant at all, using heat pipes that move heat out of the core by evaporating and condensing a liquid metal, eight or more full-power years before refuelling, and a footprint under two acres.

None of this is new. Between 1954 and 1977 the US Army Nuclear Power Program built eight reactors of that size and put them to work wherever hauling fuel was the hard part. The SM-1 at Fort Belvoir, Virginia, went critical on April 8, 1957 and became the first US reactor connected to an electrical grid, months before Shippingport. The PM-2A was assembled under the Greenland ice at Camp Century. The PM-3A gave McMurdo Station in Antarctica electricity, steam heat and desalinated water from March 1962 to 1972. And the MH-1A Sturgis was a power plant mounted on a cargo ship, moored in the Panama Canal Zone from 1967 to 1977.

Two lessons from that program still govern design today. The first came from the ML-1, the only genuinely mobile one: truck-mounted, with a closed-cycle gas turbine, it went critical on March 30, 1961, never got past 140 kW against a 300 kW design target, and logged only a few hundred hours of testing before being shut down in 1965. Moving a reactor turned out to be considerably harder than shrinking one. The second came from the SL-1: on January 3, 1961, during a maintenance procedure, an operator withdrew the central control rod about 50 centimetres instead of the intended 10, the reactor went prompt critical within milliseconds, and all three operators died — the only fatal reactor accident in US history. That is where the stuck-rod criterion comes from: a reactor has to be able to shut down with its most reactive rod stuck out, and no single rod may be able to make it dangerous.

What closed that program was not physics but arithmetic: at remote sites fossil fuel was expensive, but less expensive than operating and then decommissioning your own reactor. Today's microreactors are coming back for a reason other than the price of a megawatt-hour. On September 9, 2026 the Department of War announced the Navy's first shore-based microreactor, at Naval Weapons Station Crane, Indiana, no later than September 2028, commercially owned and operated and under the Army's Janus program: what a base buys with that is not cheap electricity, it is staying up when the commercial grid is not. What remains unsettled is regulatory rather than technical. In SECY-24-0008, dated January 24, 2024, the Nuclear Regulatory Commission put the open questions of this deployment model in writing: loading fuel at the factory, operating the reactor for testing at the factory, transporting it already fuelled, and swapping the whole reactor instead of refuelling it.