EnergyBasicsRadiant’s Kaleidos microreactor left El Segundo, California, in August on a trailer bound for Idaho National Laboratory. It is a 1 MWe, helium-cooled, TRISO-fueled unit, and the Department of Energy says its parts fit inside a single shipping container. 🚚 That image explains the headline. A box that arrives by truck, plugs in and runs for years without refueling sounds like the perfect answer to a grid knocked flat by a storm.
The honest answer is “probably not your next outage.” ⚡ As far as public records show, no commercial microreactor is delivering power to a US grid today. The first Army unit is due in September 2028, and it will serve a base, not a neighborhood. What exists now is a fast-moving set of tests, contracts and rulemakings that make the disaster-response idea plausible for the early 2030s. 🔬
The container is real, the outage fix is not yet
The disaster-recovery case rests mostly on modeling so far. An engrXiv preprint simulates 500 kW to 5 MW container-sized units restoring grids after severe weather, and it frames each one as a “nuclear battery” with a 1 to 10 year life. Preprint, so treat it as a hypothesis, not a result. 📄
The ship version is older and bolder. BWXT and Crowley announced a concept in 2023 for a 378-foot shallow-draft vessel carrying a 5 to 50 MW reactor to power shore facilities after disasters. It is a memorandum of understanding, and nothing in the search results shows a hull in the water. That fact may be stale. 🚢
Kaleidos (Radiant): 1 MWe / 3.5 MWt, five years between refuelings, 20-year asset life
eVinci (Westinghouse): 5 MWe / 15 MWt, eight or more years between refuelings
R1 (Antares): 100 kWe to 1 MWe per unit, sodium heat-pipe cooling
BWXT/Crowley ship: 5 to 50 MW, concept stage only
Would a utility in your region pay for a standby reactor the way it pays for spare transformers? Worth asking them.
Three designs, three different test records
Antares Nuclear moved first. Its Mark-0 sodium heat-pipe reactor became the first to go critical under the DOE Reactor Pilot Program, at Idaho National Laboratory in June. 🏁 Four projects crossed the line by July 4: Antares, Valar Atomics, Deployable Energy and Aalo Atomics. All four were zero-power demonstrations, which is a physics check and not a power plant.
Westinghouse completed zero-power criticality on its eVinci test reactor on August 24 at the NNSA’s National Criticality Experiments Research Center in Nevada. DOE then reported that the test reached criticality at 663 degrees Celsius on September 3. The test unit is a one-fifth-scale version of the commercial design, so the full-size reactor is still ahead. 🔥
Radiant has the hardest exam. Its plan has five phases, ending with 150 hours at full power without operator intervention, and it targeted completion by September 30. Radiant itself says going critical alone does not generate meaningful power. The search results show no public confirmation that the campaign finished, so the status below is unverified. 🧪
Antares: Mark-0 critical in June 2026, HALEU TRISO fuel, zero power
Westinghouse: zero-power criticality August 24, hot test at 663 C on September 3
Radiant: fuel delivered to DOME in July, full-power results not confirmed
Deployable Energy: 1-MWe gas-cooled Unity unit critical at INL by July 1
The Army writes the first checks
On August 26, the Army awarded up to $2.2 billion over five years under its Janus Program to Antares, BWXT Advanced Technologies, General Atomics Electromagnetic Systems, Radiant and Westinghouse Government Services. 💰 Each company gets a different base. About 20 microreactors are expected, and the goal is a first unit running by September 2028.
Radiant’s piece is a binding agreement worth up to $750 million for 15 Kaleidos units, starting with a three-unit installation. Antares plans three R1 units at Fort Bragg, and Westinghouse takes Fort Drum, New York. The Army regulates these reactors itself, not the NRC, which is a large part of why the schedule is aggressive. 🪖
The economics are specific. An Army official says the military pays about 40 cents per kilowatt-hour to power Arctic installations. Janus is partly a test of whether factory production pulls that number down, which is the same question a utility would ask before buying a standby unit. Every project, deal, company, and regulatory event we cover is logged, verified, and searchable in SMRbrief Pro. The bases stay connected to the local grid, so these reactors add resilience, not independence.
Total award: up to $2.2 billion, five years
Radiant: up to $750 million, 15 Kaleidos units, three-unit first phase
Target: first reactor operating September 2028
Regulator: the Army, not the NRC
Regulators decide how fast the boxes ship
The NRC proposed 10 CFR Part 57 in April, a licensing framework for microreactors and other low-consequence designs. Part 57 targets reactors of about 100 MWe or smaller, and the NRC estimates it saves the agency and industry $3.76 billion to $11.84 billion. 📉 Law firm Pillsbury reads the proposal as potentially cutting licensing to 6 to 12 months, and the rule contemplates remote and autonomous operation.
For a mobile unit, this matters more than any single test. Radiant’s R-50 factory license in Oak Ridge, Tennessee, is under NRC review, with approval targeted for the fourth quarter. 🏭 A reactor that rolls off an assembly line needs a factory license first and a customer license second. Part 57 is still a proposed rule, so the final text could change.
Part 57: proposed April 2026, not final
R-50 Part 70 application: accepted May 2026, expedited review
Operations: remote and autonomous monitoring contemplated
Open question: how a unit relocates between sites under a license
What an outage response would actually take
Moving a reactor into a storm zone needs more than a truck. It needs a site-specific environmental review, a trained crew, a utility interconnection and a licensing path that allows relocation. Radiant’s own 2024 slides note that each reactor requires its own environmental impact statement, and its listed shutdown-to-shippable time is up to 30 days. 🌪️ That is far slower than a diesel generator, which means the first civilian use is more likely planned backup for hospitals and data centers than emergency response.
The sensible forecast is the 2030s, and only if Janus works and Part 57 survives comment. Radiant tells Reuters that data center customers arrive around 2030. 📈 Disaster relief probably follows once the licensing and logistics for moving units are proven.
So who gets the first container that moves in response to a blackout: a base, a hospital, or a whole town, and who pays for the standby unit while it waits?