Radiant Kaleidos DOE Full-Power Approval

The U.S. Department of Energy has approved the Documented Safety Analysis for Radiant Industries’ Kaleidos, so a commercial microreactor design can seek an extended full-power run at Idaho National Laboratory instead of another zero-power criticality test. The planned campaign at the National Reactor Innovation Center’s Demonstration of Microreactor Experiments (DOME) test bed reaches 60 effective full-power days, Nuclear News Network reported. Radiant, the El Segundo company building the unit, said the approval lets it advance to DOE’s Readiness Review, then fuel the reactor and start a five-phase test campaign, in its announcement carried in the final days of October 2026.

Radiant is still private, so this gate does not print through a ticker. The price signal for infrastructure buyers sits in the nameplate, the contract ceilings, and the distance between a DOE test authorization and an operating license. The Nuclear Regulatory Commission lists Kaleidos as a transportable high-temperature gas-cooled reactor designed for 3 megawatts thermal and about 1 megawatt electric, on a pre-application docket open since October 2022. A machine at that scale can matter at a remote site or as a first on-site block. It does not retire the interconnect queue in front of a large training or inference hall.

Key Takeaways:

DOE approved the Kaleidos Documented Safety Analysis for power operation, with a planned 60 effective full-power-day campaign at INL’s DOME. A Readiness Review still sits ahead of fueling.

NRC lists the design at 3 MWth and about 1 MWe. Equinix’s 20-unit preorder is a fleet signal, not a hyperscale power plant.

Peer 2026 milestones for Aalo-X, Antares, Westinghouse eVinci, and DOE’s MARVEL are criticality or startup gates. They are not multi-week power runs.

A DOE safety analysis for a test at Idaho is not an NRC commercial operating license. Radiant’s NRC engagement is still pre-application.

What the design is specified to deliver

Kaleidos is a factory-oriented, transportable high-temperature gas-cooled microreactor. That description comes from the NRC’s pre-application page, not from a marketing line. Thermal output is listed at 3 MWth. Electric output is listed at approximately 1 MWe. Radiant’s own media-room listing for the Equinix preorder uses the round figure of 1 megawatt. Those two public numbers line up. They also set the ceiling on what a buyer can honestly model before the Idaho campaign produces hours at power.

The buyer set already on paper is split between defense installations and data-center operators. Equinix signed a preorder for 20 Kaleidos units, World Nuclear News reported, and an IAEA compilation dated October 19, 2026 says that agreement includes customer deposits. The same World Nuclear News story says Equinix intends to sign a power-purchase agreement with Dutch developer ULC-Energy. That PPA intent is a separate commercial thread. It is not a statement that ULC-Energy will operate the 20 Radiant units.

World Nuclear News filed the preorder in a story about small modular reactor developers. The NRC rating does not support that category slip. About 1 MWe is a transportable block, closer to a remote feeder or a small on-site slice than to a multi-hundred-megawatt campus plant. Twenty units at the NRC figure remain a low-tens-of-megawatts fleet if every machine hits nameplate. Infra leads should underwrite the preorder as an option on a tested product, with deposits at risk if the Idaho hours slip, not as firm capacity dated to a building schedule.

The defense side is larger in dollars and earlier in deployment language. On August 26, 2026, Radiant said the U.S. Army, with the Defense Innovation Unit, had executed a binding agreement of up to $750 million to develop and deploy 15 Kaleidos units through the Army’s Janus program, in the company release. Power Magazine described the same ceiling and framed Janus as a five-company push with commercial microreactors starting in 2028. A Yahoo Finance report put the wider Army effort at $2.2 billion and said Janus, unveiled in October 2025, aims to put portable microreactors on military bases by 2028, in a class it described as typically 1 to 20 megawatts. Kaleidos sits at the bottom of that band on the NRC’s electric figure.

The safety analysis, the fuel shipment, and DOME

The sequence in front of those 60 days is still operational, not ceremonial. Radiant said the next step is DOE’s Readiness Review, which assesses whether the reactor and the operating team are prepared to begin nuclear operations. Only after that review does the company say it will fuel Kaleidos and open the five-phase campaign. The public releases do not name the five phases. What they do establish is that fueling is not done.

Fuel hardware moved earlier than the final safety document. On July 3, 2026, World Nuclear News reported that the first shipment of fuel for full-power, full-temperature testing of Kaleidos technology had been delivered to DOME at Idaho National Laboratory. Delivery to the test bed is not the same as loading the core. The July shipment shows the fuel line ran ahead of the October DSA. The October document shows the safety case caught up on paper. The Readiness Review is the step that has to join those two facts before a chain reaction at power is even attempted.

By mid-February 2026, coverage of an earlier DOE milestone still described Radiant as preparing its final documented safety analysis and conducting readiness reviews. The company has a separate release on approval of a preliminary documented safety analysis, at radiantnuclear.com. The late-October decision is the later document: the one Nuclear News Network ties to power operation and a 60-day effective full-power campaign. Collapsing the February story and the October story into one “DOE said yes” erases eight months of remaining safety work.

DOME itself is a test bed, not a customer site. The National Reactor Innovation Center describes it as a place where experimental reactors can go critical for the first time, built around the containment system of a decommissioned Experimental Breeder facility, on the NRIC DOME page. NRIC has also selected Deployable Energy to test that company’s nuclear battery inside DOME in 2027, according to the center’s site. A solicitation for later industry applicants says fueled experiments after the first reactor experiment would be scheduled once that first experiment is complete, projected for July 2027, in the NRIC request for applications. The solicitation does not name Kaleidos as that first experiment. What it does tell a buyer is that DOME time is a queue, and a slip in the first campaign moves everyone behind it.

Army, Navy, Equinix, and the capital stack

Three commercial or government commitments are public enough to underwrite, and they do not say the same thing.

Counterparty	Instrument	Scale in the source	Source
Equinix	Preorder for Kaleidos units, with customer deposits noted in an IAEA compilation	20 units	World Nuclear News; IAEA PDF, October 19, 2026
U.S. Army with the Defense Innovation Unit	Binding Janus agreement announced August 26, 2026	Up to $750 million for 15 units	Radiant release; Power Magazine
U.S. Navy	Selection to deploy Kaleidos at a weapons station	Naval Weapons Station Crane, Indiana	Radiant announcement via Yahoo Finance; NucNet, October 2026

The Army figure is not clean across outlets. Radiant and Power Magazine both use up to $750 million for 15 units. The American Nuclear Society, on October 23, 2026, described a Janus figure of up to $650 million to deploy 15 Kaleidos reactors, in a story whose main subject was the naval-station selection. Those ceilings differ by $100 million. Until the award documents are public, the contract value is a range reported by the company and a professional society, not a single settled obligation. “Up to” also means the ceiling is not revenue. Development milestones can leave a large share unearned.

The Navy selection is a site, not a second $750 million headline. Radiant said the Department of the Navy selected Kaleidos for Naval Weapons Station Crane, in an announcement carried by Yahoo Finance. NucNet reported the Indiana base deployment in October 2026 and tied a prototype to Janus. No public dollar figure for the Crane scope appears in those accounts. A base deployment and a 15-unit Army agreement can share a reactor design without sharing a budget line.

Venture capital behind the program is easier to date than to total. On November 14, 2024, Radiant announced a $100 million Series C. On May 28, 2025, it said it had closed a $165 million Series C and that total venture funding then stood at $225 million, in the Series C close note. Those two 2024-2025 announcements should not be added together. The later close is the round size the company published, and $225 million is the cumulative venture figure it published that day. On December 17, 2025, Radiant said it had raised more than $300 million in new funding, in its Series D announcement. A Tracxn profile dated September 29, 2026 put cumulative funding at $518 million over seven rounds. That tracker total is not a company confirmation, and it should not be merged with the December 2025 “more than $300 million” into a single Radiant-stated sum. The usable fact is narrower: by late 2025 the company was raising in nine-figure increments, and by late 2026 third-party trackers still had it private with roughly half a billion dollars of recorded capital.

That capital is funding a test article and a factory thesis at the same time. The Idaho campaign has to show that a fueled high-temperature gas-cooled core can hold power for weeks inside DOME. The Janus language points at military bases in 2028. The gap between a Readiness Review in late 2026 and a base deployment two years later is the schedule a defense buyer is actually buying. A data-center preorder with deposits is a cheaper option on the same evidence. Neither converts to firm megawatt-hours until the 60-day campaign, or a successor test, produces them.

Peer milestones that are not a power run

2026 produced a cluster of advanced-reactor headlines. Most of them cleared criticality, which is a physics gate. Kaleidos’s October document is a safety-analysis gate for a power campaign that has not started. Mixing those boxes is how a procurement memo overstates how close any of these machines is to a busbar.

Design	Developer	2026 milestone	Site	Bound of the result
Kaleidos	Radiant Industries	Documented Safety Analysis approved for power operation; 60 effective full-power days planned	INL DOME	Readiness Review and fueling still ahead of the campaign
eVinci test core	Westinghouse Electric Company	Zero-power criticality, August 2026	NCERC, Nevada National Security Site	No electricity and no usable heat from that test
Aalo-X	Aalo Atomics	Zero-power fueled criticality on July 4, 2026	Idaho National Laboratory	Zero-power test, counted by DOE as the fourth such milestone
MARVEL	Idaho National Laboratory	Final safety document approved as the primary coolant system arrived, October 2026	Idaho National Laboratory	Prepared for initial startup tests, not a commercial power campaign

The July 4 cluster came from a deadline, not from a sudden change in reactor physics. DOE said four privately designed reactors completed zero-power fueled criticality tests before July 4, 2026, and that Aalo Atomics’ Aalo-X was the fourth, reached at INL on that date, in the department’s fact sheet and its July 6, 2026 note. Idaho National Laboratory reported that Antares Nuclear’s advanced microreactor was the first to reach criticality under DOE’s Reactor Pilot Program, on June 4. NucNet reported Deployable Energy’s Unity test as the third. Reporting carried by MSN tied the deadline to a presidential executive order issued in May 2025. A DOE post identified the program as a June 2025 response to Executive Order 14301. Kaleidos was not one of the four names DOE celebrated against that date. Its later gate is a different document.

Westinghouse’s eVinci result is the cleanest illustration of what zero-power criticality does not buy. The company said it completed zero-power criticality testing on a test reactor with Los Alamos National Laboratory and Idaho National Laboratory at the National Criticality Experiments Research Center, validating core models, in its August 2026 release. mgrid.org noted that the 5 MW design produced no electricity and no usable heat at the power level of that test. Westinghouse has separately said a zero-power high-temperature criticality test reached 663 degrees Celsius, with a peak core temperature of 1,011 degrees Celsius, in a company release. Those temperatures are a vendor measurement from a zero-power test. They are not electric output, and no independent benchmark in the public record restates them. Coverage of the Nevada work also described a design target of eight years of continuous power without refueling. That is a design target, not a result from the August test.

MARVEL is a federal research reactor on the same Idaho site, and it is not a commercial alternative to Kaleidos. In October 2026, Idaho National Laboratory said DOE had approved MARVEL’s final documented safety analysis and that the primary coolant system had arrived, with initial startup tests aimed at the coming winter, in the laboratory release. Separate reporting has described MARVEL as sodium-potassium cooled, a different coolant class from Radiant’s high-temperature gas-cooled design. MARVEL’s job, on INL’s own description, is a DOE test bed for microreactor applications. It does not carry an Equinix preorder or a Janus ceiling.

The license gap in front of a campus megawatt

DOE’s safety analysis governs a test inside a DOE facility. The NRC docket is the path to a machine a private operator could license at a site that is not Idaho National Laboratory. Those processes can run in parallel. They do not substitute for each other. Radiant has been in NRC pre-application since October 2022. Public submittals on that docket include a white paper on fleet operator training, qualification, and staffing, and a gap analysis against 10 CFR 55, the operator-licensing rule. Pre-application discussion is how a vendor learns where the rule does not fit a transportable core. It is not an accepted application, and it is not a construction permit or an operating license.

That split is the practical risk for a data-center energy lead. A successful 60-day run at DOME would be evidence about the fueled system under a DOE safety case. It would not, by itself, authorize Equinix or anyone else to install a fueled unit behind a meter in a U.S. commercial park. The Army and Navy paths may move under defense authorities on a different clock than an NRC combined license or operating license for a civilian site. A buyer who treats the October DSA as “licensed for deployment” is buying a press milestone, not a permit.

The commercial paper has the same lag built in. Equinix’s 20-unit preorder, already public months before the October DSA, is a deposit-backed reservation on a product whose power run had not been authorized when the reservation was being discussed in mid-2026. Janus points at 2028 for base deployments, while the test article in late 2026 is still in front of a Readiness Review. If the review finds the operating team or the safety basis short, the 60-day clock does not start, and both the Army ceiling and the Equinix deposits wait on a redesign or a rewritten analysis. If the review clears and the campaign starts, the number that will matter is how many of those 60 effective full-power days the core actually delivers, at what electric output, and with what trips. None of those operating figures exist yet, because the run has not begun.

The next documents that change the underwriting are specific. One is the outcome of DOE’s Readiness Review, which Radiant has named as the step before fueling. Another is whether NRC pre-application converts into an application the staff can docket, rather than another white paper. A third is a public reconciliation of the $750 million company figure and the $650 million figure in the American Nuclear Society report, preferably from the award itself. A fourth is any Equinix filing that converts a 20-unit preorder into a site, a delivery year, and a price per megawatt-hour. Until those four exist, Kaleidos is a well-funded test program with real government ceilings and a real safety-analysis approval for a power campaign at DOME. It is not yet a megawatt on a data-center bus.

More in-depth coverage from this blog on closely related topics:

Sources and References

Sources cited while researching and writing this article: