Energy / Nuclear · Deep dive

Radiant

Portable 1 MW nuclear microreactors off a production line — TRISO-fueled, helium-cooled, shipped fueled in a container, aimed squarely at the diesel generator.

emerging

The question that decides it: Fuel is in the building but the chain reaction is not yet public: Kaleidos took delivery of its TRISO core at INL's DOME in July 2026 for a five-phase campaign ending in a 150-hour unattended full-power run, while the NRC races a December 2026 deadline on the license that lets Radiant fuel reactors at its Oak Ridge factory. If both land, does a factory-fueled 1 MW HTGR actually beat delivered diesel on cost at defense and remote sites — or do DOE-rationed HALEU, academic microreactor cost estimates of $140-410/MWh, and a 50-unit-per-year factory built ahead of any commercial license leave Radiant a defense-subsidized niche carrying a $1.8B price tag?

My take ★★ Very interesting ★ Interesting − Maybe ✗ Not interesting

Snapshot

Radiant is the purest expression of the “SpaceX for nuclear” thesis: a five-year-old El Segundo company trying to be the first in the world to mass-produce nuclear reactors — 1 MW, truck-transportable Kaleidos units that ship fueled from a factory and replace diesel generators at military bases, data centers, and remote sites. It has raised roughly $575M (including a $350M Series D at a $1.8B valuation, Dec 2025-Jan 2026), holds the first slot in the DOE’s DOME microreactor test bed at Idaho National Laboratory, and received its TRISO fuel there on July 1, 2026. It has also never generated a watt of commercial power or a dollar of product revenue, and its entire model depends on regulatory and fuel-supply machinery the US government is still assembling.

Founding story

Doug Bernauer spent about twelve years at SpaceX. His formative assignment was unglamorous and clarifying: figure out how to power a Mars settlement. He built a coded trade-study model comparing energy options and concluded that nuclear fission beat everything else by a wide margin — solar with storage could not carry a Martian winter dust season, and chemical fuel logistics were absurd. The follow-on realization was terrestrial: the same logic applies anywhere on Earth where fuel must be trucked, flown, or barged in. When it became clear SpaceX was not going to build the reactor, Bernauer left and founded Radiant in 2020 with $1.2M in angel money (GlobeNewswire, Sep 2020) and a team drawn heavily from SpaceX, including co-founder and CTO Bob Urberger. The company’s stated strategy is unusually explicit about its own romance: mass-produce reactors for Earth, make money, then send one to Mars on a Starship. Along the way Radiant bolted nuclear-establishment credibility onto the aerospace core — Rita Baranwal, who ran the DOE’s Office of Nuclear Energy, joined as Chief Nuclear Officer.

How it works

Kaleidos is a high-temperature gas-cooled reactor (HTGR) that fits, with its shielding, in a standard shipping-container form factor. The core is built from prismatic graphite blocks holding TRISO fuel — poppy-seed-sized kernels of high-assay low-enriched uranium (HALEU, enriched between 5% and 20%) wrapped in ceramic layers that retain fission products up to roughly 1,600°C, well above anything the core can reach. Pressurized helium flows through channels in the fuel blocks and carries heat to a power conversion system producing about 1 MW electric from roughly 3 MW thermal (World Nuclear News, 2026). Helium is the load-bearing safety choice: it does not become radioactive, so a leak disperses harmlessly, and the reactor needs no water at all — cooling in an emergency is passive air circulation. Radiant demonstrated the passive cooldown in a full-scale test (simulated helium pump failure), and says the safety system scrams the reactor within 300 milliseconds of detecting a fault.

The operating model is the actual invention. A unit arrives by truck, connects to a microgrid within about 48 hours, runs roughly five years without refueling, and then goes back to the factory for a core swap — a “heart swap” model in which customers never handle fuel. That only works if the factory itself is licensed to load fuel, which is why Radiant’s Part 70 special nuclear material license application for its R-50 factory in Oak Ridge, Tennessee — designed to build up to 50 reactors a year on a Manhattan Project-era site — is on an accelerated NRC review with a target completion of December 18, 2026 (NRC, 2026).

Product and business overview

There are effectively four artifacts. The Kaleidos Demonstration Unit is the test article now sitting in INL’s DOME facility — the containment dome of the old EBR-II reactor — for a fueled, full-power campaign. The commercial Kaleidos is the 2028-targeted product: 1 MWe / ~3 MWt, five-year core life, ~20-year design life. The R-50 factory in Oak Ridge is the scaling thesis made concrete — reactors as a manufactured product, not a construction project. The fuel chain is deliberately outsourced but co-located: TRISO fabricated to Radiant’s spec by Standard Nuclear in Oak Ridge, using HALEU allocated by the DOE — Radiant has now won two allocations, the second announced alongside NASA in July 2026 (ANS).

Business model and pricing

Radiant plans to sell reactors outright or deliver electricity under power purchase agreements, with Radiant or a utility operating the unit (TechCrunch, Dec 2025). No price per unit or per MWh has been published — a notable silence at a $1.8B valuation. What exists instead is committed demand of two kinds. Commercial: Equinix signed a preorder with deposits for 20 Kaleidos units in August 2025, the first data-center prepayment for microreactors. Defense: an August 2025 agreement with the Defense Innovation Unit under the Air Force’s Advanced Nuclear Power for Installations (ANPI) program to deliver a first-of-a-kind reactor to a US military base in 2028, value undisclosed, following earlier DoD-funded siting work at Hill Air Force Base. The economic frame Radiant sells against is delivered diesel: at remote and contested sites, fuel convoys and logistics push effective costs to multiples of grid power, and academic work finds microreactors competitive where diesel fuel exceeds roughly $1.50/liter (ScienceDirect, 2023). The uncomfortable other half: a 2025 University of Michigan study put microreactor levelized costs at $140-410/MWh — competitive with almost nothing except that diesel edge case, and generally losing to solar-plus-storage in most regions. Radiant’s bet is that mass production bends that curve; nobody has yet manufactured reactor number two, let alone number fifty.

Traction over time

Headcount is roughly 150-200 (Glassdoor band 51-200, 2026). Revenue: none; deposits and government awards only.

Market analysis

The addressable wedge is every generator that burns trucked-in fuel. The global diesel generator market is projected at $19.3B in 2026 growing to $25.6B by 2031 (MarketsandMarkets); data-center generators alone were ~$9.2B in 2025, headed to $17.3B by 2034 (Precedence Research). Defense installation energy is the beachhead with the least price sensitivity — the Air Force ANPI program alone named eight vendors for on-base microreactors by 2028. The structural tailwinds are real and recent: AI-driven load growth, the ADVANCE Act, 2025 executive orders that created the DOE Reactor Pilot Program and its July 4, 2026 criticality sprint, and a DOE HALEU program standing up domestic enriched-fuel supply. The structural headwind is equally real: in 2023 the US produced roughly 20 kg of HALEU (OilPrice.com, 2025) — a 50-reactor-a-year factory needs a fuel industry that does not yet exist at scale.

Competitive intel

The frontmatter carries the set; the shape of the field matters more than any single rival. Radiant is the smallest unit size and the only player whose core thesis is the factory rather than the reactor. Above it, Oklo, X-energy, and Westinghouse fight for multi-megawatt data-center and utility deals with more power and more institutional heft; X-energy’s owned TRISO fuel plant directly attacks Radiant’s most fragile dependency. Beside it, BWXT’s Project Pele will likely be the first transportable TRISO microreactor to actually run — on the government’s dime and schedule. And behind it, the DOE pilot cohort — Antares, Valar Atomics, Deployable Energy, Aalo — hit zero-power criticality in June-July 2026 while Radiant’s more ambitious full-power DOME campaign was still loading fuel. Radiant’s counter to all of them is the same sentence: nobody else has a fueled full-power test slot, a licensed-track fuel-loading factory, and a deposit-backed commercial order book at once.

History and evolution

• Sep 2020 — Founded by Bernauer and ex-SpaceX engineers; $1.2M pre-seed; pitch is explicitly diesel replacement plus Mars.
• 2022-23 — NRC pre-application work begins; Series A (USV) and Series B (a16z); FEEED award puts Kaleidos among three designs funded toward DOME.
• 2024 — Full-scale passive cooldown test; DOE names Radiant a first DOME experiment (Oct).
• 2025 — The compounding year: $165M Series C, first-in-DOME selection, HALEU allocation, Equinix preorder, DIU 2028 delivery agreement, Oak Ridge factory decision (after a Wyoming flirtation that drew local nuclear-waste pushback in Casper trade press), $300M Series D at $1.8B.
• 2026 — DARK full-power approval (Feb); DOME possession (Apr); accelerated NRC factory-license review opens; fuel on site (Jul 1). Notably absent: Radiant was among the 11 firms in DOE’s Reactor Pilot Program, but was not among the four that demonstrated criticality by the July 4 deadline — its test architecture is different and more ambitious, but the optics ceded “first” headlines to Antares, Valar, Deployable, and Aalo.

What people say

The case for. The enthusiasm is loud and comes from serious places. Packy McCormick’s Not Boring profiled Radiant as the company that could make nuclear a product rather than a project; DCVC and Draper framed the Series C/D as backing the first real reactor assembly line; Energy Secretary Chris Wright has repeatedly showcased the DOME program. Chief Nuclear Officer Rita Baranwal’s line — de-risking a commercial product to be “manufactured and delivered within 18 months” (WNN, Jul 2026) — is the crispest statement of the thesis. Employees rate it 3.9/5 on Glassdoor (small sample), 78% recommend; reviews describe smart, driven, mission-obsessed colleagues and high ownership.

The complaints. Three clusters. Economics: peer-reviewed work (Michigan, 2025; MIT/Nuclear Technology reviews) puts microreactor LCOE at $140-410/MWh, uncompetitive with solar-plus-storage almost everywhere and with diesel except at genuinely remote or militarized sites; nuclear commentator Chris Keefer’s Decouple essay on microreactors-versus-diesel makes the blunt case that gensets are absurdly cheap, mass-produced, and serviceable by any mechanic on Earth. Fuel: HALEU supply is DOE-rationed, US production was measured in tens of kilograms as recently as 2023, and one estimate puts ten years of fuel for a single small microreactor north of $32M (OilPrice.com, 2025) — Radiant’s entire fleet plan rides on allocations it does not control. Regulatory: a DOE test authorization at a national lab is not an NRC commercial license; the Part 70 factory review is accelerated but unfinished, and the commercial reactor licensing path beyond the DOME data is still being written. Glassdoor’s minority reports echo any hard-tech sprint: nonexistent work-life balance and, in one pointed review, a favoritism culture where relationships outrank results.

Outlook: the open question

The bull case is confirmed if, by late 2027, three things are true: Kaleidos completes its 150-hour unattended full-power run at DOME on prototypic fuel and coolant; the NRC grants the R-50 Part 70 license near its December 2026 target so Oak Ridge can load fuel; and either Equinix’s deposits convert to firm-priced orders or the Air Force delivery lands in 2028 at a disclosed cost anywhere near diesel-parity for defense sites. At that point Radiant is what its investors already priced: the only company with a proven product, a licensed factory, and a paying queue in the one nuclear segment where speed of manufacture is the moat. The bear case is confirmed if the DOME campaign slips or surfaces fuel-performance issues, HALEU allocations tighten under competing defense and NASA claims, or the first published price lands at the high end of the academic $140-410/MWh range — in which case Kaleidos is a brilliant piece of engineering serving a market small enough for BWXT and Westinghouse to absorb on government contracts, and the $1.8B valuation was paid for optionality on a factory that has no license to run. The honest scorecard as of August 2026: every milestone Radiant controls, it has hit early; every input it does not control — fuel, licenses, price discovery — is still someone else’s decision. That asymmetry, not the physics, is the investment question.

How a challenger would attack it

The wedge. Radiant’s single most fragile input is fuel, and a challenger attacks by simply not needing it. Last Energy’s contrast is the template: conventional LEU in a 20 MWe containerized PWR sidesteps the HALEU bottleneck entirely — US production was measured in tens of kilograms in 2023, allocations are DOE-rationed against competing defense and NASA claims, and one estimate puts ten years of fuel for a single small microreactor above $32M. A challenger building on LEU, or buying from X-energy’s owned TRISO-X plant, controls the input Radiant must petition for. The second vector is price disclosure: Radiant has published no price per unit or per MWh at a $1.8B valuation, while academic LCOE estimates run $140-410/MWh; a rival that publishes a firm PPA price in the Last Energy $130-200/MWh style forces Radiant into price discovery on someone else’s terms, in front of the Equinix-class buyers whose deposits are still refundable. Third: the diesel incumbent itself — gensets are mass-produced, absurdly cheap, and serviceable by any mechanic on Earth, and a hybrid solar-storage-genset microgrid vendor can beat Kaleidos on delivered cost everywhere except the contested-logistics defense edge case. Radiant’s factory-ahead-of-license sequencing means every regulatory slip past December 2026 is capacity burning cash.

Same playbook, new buyer

The factory-not-project thesis is portable to buyers Radiant’s 1 MW, five-year-core architecture happens to fit better than anyone else’s. The nearest shift is heat, not electricity: a ~3 MWt HTGR delivers high-temperature process heat that diesel gensets and heat pumps cannot, and remote mining operations, Arctic and island communities, and forward industrial sites currently burning trucked fuel for heat-plus-power are the segment where the $1.50/liter diesel-parity math already closes — no data-center price war required. Second: allied-government sales. Canada’s remote north, Pacific island territories, and NATO installations all face the same fuel-convoy economics as US bases, and a partner-operated fleet model (Radiant’s own sell-or-PPA structure, franchised to national utilities) reaches them without Radiant building foreign regulatory muscle. The heart-swap model is the differentiator no incumbent will copy: BWXT and Westinghouse are institutionally wired for cost-plus government prototypes and decade-long site licensing, not for a return-to-factory refueling loop that requires owning a production line. Whoever runs the factory playbook for the 5-20 MW industrial-heat tier — above Kaleidos, below Oklo’s utility scale — inherits the same logic in a less crowded weight class.

Sources and further reading

• Radiant — Series D announcement: $300M+ to mass-produce portable nuclear reactors (Radiant Nuclear, Dec 17, 2025)
• Axios Pro — Radiant boosts Series D to $350M at $1.8B valuation (Jan 7, 2026)
• TechCrunch — Radiant Nuclear raises $300M for its semi-sized 1 MW reactor (Dec 17, 2025)
• World Nuclear News — TRISO fuel delivered for Kaleidos reactor experiment (Jul 2026)
• ANS Nuclear Newswire — NRC to review Radiant R-50 Part 70 license application (2026)
• Radiant — Equinix preorder and deposits for 20 Kaleidos microreactors (Aug 14, 2025)
• ANS Nuclear Newswire — Radiant signs contract on microreactors for the military (DIU/ANPI) (Aug 14, 2025)
• Neutron Bytes — Four microreactors hit criticality milestones by July 4 (Jul 8, 2026)
• University of Michigan NERS — Evaluating the economic viability of microreactors in today’s electricity markets (Dec 2025)
• Contrary Research — Radiant business breakdown and founding story (2025)
• Not Boring — Radiant (Packy McCormick, 2024)