Radiant Nuclear

Phase A — Understand the business

Lens 1 · Company Overview

What it is. Radiant Nuclear (legal entity Radiant Industries, founded 2019–2020, HQ El Segundo CA, manufacturing relocated to Oak Ridge TN Oct 2025) is building Kaleidos — a 1 MW-electric / ~1.9 MW-thermal portable microreactor that fits inside a single 8×20 ft shipping container and is designed to be mass-produced in a factory, not stick-built on site.

The product in plain terms. Kaleidos is a high-temperature gas-cooled reactor (HTGR): TRISO fuel pebbles/compacts, helium coolant, prismatic graphite core, and — the design signature — a closed-loop CO₂ (supercritical) Brayton power-conversion cycle rather than the water/steam Rankine cycle that forces every conventional reactor to sit next to a river. It is air-cooled (fans + an air jacket, natural convection), so it needs no on-site water, can drop into a desert forward base or a rooftop-constrained data-center campus, and is trucked or flown in. Design life ~20 years; refuel every ~5 years by swapping the whole unit back to the factory (no on-site refueling). A five-year fuel load displaces roughly 22,000 tons of diesel.

How it makes money. Dual model: direct sale or Power-Purchase Agreement (PPA), with Radiant retaining the fuel-swap and end-of-life-removal obligation. Economically this is a razor / razor-blade: hardware (or a per-MWh PPA) plus a recurring fuel-services annuity every ~5 years, with fleet-management software ("SimEngine") monetizing autonomous operation. That recurring fuel/service leg — not the box itself — is where a durable margin can live.

Customers, suppliers, competitors (named).

• Customers / demand book: Equinix — 20-unit pre-order with deposits (Aug 2025); U.S. Air Force / DoD (via Defense Innovation Unit, ANPI program) — agreement to deliver a working microreactor to a base by 2028, Radiant paired with Buckley Space Force Base, Colorado. Target segments: remote/forward military bases, data centers, disaster relief, off-grid communities, mining, industrial heat.
• Key supplier (single-source, critical): Standard Nuclear (Oak Ridge TN) fabricates the TRISO fuel; Centrus / DOE supply the HALEU feedstock (conditional allocation). (See Lens 2 — this is the chokepoint.)
• Competitors: Westinghouse eVinci, Oklo (Aurora), BWXT Project Pele, Antares R1, Aalo Atomics (Aalo-X), Last Energy, X-energy, Nano Nuclear, General Atomics, Kairos.

Contract structure / concentration. Order book today is thin and concentrated: one anchor commercial customer (Equinix, 20 units) and one government demonstration agreement (USAF/DIU). No disclosed backlog dollar value. Both are framework/pre-order commitments contingent on a working reactor — not firm, priced, delivered revenue.

Lens 2 · Supply Chain

Map: HALEU feedstock → TRISO fuel fabrication → Radiant (reactor integration, Oak Ridge R-50 factory) → containerized unit → end customer (base / data center / off-grid site) → factory fuel-swap every ~5 yrs → spent-fuel handling.

Named stakeholders and chokepoints:

• Enrichment (the national chokepoint): Centrus Energy (American Centrifuge Plant, Piketon OH) is the only operating U.S. HALEU producer; it had delivered ~900 kg to DOE by Jan 2026. DOE has committed >$2.7B to build domestic HALEU/LEU capacity, with Orano also awarded. The gap is structural: by 2030 DOE projects ~40 MT of HALEU demand against ~12 MT of funded Centrus capacity — a ~28 MT shortfall.
• Fuel fabrication (single-source, itself pre-revenue): Standard Nuclear (the successor to Ultra Safe Nuclear's TRISO fuel business) fabricates Kaleidos fuel "to Radiant's specifications" and delivered the first tranche to DOME on July 1, 2026. Standard Nuclear is itself filing an S-1 to go public and warns in its own draft disclosures that HALEU supply "is currently limited and we do not know when, or if, supplies will increase." So Radiant's most safety- and schedule-critical input runs through a single, early-stage, not-yet-public vendor dependent on a nationally scarce feedstock.
• Reactor integration / manufacturing: Radiant's R-50 factory (Oak Ridge TN, ~$280M, targeting ~50 units/yr) is the intended mass-production node; first Kaleidos was assembled on the Manhattan Project site.
• Test-bed infrastructure (government-provided moat): INL's DOME facility (repurposed EBR-II containment, up to 20 MWth) — Radiant took one-year possession April 1, 2026. Access to a licensed DOE full-power test bed is scarce and hard to replicate.

Verdict on the chain: vertically light on the reactor, dangerously thin on fuel. The whole deployment thesis is gated by HALEU/TRISO — an input Radiant does not control and that the entire advanced-reactor cohort is fighting over at once.

Lens 3 · Competitive Advantages (moats)

Where the moat is real:

1. Execution lead on the test-bed track. Radiant is the first company to test a reactor in INL's DOME and has fuel on site (July 1, 2026). That's tangible, capital-and-clock-intensive progress a new entrant cannot leapfrog.
2. Regulatory human capital. Dr. Rita Baranwal (ex-DOE Assistant Secretary for Nuclear Energy; ex-Westinghouse AP300 CTO) as Chief Nuclear Officer, plus Steve Burns (ex-NRC Commissioner) on the board. In an industry where the binding constraint is licensing, that roster is a moat.
3. Form-factor + no-water design. Container-scale, air-cooled, factory-swap refueling — a genuinely differentiated deployment envelope vs. site-built SMRs. The CO₂ Brayton cycle removes the water dependency that constrains most peers.
4. Anti-stealth transparency as a go-to-market flywheel — public milestone cadence builds regulator, customer and capital confidence (rare in nuclear).
5. Prospective manufacturing scale + fuel-services annuity — if the R-50 factory hits ~50/yr, unit-cost learning curves and a recurring fuel-swap revenue stream are the durable-moat prize.

Where the moat is thin / contested:

• No IP-lock on the physics. HTGR/TRISO is a well-trodden design space; BWXT Project Pele (1.5 MWe HTGR) and X-energy (TRISO-X) are equally credentialed and, in BWXT's case, backed by a multibillion-dollar naval-reactor incumbent.
• Bargaining power is currently against Radiant on the input it needs most (HALEU/TRISO) and neutral on demand (customers hold pre-orders, not firm irreversible commitments).
• The scale moat is prospective, not present — it exists only if the factory ramps, which no first-of-a-kind nuclear program has yet done on time or on budget.

Lens 4 · Segments (private swap → product/market segmentation & traction — no reported revenue segments)

No revenue, so no segments.csv breakout exists — n/a — private, pre-revenue, not disclosed for any $-segment. The meaningful segmentation is by end-market wedge:

The strategic read: defense and data centers are the right first wedges — both are price-insensitive relative to reliability and both have the fully-burdened-cost umbrella that lets a $10–15M microreactor clear. The disaster-relief / off-grid-village framing is TAM-marketing; those buyers are the most price-sensitive and least bankable. Judge Radiant on defense + data-center conversion, not the $600B slide.

---

Phase B — Measure performance

Lens 5 · Funding & Valuation Trajectory (private swap for "Earnings Result")

Round history (all ``, unaudited):

• Post-money valuation: ~$1.8B+ (Series D, Dec 2025). This is the only hard valuation anchor.
• Total venture raised ≈ $525M.

Burn read. 311 employees + a ~$280M factory + a fueled test campaign → this is a capital-furnace. The back-to-back $165M then $300M raises six months apart signal a burn that requires near-continuous fundraising; a single missed milestone in a cooling sector would make the next round painful. Runway is a function of test success, not the balance sheet.

Lens 6 · Narrative & Founder Sentiment (private swap for "Earnings Calls")

No earnings calls. The tape here is founder communication + milestone cadence, and the trend is strongly, consistently on-message and accelerating:

• "Anti-stealth" is the deliberate posture — Radiant publishes design detail and milestone dates most nuclear startups hide (Not Boring / Packy McCormick profile; NC State seminar; blog cadence). The tone is SpaceX-transplant: hardware-rich, deadline-driven, "make nuclear boring/portable."
• Sentiment shift over 2023→2026: from "can a SpaceX EE really build a reactor?" skepticism (Forbes 2023) → establishment validation (DOME selection, Baranwal hire, DOE authorizations, NRC accepting the factory license). The things they now say — "first to test in DOME," "fuel on site," "mass-produced" — are milestone claims, not vision claims. The thing to watch them stop saying: the original "criticality by July 4, 2026" deadline (missed — see Lens 8/10).
• Risk in the narrative: anti-stealth transparency raises the cost of a slip — every public deadline is a public scoreboard. That is a double-edged moat.

Lens 7 · Cap Table, Secondary Marks & Public-Comp Cross-Check (private swap for "Comps")

Syndicate quality (the +private tell). Genuinely tier-1 and broadening: a16z, Founders Fund, DCVC, Union Square Ventures, Draper (classic deep-tech VC) + strategics (Chevron Technology Ventures = energy major; Hanwha = Korean defense/energy conglomerate; Equinix = customer-investor) + crossover-adjacent (StepStone private-markets, ARK Venture Fund = Cathie Wood's retail-facing private vehicle). What's missing is the classic IPO-proximity crossover (Fidelity / T. Rowe / Coatue). Read: the cap table is maturing toward public-market readiness but is not yet signaling an imminent S-1 — the strategics and ARK are the "we want public exposure to this" tell, without the mutual-fund crossover that usually precedes a filing by 6–12 months.

Public-comp cross-check (peers are ``; there is no private secondary mark for Radiant beyond the $1.8B primary):

Relative-value read (opinionated): Radiant's $1.8B private mark sits between Nano ($1.4B) and NuScale ($5.9B) and at ~15% of Oklo ($10–12B) — yet Radiant arguably has the most tangible near-term hardware milestone of the group (an actual fueled full-power test in a DOE bed). Two mutually exclusive conclusions, and which one is right is the trade: (a) Radiant is cheap vs. Oklo's pre-revenue euphoria and should re-rate up on a successful test; or (b) the entire cohort is a momentum bubble (Oklo at ~3.3x book with $0 revenue) and Radiant's private mark will compress with the sector when the first FOAK cost/schedule reality lands. No multiple can honestly be applied to $0 revenue — EV/Sales, P/E, ROE = n/a / not meaningful. The comp is a staging comp (how far along, at what mark), not a valuation multiple.

Lens 8 · Perception Catalysts (the events that re-rated the story)

Radiant is private, so "the tape" is funding/milestone step-ups. The events that moved perception (and the private mark):

• 2023 Series B ($40M, a16z) — first tier-1 validation that a SpaceX EE could do nuclear.
• 2024 DOME selection (first design to test in DOME) — the credibility inflection.
• Aug 2025 Equinix 20-unit pre-order — first marquee commercial demand.
• Aug 2025 USAF/DIU agreement — bankable government demand + regulatory tailwind.
• Feb 2026 DOE DARK / PDSA approvals — first full-power test authorization in the program.
• Apr 1, 2026 DOME possession; May 5, 2026 NRC accepts R-50 Part 70 factory license (expedited, ~8-month review).
• Jul 1, 2026 first TRISO fuel on site — cleared the path to full-power testing.
• The one that cut the other way: Radiant did not achieve criticality by the July 4, 2026 deadline — four peers (Aalo, Antares, Valar, Deployable Energy) did; Radiant only received fuel. The pattern the market reacts to: authorizations and hardware milestones re-rate it up; a missed public deadline is the first real dent in the anti-stealth scoreboard.

Traction / unit-economics add (/, unaudited): no per-unit price disclosed. Sector capex benchmarks: microreactor $5,000–25,000/kW; Oklo's NRC filing cites ~$6,700/kW for a 1.5 MWe unit. → a 1 MWe Kaleidos at $7–15k/kW ≈ $7–15M/unit installed capex. Competitive PPA reference: Last Energy $130–200/MWh; solar ~$50/MWh; FOAK SMR LCOE $80–150/MWh, NOAK target $50–80/MWh (with PTC). Radiant only clears where the alternative is diesel at $200–1,000+/MWh — i.e., the defense/remote wedge — not on-grid vs. solar.

---

Phase C — Judge people & books

Lens 9 · Management

• Doug Bernauer — CEO / co-founder. Case Western EE; ~11–12 yrs at SpaceX (joined 2007/08) as R&D engineer — avionics for Grasshopper (first reusable-rocket legs), plus Hyperloop/Boring side projects. Discovered the microreactor concept while scoping Mars-colony power; pivoted to DoD's 2019 portable-microreactor RFP for a nearer, bankable customer. Archetype: founder-engineer, hardware-first, deadline-culture transplant from SpaceX. Strength: velocity, transparency, first-principles. Watch: no prior nuclear-licensing or manufacturing-at-scale track record — he's buying that in.
• Bob Urberger — CTO / co-founder. ex-SpaceX (Grasshopper structural).
• Dr. Rita Baranwal — Chief Nuclear Officer (Jun 2025). ex-DOE Assistant Secretary for Nuclear Energy (2019–21), ex-Westinghouse AP300 CTO/SVP, ex-EPRI, ran DOE's GAIN. This is the single most important hire — it converts a SpaceX startup into a licensable nuclear operator, and she came from the incumbent (Westinghouse eVinci's parent).
• Ben Betzler — Dir. Nuclear Engineering (ex-Oak Ridge National Lab); Roger Chin — Software Architect (ex-SpaceX); Steve Burns — board (ex-NRC Commissioner).
• ~311 employees (latest).

Capital allocation / skin in the game. Founders hold meaningful equity (typical for a Series-D founder-led hardware co; exact % n/a — private, not disclosed). Capital allocation so far is concentration into one bet — Kaleidos + the R-50 factory — which is the right focus but leaves no diversification if the design or fuel disappoints. No buybacks/dividends (pre-revenue). Red flags: none material found — no related-party, promotional, or governance issues surfaced in public sources. The only judgment-watch item is founder-vs-professional-manager: a first-time nuclear operator scaling manufacturing, mitigated (not eliminated) by the Baranwal/Burns bench.

Lens 10 · Forensic Red Flags + Regulatory

Accounting/forensic: n/a — private, no audited financials, no SEC filings. There is nothing to forensically test (no income statement, no cash-flow-vs-earnings divergence, no receivables/inventory, no SBC disclosure). The honest red flag is that absence: a would-be secondary/pre-IPO buyer is underwriting unaudited numbers and self-reported milestones. Standard institutional-diligence items (cash runway, burn rate, contract terms, deposit refundability on the Equinix/USAF pre-orders) are not public — treat all as unverified.

Regulatory findings (required sub-section):

• SEC (LR / AAER): None — Radiant has no CIK and is not an SEC filer; no EDGAR enforcement search is possible.
• Non-SEC (FTC/DOJ/FDA/DOE/NRC enforcement): No material adverse actions found via web search. To the contrary, the regulatory posture is a positive: DOE DARK/PDSA authorizations (Feb 2026) and NRC acceptance of the R-50 Part 70 license application (May 5, 2026; decision targeted Dec 18, 2026, ~55% faster than standard).
• Item 3 / Legal Proceedings: n/a — no 10-K exists (private).
• The one genuine regulatory risk (not an enforcement finding): the Part 70 license is for the factory (special-nuclear-material handling) — it is not a commercial reactor operating license. The DOME test runs under DOE authorization (a deliberate regulatory arbitrage that lets Radiant test now without NRC reactor licensing). Commercial deployment on non-federal sites will require NRC reactor licensing — a separate, slower, unproven-for-microreactors path. Also noted: a 2025 Wyoming siting project was withdrawn amid local political opposition — a data point on state/local permitting friction.
• Summary: No material regulatory or legal findings — verified via SEC EDGAR EFTS (LR/AAER, zero — no CIK), web search (no FTC/DOJ/DOE/NRC actions), and the absence of any 10-K, as of 2026-07-10. The regulatory exposure is prospective (commercial reactor licensing), not a live enforcement matter.

---

Phase D — Project & stress-test

Lens 11 · IPO-Readiness & Path-to-Tradeable (private swap for "Forward Projection")

No private-watch.json entry exists for Radiant — recommend creating one (stage, ipo_readiness, catalyst, and this dossier's path) so privates.ts shows the name dossier-warm. My fresh assessment:

• Stage: late Series D (~$1.8B post, Dec 2025), pre-revenue, first fueled test in progress (2026), commercial delivery targeted 2028, factory under NRC review.
• Milestones that unlock an S-1 / tradeable event (in order):

1. Zero-power criticality at DOME — imminent, pending (missed the July 4 2026 deadline; expected Q3 2026). This is the binary go/no-go.
2. Full-power / full-heat / 150-hr unattended operation (over the one-year DOME campaign, ~through Apr 2027).
3. NRC R-50 Part 70 factory license decision (targeted Dec 18, 2026).
4. R-50 factory construction + first customer/DoD delivery (2028).

• IPO window / path-to-tradeable: Two realistic paths. (a) Opportunistic sector-IPO: precedent is strong — Oklo (SPAC) and Nano Nuclear (IPO) both listed pre-revenue — so a successful criticality test in a hot sector could support a 2026–2027 listing well before commercial revenue. (b) Fundamentals-gated: wait for the 2028 factory ramp. My read: path (a) is live but not imminent — the missing crossover investors (Fidelity/T.Rowe/Coatue) argue against an S-1 in the next ~2 quarters; a clean full-power test + the Part 70 license by Dec 2026 is the more likely trigger for either a crossover round or a listing.
• Value framing (not a price target — no tradeable security): at $1.8B today, a successful full-power DOME test + Part 70 license could support a $3–5B next-round/IPO mark; a failed or materially delayed test re-rates toward Nano's ~$1.4B or below. No EPS projection is possible (pre-revenue) — EPS FY26/27/28 = n/a — pre-revenue.
• No Brier forecast logged (per --watchlist rule; and no honest binary I'd stake without confirming the exact criticality date). The natural scoreable claim to log later: "Radiant Kaleidos achieves zero-power criticality at DOME before 2026-12-31" — I'd seed that ~0.80 given fuel is already on site, but I am not committing it in this unattended pass.

Lens 12 · Bull vs Bear

Bull case. Radiant is the cleanest execution proxy in portable nuclear: it has fuel in a licensed DOE test bed (most peers are still on paper or at zero-power), an A-team that converts a SpaceX hardware culture into a licensable nuclear operator (Baranwal/Burns), a genuinely differentiated no-water, factory-swap, container form factor, and it plays in the one market where microreactor economics unambiguously win — diesel displacement at defense/remote sites where the alternative is $200–1,000+/MWh. The recurring fuel-swap + SimEngine annuity is a real margin engine if the fleet scales. If the factory learning curve is even half of SpaceX's, NOAK unit costs fall into a defensible $50–80/MWh and the ~$600B off-grid TAM is not fantasy. At $1.8B — a fraction of Oklo's mark — the risk/reward on a successful test is asymmetric to the upside.

Bear case (2–3 permanent-impairment risks).

1. Fuel. The thesis is hostage to HALEU/TRISO — single-sourced through pre-revenue Standard Nuclear into a ~28 MT national shortfall by 2030. No fuel at scale = no fleet, regardless of how good the reactor is. This is the risk that can permanently cap the business.
2. FOAK cost/schedule. Nuclear's history is optimistic estimates meeting reality: NuScale's CFPP went $5.3B→$9.2B then cancelled. A 2028 factory delivering 50/yr at competitive cost is a bet against the entire industry's track record.
3. Sector de-rating. Radiant's mark is set by a market that has Oklo at ~$10–12B on $0 revenue. If any peer's FOAK slips publicly, the whole cohort — Radiant's private mark included — compresses.

Pre-mortem (18 months out, thesis broke): Criticality slipped into 2027 on a fuel-qualification or test issue; the sector cooled after a high-profile peer cost-overrun; Radiant's Series E was a flat/down round; the Equinix/USAF pre-orders proved non-binding when timelines moved. Every element of that story is plausible from today's facts.

Are multiples too high? There are no multiples — but the staging mark ($1.8B pre-criticality, pre-revenue) prices in successful execution. It is not a value entry; it is a momentum entry.

Contrarian view (what the market refuses to see): The bull crowd fixates on the $600B off-grid/data-center TAM. The real, bankable, near-term business is a narrow, unglamorous one — displacing diesel at defense and remote industrial sites — and that niche alone can justify the company if (and only if) the fuel supply exists. The market is mispricing which story matters: fuel and defense logistics, not AI-data-center slides.

Lens 13 · Devil's Advocate (short-seller)

Dismantling the bull case:

• Structural break-point = fuel, not physics. Even a flawless reactor is worthless without HALEU/TRISO at volume. Radiant controls neither enrichment (Centrus, nationally short) nor fabrication (Standard Nuclear, a single pre-revenue vendor). One vendor stumble or one DOE reallocation and the 2028 ramp is fiction.
• Revenue concentration is zero real revenue dressed as a backlog. The "order book" is 20 Equinix units (deposit, not delivery) + one USAF demo agreement — both contingent on a working reactor and both cancellable if timelines slip. There is no take-or-pay, no delivered unit, no priced PPA.
• The moat bulls overrate: HTGR/TRISO is not proprietary. BWXT (Project Pele, 1.5 MWe HTGR) is the most dangerous competitor bulls underestimate — a multibillion-dollar naval-reactor incumbent that already builds reactors for the U.S. government and has also delivered fuel to INL. Westinghouse (eVinci, and Baranwal's former employer) has balance-sheet and utility relationships Radiant can't match.
• Capital-allocation risk: back-to-back $165M/$300M raises six months apart = a burn that requires permanent access to capital. In a cooling sector, the next round is dilutive or flat/down — and there's no revenue to self-fund.
• The missed deadline is a tell. Radiant made "criticality by July 4, 2026" a public commitment (anti-stealth) and missed it while four peers hit it. If the first public reactor deadline slips, why underwrite the 2028 factory date?
• What must hold for $1.8B: flawless 2026 test, HALEU at scale, a 2028 factory at competitive cost, binding conversion of pre-orders, and a sector that stays hot. If growth/timeline disappoints 20–30%, the mark compresses toward Nano's ~$1.4B or below, and a down-round resets the cap table.
• Single scenario that permanently impairs: a fuel-qualification or in-test safety event at DOME that forces a redesign — it would blow the schedule, spook capital mid-burn, and hand the defense contracts to BWXT/Westinghouse. Plausibility: low-to-moderate, but non-trivial for a first fueled test of a new design.

Lens 14 · Fifteen Questions for the CEO (ordered by information value)

1. HALEU/TRISO: What is your contracted fuel supply through 2030 in kg, who are the counterparties beyond Standard Nuclear, and what is your second source if Standard Nuclear's fabrication or the national HALEU supply slips?
2. Criticality: You received fuel July 1 but did not hit the July 4 criticality deadline — what is the specific gating item, and what is your confidence on zero-power criticality before year-end 2026?
3. Unit economics: What is your target installed capex per unit and PPA $/MWh at NOAK (50/yr), and what learning-curve slope must the R-50 factory achieve to get there?
4. Order book quality: Are the Equinix 20 units and the USAF agreement binding with non-refundable deposits and priced delivery, or framework commitments — and what triggers cancellation?
5. Commercial licensing: Part 70 covers the factory; what is your path and timeline to an NRC reactor operating license for non-federal commercial sites, and does your model depend on federal-site/DOE-authorization deployments to reach scale first?
6. Runway: What is current monthly burn, cash on hand post-Series D, and how many milestones does it fund before you must raise again?
7. BWXT/Westinghouse: BWXT (Pele) and Westinghouse (eVinci) are HTGR/microreactor peers with government relationships and balance sheets — what is your durable advantage against incumbents, specifically on defense procurement?
8. Factory: What are the long-lead items, qualified-supplier gaps, and workforce constraints on standing up ~50 units/yr by 2028, and what's your confidence interval on that date?
9. Spent fuel / back-end: Who owns spent-fuel liability under a PPA, and how is factory-swap logistics + waste handling priced and regulated across states?
10. Margin architecture: What share of lifetime unit value is hardware vs. the recurring fuel-swap/SimEngine annuity, and what gross margin do you underwrite on each?
11. Defense scaling: Beyond Buckley SFB, what is the realistic base-count and revenue if the ANPI program scales — and how exposed is that to an administration/budget change?
12. Water-free CO₂ Brayton cycle: What efficiency and reliability data do you have on the closed-loop supercritical-CO₂ conversion at 1 MWe, and what are the failure modes in field conditions?
13. Talent: How do you retain nuclear-licensing and manufacturing expertise (Baranwal/Betzler) against incumbents, and what's the bench if a key hire leaves?
14. Liquidity for holders: What is your realistic path and timeframe to a tradeable event (IPO/secondary), and what milestones gate it?
15. The bear: Which single risk keeps you up at night — fuel, FOAK cost, or licensing — and what have you already done to de-risk it?

---