The Infinite Unknown

Overview

Radiant Nuclear is developing the Kaleidos microreactor, a factory-built, modular high-temperature gas-cooled reactor (HTGR) designed for distributed deployment at data center colocation facilities. At 1.2 MWe per unit, Kaleidos is small enough to fit on a single truck, yet can be clustered to scale from 2.4 MW (two units) to 10+ MW (eight+ units) without significant site infrastructure.

Technology: Kaleidos HTGR Design

Core Reactor Physics

Kaleidos is a pebble-bed or prismatic-block high-temperature gas-cooled reactor (HTGR variant), operating at ~700–800°C helium outlet temperature:

Design features:

• Helium coolant: Inert, excellent heat transfer properties, no water corrosion concerns
• TRISO fuel: Tristructural Isotropic coated particles (uranium oxycarbide kernel surrounded by ceramic layers) — thermally and mechanically robust; can tolerate 1200°C+ peak temperatures
• HALEU fuel: Low-enriched uranium (typically 10–20% U-235), higher density than LEU but less restricted than weapons-grade HEU
• Passive cooling: No active pumps required in emergency; helium natural circulation cools the core to ambient via conduction through vessel walls
• Walk-away-safe: Core can withstand full loss of coolant and full loss of heat removal without melting — lowest decay heat, higher surface-to-volume ratio

Factory Manufacturing Advantage

Unlike large LWRs (built on-site over 5+ years), Kaleidos units are factory-constructed and shipped:

1. Modular construction: Reactor vessel, steam generator, control systems all assembled in a factory (Radiant’s planned Albuquerque-area facility or partner fab)
2. Quality control: 100% inspection and testing in controlled environment before shipment
3. Transport: Modules fit in standard truck containers or rail cars; no custom transport infrastructure required
4. Site installation: Plug-and-play setup; minimal on-site welding or heavy fabrication (weeks to months vs. years)
5. Scaling: Multiple units shipped to same site, assembled in parallel, commissioned independently or linked

Timeline advantage: Factory + transport + site assembly takes 12–18 months from order to power generation, vs. 36–48 months for on-site constructed SMRs.

Modular Scaling Strategy

Radiant’s key innovation is scalability through clustering:

Advantage: Equinix (or any colocation player) doesn’t have to commit to a single large reactor; can deploy 1–2 units initially, add capacity incrementally as site grows. This matches the modular, incremental capex of data center buildout.

Market Opportunity: The Equinix Deal

Equinix 20-Unit Preorder (August 2025)

Announced: August 2025 Capacity: 20 × 1.2 MWe = 24 MW+ deployed across multiple Equinix facilities Timeline: Phased deployment 2027–2030+ Commercial terms: Binding preorder with supply and deployment schedule milestones

Significance:

1. Largest microreactor order to date: 24 MW exceeds any single company’s microreactor commitment
2. Colocation validation: Equinix operates 260+ data centers globally; 20-unit deployment spans multiple geographies and customers
3. Production signal: Equinix’s order provides production volume certainty, enabling Radiant to scale manufacturing
4. Competitive positioning: Signals that colocation players (not just hyperscalers like Meta/Google) are betting on microreactors

Strategic context:

• Equinix generates revenue from customer power consumption; on-site nuclear BTM increases customer stickiness and SLA guarantees (99.99%+ uptime)
• 20-unit order diversifies Equinix’s BTM portfolio (also evaluating Oklo SMRs, Bloom fuel cells, geothermal)
• Incremental deployment model matches Equinix’s geographic expansion strategy

Technology Validation: KDU Demonstration

Kaleidos Demonstration Unit (KDU) at Idaho National Lab

Location: INL DOME facility (Dome Expansion Module — new experimental hall) Unit size: 1.2 MWe / 5 MWth (full-scale commercial module) Target startup: 2026 (as of this update; some slippage likely) Duration: Minimum 18–24 months continuous operation

Key milestones:

• Fuel loading: Completion of HALEU TRISO fuel verification, loading into reactor
• Criticality: First chain reaction (confirms physics models)
• Power ramp: Step-wise increase from 10% to 100% power; validate thermal-hydraulics, control systems
• Load-following tests: Variable power output to simulate data center demand response
• 24/7 operation validation: Minimum 1,000-day uninterrupted operation to demonstrate reliability

Importance: KDU is the make-or-break milestone for Radiant. Successful operation validates:

• TRISO fuel thermal performance under HTGR conditions
• Helium loop integrity and heat transfer
• Modular factory-built assembly and commissioning process
• Control system responsiveness and safety systems
• Data generation for NRC operating license applications

Risk: INL DOME facility has experienced delays due to infrastructure constraints. KDU startup could slip from target 2026 to 2027–2028.

Regulatory & Licensing Path

NRC Pre-Application Review

Radiant is in pre-application phase with NRC:

• Design Certification Application (DCA): Target filing 2026–2027 (non-specific to any plant)
• Standard Design Certificate: If approved, greatly accelerates individual project licensing
• KDU DOME license: Experimental reactor license; lower regulatory burden than commercial COLAs

Equinix Site-Specific Licensing

For Equinix deployments, Radiant will pursue:

• Site-specific Combined License Applications (COLAs): One per facility or cluster of facilities
• NRC review: 18–24 months per COLA (standard path for established reactor designs; KDU validation + DCA may shorten this)
• State/local siting: Equinix facilities in various states; each may require state environmental review

Advantage vs. larger SMRs: Smaller reactor (1.2 MW) often faces less local opposition and simpler site reviews. Many Equinix colocation sites are industrial parks with existing infrastructure and permitting precedent.

Manufacturing & Supply Chain

Production Facility Plan

Target: 50 Kaleidos reactors/year by late 2020s Location: Primary fab TBD (likely Albuquerque area or partnership with existing reactor OEM) Capex for fab: Estimated $200–400M (factory, tooling, quality systems) Ramp timeline: Pilot production 2027–2028 (10–15/year); scale to 50/year by 2029–2030

Partners:

• Fuel supplier: BWXT (TRISO) — shared bottleneck with Kairos, X-energy, USNC
• Pressure vessel fabrication: ASME-N stamped shops (heavy industrial suppliers)
• Steam turbine/generator: GE or Siemens (OEM surplus or custom small-scale units)
• Instrumentation & control: Conventional DCS suppliers (Siemens, Emerson, Foxboro)

HALEU & TRISO Fuel Supply

Risk: TRISO fuel is a shared constraint across Kaleidos, Kairos Hermes, X-energy, and USNC KRONOS:

• BWXT is the sole U.S. TRISO supplier
• Production capacity: ~500–1,000 kg TRISO/year as of 2025; expanding to support SMR + microreactor pipeline
• Radiant’s need: 20 units × 1.2 MW requires ~500 kg TRISO per year for supply + contingency
• Competitive pressure: If Kairos deploys 6–7 units (Google partnership) and X-energy scales, BWXT capacity may be exceeded by 2028–2030

Mitigation:

• Radiant is funding BWXT capacity expansion (co-investment)
• Exploring alternative TRISO sources (international suppliers, e.g., France, Japan)
• Possible fuel swap agreements with competitors (e.g., Radiant delays some units if Kairos needs priority fuel supply)

Timeline concern: TRISO supply may limit Equinix deployment to <10 units by 2030; full 20-unit order likely extends to 2032–2035.

Competitive Landscape

vs. USNC KRONOS MMR (15 MWe)

Advantage: Radiant’s modularity and Equinix order provide credibility and production certainty. USNC bankruptcy and acquisition by NANO (mid-stage company) raised questions about KRONOS viability.

Disadvantage: Smaller unit (1.2 MW) means higher per-MW capital cost if not mass-produced. KRONOS’s larger 15 MW size could offer better economics if successfully deployed.

vs. Westinghouse eVinci (5 MW Heat-Pipe Microreactor)

Advantage: Kaleidos has proven data center market demand (Equinix) and factory-manufacturing experience. eVinci is larger unit, simpler cooling design, but unproven at commercial scale.

Watch: Westinghouse could rapidly scale eVinci if Penn State prototype succeeds and industry adopts it as standard. But as of April 2026, Kaleidos has stronger market validation.

vs. Oklo Aurora & Kairos Hermes (SMRs, Baseload Focus)

Radiant’s niche: Colocation players and mid-scale data center operators who want 1–5 MW increments, not 50–345 MW units. Equinix’s 20-unit strategy is exactly this model.

Capex & Economics

Capital Cost Trajectory

Unit 1 (KDU): Estimated $60–100M (prototype, higher cost) **Units 2–5 (pilot production):** ~$50–70M per unit Units 6–20 (scaling): ~$40–60M per unit (target with 50/year production) **Units 21+ (mature production):** ~$30–50M per unit (if 50+/year achieved)

$/MW capital cost:

• Early units: $42–83M/MW (high per-MW cost due to small unit size)
• Mature production: $25–50M/MW (if capital cost targets hit and volume ramps)

Comparison:

• SMRs (Oklo, Kairos): ~$2.5–4.0M/MW all-in
• Large gas turbines: ~$1.5–2.0M/MW
• Bloom fuel cells: ~$2–3M/MW

Assessment: Kaleidos per-MW capex is higher than SMRs or gas turbines at modest scales (1–5 MW). However, for colocation sites with existing infrastructure, avoidance of large-scale site work, permitting delays, and regulatory queues may justify the premium. Also, Equinix’s revenue-per-MW-deployed is high; colocation economics differ from hyperscaler BTM.

Levelized Cost of Power (LCOP)

Assumptions:

• Capex: $45M/unit × 1.2 MW = $37.5M/MW (mature production)
• Debt: 70% @ 5% interest
• Equity: 30% @ 12% return
• Capacity factor: 85% (data centers run ~24/7, slightly lower than grid baseload)
• O&M: $500–800K/unit/year (~$0.4–0.7M/MW/year)
• Fuel cost: Negligible (HALEU once-through fuel)
• Decommissioning reserve: ~$10M/unit amortized

Estimated LCOP: $60–90/MWh (2026$)

Context:

• Grid wholesale power (2025–2026): $30–60/MWh (varies by region)
• Bloom fuel cells (running on NG): $80–120/MWh levelized
• Solar + battery (100-hour storage): $100–150/MWh levelized
• Hyperscaler’s grid interconnection + transmission: $40–80/MWh (including transmission charges)

Interpretation: Kaleidos’s LCOP is competitive with grid + transmission + reliability premium in many data center markets, especially if colocation operators are willing to pay premium for 24/7 carbon-free + guaranteed availability.

Risks & Challenges

⚠️ KDU Delay Risk

Current target: KDU startup 2026 Realistic expectation: Slippage to 2027–2028 is likely (INL DOME infrastructure constraints, NRC review bottlenecks) Impact: Each 6-month delay pushes commercial production timeline back 12+ months (licensing relies on KDU data)

⚠️ TRISO Fuel Supply Bottleneck

Kaleidos uses HALEU TRISO, shared with:

• Kairos Hermes (50 MW Google contract)
• X-energy Xe-100 (5 GW Amazon contract, though distributed)
• USNC KRONOS (16 MW demo pipeline)

BWXT capacity: ~1,000 kg/year as of 2025; demand from all sources could exceed 1,500 kg/year by 2029 → supply crunch.

Mitigation: Radiant funding BWXT expansion, but capital and licensing timelines uncertain. Real risk: Equinix’s full 20-unit deployment extends to 2032–2035 instead of 2028–2030.

⚠️ Per-MW Capex Remains High

At $40–60M/unit, Kaleidos’s per-MW capital cost is 1.5–2.5× higher than gas turbines or Bloom fuel cells. For small colocation facilities (1–2 MW need), capex-per-MW may be prohibitive unless:

• Radiant achieves <$30M/unit through mass production (uncertain)
• Colocation operators accept premium for 24/7 carbon-free + longevity (25+ year reactor life vs. 10–15 year turbine)
• Regulatory/permitting advantages of microreactor (faster licensing, lower local opposition) justify capex premium

⚠️ Colocation Customer Appetite Uncertain

Equinix’s 20-unit order is a preorder, not yet deployed. Key questions:

• Will Equinix customers (hyperscalers, enterprises) actually commit to multi-year 1.2 MW BTM contracts?
• Will Equinix’s colocation customers accept higher power costs if BTM price premium exceeds grid+transmission?
• What happens if hyperscalers (AWS, Azure, GCP) begin building their own colocation-like facilities? Equinix’s BTM value proposition erodes.

Execution risk: Equinix could reduce or cancel 20-unit order if market fundamentals shift 2026–2028.

2026–2030 Milestones

Outlook: Why Radiant Matters for Colocation

Kaleidos is the only microreactor with proven customer demand (Equinix 20-unit preorder). While Westinghouse eVinci and USNC KRONOS are technically sound, they lack market validation.

For colocation operators:

• ✅ Modular scaling (1.2 MW units) matches incremental capex cycles
• ✅ Factory-built shortens deployment timeline (12–18 months vs. 3–5 years for grid interconnection)
• ✅ 24/7 carbon-free power enables “green colocation” marketing
• ⚠️ High per-MW capex ($30–60M/MW) requires premium colocation customer willingness-to-pay
• ⚠️ First-of-a-kind regulatory and licensing complexity

Competitive dynamics:

• If Kaleidos succeeds at Equinix (2028–2029), other colocation players (Digital Realty, CoreWeave, Zenlayer) will follow
• If Kaleidos experiences delays or capital overruns, colocation market will default to Bloom fuel cells + gas turbines (proven, cheaper per-MW)

Key watch: KDU startup (target 2026, likely 2027), Equinix site selection and licensing (2027–2028), and capital cost trajectory for units 2–5 will determine Radiant’s viability.

Key People

Doug Bernauer — CEO & Co-Founder

• LinkedIn: TODO: verify slug
• Prior: SpaceX propulsion and systems engineering; co-founded Radiant Nuclear 2020 to apply aerospace manufacturing discipline to microreactors

Zeke Hausfather — Board / Scientific Advisor

• LinkedIn: public profile; prominent climate scientist; prior Stripe / Berkeley Earth
• Note: The steering summary listed Hausfather as “CEO, board” but primary CEO is Doug Bernauer; Hausfather’s specific role requires verification — TODO: confirm exact title

Eric Ingersoll — Board Advisor

• Former NRC official; deep nuclear licensing expertise
• No confirmed LinkedIn; see Radiant public press releases for bio

People — Last Reviewed: 2026-04-25

Related Profiles

• /research/datacenters/behind-meter-power/oklo-btm/ — Oklo Aurora SMR (75 MW, largest order book)
• /research/datacenters/behind-meter-power/kairos-power-btm/ — Kairos Hermes SMR (50 MW, molten-salt cooled, Google partner)
• /research/datacenters/behind-meter-power/westinghouse-evinci/ — Westinghouse eVinci microreactor (5 MW, heat-pipe design)
• /research/datacenters/behind-meter-power/usnc-microreactor/ — USNC KRONOS MMR (15 MW, post-bankruptcy)
• /research/datacenters/behind-meter-power/terrapower-btm/ — TerraPower Natrium (345 MW, thermal storage, Meta partner)