The Infinite Unknown

Summary

Kairos Power, an Albuquerque, New Mexico advanced reactor developer, is pioneering molten-salt-cooled small modular reactors (SMRs) explicitly designed for data center and industrial heat applications. The company’s Hermes design produces ~50 MWe at the core, with the unique capability of operating at high temperatures (650–750°C) that enable efficient electricity generation plus process heat for hydrogen production, desalination, or industrial processes.

In August 2025, Google signed the first corporate SMR power purchase agreement with Kairos Power: a binding commitment for 500 MW across 6–7 Hermes reactors, with the first unit (Hermes 2) targeted for operation at Oak Ridge, Tennessee by 2030, and full fleet deployment by 2035. This binding PPA (not just an LOI) is the most credible SMR datacenter signal to date.

Key appeal: Molten-salt technology enables higher operating temperatures than water-cooled reactors, unlocking process heat applications beyond just electricity; Google’s binding commitment provides validation and capital cushion; Hermes design is simpler than liquid metal coolers (Oklo, TerraPower).

Risks: TRISO fuel supply (shared bottleneck with X-energy, USNC); molten-salt operation unproven at commercial scale; Google’s first-of-a-kind PPA terms and pricing not disclosed (potential precedent risk for other hyperscalers).

⚑ Major Update: Hermes 2 Groundbreaking (April 17, 2026)

On April 17, 2026, Kairos Power broke ground on the Hermes 2 Demonstration Plant at Oak Ridge, Tennessee — a milestone with multiple firsts:

• First Gen IV reactor to receive an NRC construction permit in U.S. history
• First commercial-scale advanced reactor under construction in the United States
• First deployment under the Google PPA for 500 MW of SMR power across 6–7 units by 2035

Project scope:

• Plant type: Molten-salt-cooled, TRISO-fueled demonstration plant (Kairos KP-FHR design)
• Output: Up to 50 MWe; power delivered to Tennessee Valley Authority (TVA) grid to serve Google data centers in Tennessee and Alabama
• Manufacturing approach: Reactor equipment modules fabricated at Kairos’s Manufacturing Development Campus in Albuquerque, New Mexico, then shipped to Oak Ridge for assembly — pioneering the factory-built SMR model
• Design basis: Builds on operational data from Hermes 1 (low-power demonstration at INL, which commenced nuclear construction in 2025 and received reactor vessel installation completion)

Operating license: As of February 2026, Kairos is preparing to submit an operating license application for the Hermes reactor (separate from the construction permit). Operating license approval is the final step before commercial power generation.

Sources:

• Kairos Power Breaks Ground on Hermes 2 Demonstration Plant (GlobeNewswire, April 17, 2026)
• World Nuclear News: Kairos breaks ground for Hermes 2 reactor
• Interesting Engineering: First 4th-gen US nuclear reactor breaks ground to power Google data centers

Key Facts

• Founded: 2016 (as Kairos Aerospace, then Kairos Power)
• HQ: Albuquerque, New Mexico (nuclear innovation hub, proximity to INL)
• Type: Private (Series C/D equivalent; funding TBD, likely $200M+)
• CEO: Chris Levesque — MIT Ph.D. nuclear engineering; prior SVP of TerraPower; deep SMR credibility
• Investors: Breakthrough Energy Ventures (Bill Gates fund), Argonne National Laboratory (R&D partnership), U.S. Department of Energy (funding)

Core reactor design — Hermes:

• Coolant: Molten fluoride salt (FLiNaK or similar eutectic mixture)
• Fuel: TRISO fuel (tristructural isotropic particles) in pebble-bed or block form
• Operating temperature: 650–750°C (enabled by salt coolant; water reactors max ~320°C)
• Output: ~50 MWe per unit (can be scaled to larger configurations)
• Heat recovery: Molten salt enables direct power cycle (gas turbine or Brayton cycle) + process heat extraction
• Safety: Passive cooling (molten salt has high heat capacity, enabling natural circulation)
• Fuel burnup: ~180 GWd/MT (high burnup, reduced fuel consumption)
• Refueling interval: ~2–3 years (frequent replacement, manageable from security standpoint)

Test reactor — Hermes 1:

• Located at Idaho National Laboratory (DOME facility)
• Startup target: 2027
• Purpose: Validate molten-salt operation, licensing basis, TRISO fuel performance
• ~10 MW thermal (small-scale demonstration)

Commercial reactor — Hermes 2 (first Google unit):

• Location: Oak Ridge, Tennessee (Kershaw site, partner: Tennessee Valley Authority)
• Target operational date: 2030
• Output: ~50 MWe
• Customer: Google (binding PPA for power + thermal energy recovery)

Google’s 500 MW Binding PPA: Game-Changer for SMRs

Announcement Details (August 2025)

Google announced a binding Power Purchase Agreement with Kairos Power for:

• Total capacity: 500 MW across 6–7 Hermes reactors
• Timeline: First unit (Hermes 2, Oak Ridge) operational by 2030; full fleet by 2035
• Power price: Not disclosed (estimated $50–80/MWh based on typical SMR PPAs)
• Thermal power: Google also expects to utilize process heat from reactors (not detailed)

Why Google (and why binding)?

Google’s datacenter strategy:

• Google committed to “24/7 carbon-free energy” across all datacenters by 2030
• Nuclear power (existing plants + new SMRs) is critical to achieving this goal
• Existing plants alone (Three Mile Island for Microsoft, others) insufficient for full fleet demand
• SMRs provide distributed, scalable, carbon-free baseload

Why Kairos over other SMR vendors?

• Hermes design maturity: Kairos has been developing since 2016; design more advanced than newer entrants
• High-temperature capability: Process heat for hydrogen/desalination aligns with Google’s sustainability goals
• Oak Ridge siting: TVA partnership and federal site reduce permitting friction
• Binding commitment: Google willing to sign PPA (not just LOI), signaling confidence

Binding vs. LOI significance:

• Binding PPA means Google has committed capital and regulatory buy-in
• Kairos can now use PPA as collateral for debt financing, attracting capital partners
• LOI (like Oklo’s) is nonbinding; PPA is binding 20–30 year contract

Financial Implications

For Kairos:

• $500 MW PPA provides revenue visibility for 30+ years
• Enables debt financing (investment-grade lender confidence) + equity raises
• First commercial reactor likely $4–6B capex; PPA supports financing structure

For Google:

• Locks in 500 MW at negotiated price (no escalation risk if technology costs decline)
• Builds strategic supply of nuclear capacity (diversifies from grid dependence)
• Supports marketing: “Google Cloud powered by nuclear” narrative

Technology Deep-Dive: Molten Salt Advantages for Data Centers

Why Molten Salt Enables Process Heat?

Conventional water-cooled reactors (LWRs):

• Operate at ~320°C max (pressurized water limits temperature)
• Thermal efficiency: ~33–35% (large temperature-to-waste-heat ratio)
• Heat recovery: Limited to space heating, desalination (low-grade heat)

Kairos molten-salt reactor:

• Operates at 650–750°C (no pressure constraint; salt is stable at high temperature and atmospheric pressure)
• Thermal efficiency: ~45–50% (better temperature-to-waste-heat ratio)
• Heat recovery: High-grade heat enables hydrogen production, high-temperature industrial processes

Data center process heat applications:

1. Hydrogen production: Electrolysis (powered by reactor electricity) + high-temp steam (from reactor exhaust) enables efficient hydrogen synthesis. Hydrogen is critical for green synthetic fuels, ammonia production, and future energy storage.
2. Desalination: High-temp water enables reverse osmosis or multi-effect distillation. Critical for water-scarce data center regions (Arizona, California).
3. District heating/cooling: Excess thermal energy can serve adjacent industrial parks, reducing total site energy consumption.

Google’s angle: If Kairos Hermes 2+ reactor can supply 50 MW electricity + 100+ MW thermal, Google could:

• Power datacenters with nuclear electricity
• Use heat for on-site hydrogen production (green hydrogen for future fuel cell vehicle fleet or power generation)
• Enable “zero-waste” energy campus model

Test Reactor (Hermes 1) Status & Timeline

Hermes 1 @ Idaho National Lab

Location: DOME (Demonstration of Microreactor Experiments) facility, INL Capacity: ~10 MW thermal (small demonstration) Schedule:

• Construction: 2024–2026
• Startup: 2027 (target; subject to NRC review and commissioning delays)
• Operation: 2027–2032 (expected 5-year demonstration)

Purpose:

• Validate molten-salt thermal-hydraulics at commercial design scale
• Demonstrate TRISO fuel performance in molten-salt environment
• Generate licensing data for NRC (supports Hermes 2 COLA)
• Train operations staff (future commercial plant teams)

Risk: INL infrastructure is shared; if NRC prioritizes other advanced reactor demonstrations, DOME facility access could slip, pushing Hermes 1 startup to 2028–2029.

Hermes 2 @ Oak Ridge — Google’s Commercial Deployment

Oak Ridge Site & TVA Partnership

Location: Oak Ridge, Tennessee (Kershaw site, near existing ORR nuclear facilities) Partner: Tennessee Valley Authority (TVA)

• TVA operates existing reactors in region (Sequoyah, Browns Ferry, others)
• Provides grid connection, land, regulatory support
• TVA has authority to license new reactors under unique federal framework

Regulatory pathway:

• TVA has exclusive state regulatory authority for nuclear power (not subject to Tennessee Public Utilities Commission)
• This is major advantage over states requiring PUC approval (CA, NY)
• Hermes 2 can proceed with TVA blessing alone

Hermes 2 Timeline & Critical Path

Phase 1: Design & Licensing (2025–2027)

• COLA Phase 1: Design-basis engineering (parallel with Hermes 1 testing)
• NRC review: 18–24 months
• Target: COLA Phase 1 approval by 2027

Phase 2: Site preparation & procurement (2027–2028)

• Parallel with COLA Phase 2 (site-specific safety)
• Equipment procurement (long lead times: 18–24 months for reactor vessel, instrumentation)
• Workforce training begins

Phase 3: Construction (2028–2030)

• Modular construction (smaller than conventional plants; 18–24 month on-site duration estimated)
• Commissioning: 12 months

Target operational: 2030 (on track if no major delays)

Hermes 2 Capex & Economics

Estimated capex: $3–4B (first-of-kind; molten-salt unproven at commercial scale)

• Reactor vessel & internals: $500M
• Molten-salt loop components: $400M
• Balance-of-plant: $800M
• Licensing, permitting, contingency: $1–1.5B

Per-MW capex: ~$60–80M/MW (higher than mature SMRs; first-commercial penalty)

Future units (Hermes 3+) likely: $2–3B/unit ($40–60M/MW) as supply chain matures and design optimization reduces cost.

TRISO Fuel Supply: Shared Bottleneck

U.S. TRISO Production Capacity (April 2026)

Current supplier: BWXT (BWX Technologies) — sole U.S. commercial TRISO fuel manufacturer

• Facility: Erie, Pennsylvania
• Current capacity: ~100 MT/year (estimated; not publicly disclosed)
• Kairos requirement (at full scale): ~20–30 MT/year (6–7 reactors)
• Other customers: X-energy (Xe-100), USNC (KRONOS MMR), Radiant (Kaleidos)

Supply constraint implication:

• BWXT capacity is sufficient for 5–10 TRISO-fueled reactors globally by 2030
• Kairos + X-energy + USNC + Radiant = potentially 15–20+ reactors if all scale
• Likely bottleneck: BWXT will need to expand capacity 2027–2029; new capacity comes online 2029–2030

Impact on Hermes 2–7 deployment:

• Hermes 1 (2027 startup): Small demonstration, limited TRISO consumption
• Hermes 2 (2030): First major TRISO load; should be within BWXT current capacity
• Hermes 3–7 (2031–2035): Dependent on BWXT expansion + qualification of new suppliers (unlikely before 2032)

Risk: If BWXT expansion stalls (cost, regulatory, technical), Kairos deployments could slow from 1 unit/year to 1 unit/2 years starting 2031+.

Competitive Position vs. Other SMR/Microreactor Vendors

Kairos competitive strength: Binding Google PPA is a major differentiator vs. Oklo’s LOIs. Google’s credibility + TVA partnership provide faster regulatory path than other SMR vendors.

Regulatory & Financing Path

Federal Support

DOE funding:

• Kairos has received multiple grants for Hermes test reactor and licensing support
• Total DOE backing estimated $200–500M (not fully disclosed)

Loan guarantees: Potential for DOE Title XVII loan guarantees for Hermes 2+ projects (large-scale projects can access $0–10B+ guarantees)

Private Financing

Google strategic investment: Google’s PPA provides equity-like returns; potential for Google to co-invest alongside debt lenders

Project finance debt: 20–30 year PPAs enable investment-grade debt financing (50–70% of capex from debt, 30–50% equity)

Timeline: Hermes 2 financing likely closes 2027–2028 (post-COLA Phase 1 approval).

Key People

• Chris Levesque (CEO) — Ph.D. nuclear engineering (MIT); LinkedIn: not confirmed — TODO: verify slug. Prior: SVP at TerraPower (overlapping with leadership period of other documented company; see TerraPower BTM entry); NRC and DOE engagement experience. ⚑ Overlap: Levesque came directly from TerraPower — the two companies share sodium-cooled SMR development heritage and compete in data center nuclear space.
• Board: Mix of advanced reactor veterans, TVA executives, and Google representatives (post-PPA). Specific names not publicly confirmed as of April 2026 — TODO: verify board composition from Kairos public disclosures.

People — Last Reviewed: 2026-04-25

2026–2027 Critical Milestones

1. ✅ Hermes 2 groundbreaking (April 17, 2026): COMPLETED — construction has begun; first NRC-permitted Gen IV reactor in U.S.
2. Hermes 1 nuclear construction progress (INL): Reactor vessel installation completed; startup sequence proceeding toward 2027 target
3. Hermes 2 operating license application submission: Kairos preparing to submit as of February 2026; submission expected mid-2026; NRC review 18–24 months
4. BWXT capacity planning: Public announcement of TRISO capacity expansion plans (critical for signaling TRISO supply resolved for Hermes 3–7)
5. Kairos financing rounds: Series D/E funding round announced (signals confidence in Hermes 2 timeline, now anchored by construction permit and groundbreaking)
6. First power delivery to TVA (target ~2030): Grid-connected power starts flowing to Google data centers in Tennessee and Alabama

Risks & Uncertainties

1. Molten-salt operational complexity: Molten salt is a proven heat transfer medium (CSP plants, industrial), but reactor-grade molten salt (with neutron activation, tritium handling, corrosion management) is unproven at commercial scale. Hermes 1 is the critical test; failures there could delay Hermes 2.

2. TRISO supply tightness: BWXT must expand capacity on schedule; if delays occur, Hermes 3–7 deployments slow.

3. Google’s PPA pricing terms: Google negotiated pricing + thermal energy terms not disclosed. If disclosed later, competitors (Amazon, Microsoft, Meta) may balk at similar PPA economics.

4. TVA regulatory integration: TVA has unique federal authority, but political shifts or safety concerns could slow approval; TVA stakeholder backlash is a non-technical risk.

Summary: Kairos as Data Center SMR Leader

Kairos Power stands out among SMR vendors due to Google’s binding PPA (not just LOI) and the unique process heat capability that aligns with data center operator sustainability goals (hydrogen, desalination). The company’s molten-salt-cooled Hermes design is technically sophisticated and differentiated from competitors.

Key inflection point: Hermes 1 successful startup (2027) + Hermes 2 COLA Phase 1 approval (2027–2028) will demonstrate regulatory pathway viability. If both succeed on schedule, Kairos could become market leader in SMR-for-datacenters segment, with 500 MW+ deployed by 2035.

Risk: Molten-salt technology is unproven at commercial scale; first-of-kind failures could add 12–24 months to schedule and $1B+ to capex, weakening economics for follow-on units.

Investment thesis: Kairos is medium-risk bet on molten-salt SMR maturation + Google’s commitment to scaled nuclear deployment. Upside: $10B+ valuation if Hermes 2–7 deployment succeeds at scale. Downside: Technical delays or molten-salt corrosion issues could crater timeline.