Overview

TerraPower, backed by Bill Gates and founded by John Deutch, is developing the Natrium sodium-cooled fast reactor specifically for industrial and data center power. The key differentiator is integrated molten-salt thermal energy storage—the reactor can dispatch 500+ MW of power for 5+ hours by drawing down heat stored in a molten-salt tank, enabling true load-following without separate battery infrastructure.

Attribute Value
Technology Sodium-cooled fast reactor (SFR)
Fuel Low-enriched uranium (LEU) metallic fuel
Baseload Capacity 345 MWe
Peak Output (with storage) 500+ MWe for 5+ hours
Thermal Storage Molten salt (enables load-following)
Status Pre-construction (site prep Wyoming)
First Online Target ~2030 (Kemmerer, Wyoming)
Data Center Partnerships Meta (8 units, 2.76 GW); Sabey Data Centers (MOU)

Technology: Natrium Reactor + Storage System

Core Reactor Design

The Natrium reactor operates at 545°C in a pool-type configuration with sodium coolant circulating through the core and three independent cooling loops—a design inherited from Integral Fast Reactor (IFR) research and validated across decades of sodium-cooled reactor operation (EBR-II, FFTF, Monju).

Key design features:

  • Passive safety: No active cooling required; if all power is lost, natural circulation and thermal conduction safely cool the core
  • Compact footprint: Pool design limits external footprint; can be sited on industrial campuses or within large data center sites
  • Fuel efficiency: Fast spectrum breeding allows multi-cycle operation on depleted uranium or LEU stockpiles
  • Metallurgical fuel: Metallic U-Zr fuel (vs. oxide pellets in light-water reactors); higher density, higher burnup capability

Thermal Energy Storage Integration

The distinguishing feature of Natrium is integrated molten-salt thermal storage. At steady state:

  1. Reactor operates at 345 MWe baseload output
  2. Excess heat is directed into a molten-salt storage tank (liquid salt maintained at ~500°C)
  3. When data center demand spikes, the salt discharges its heat to a steam generator, boosting turbine output to 500+ MW for 5+ hours without reactor power increase

Storage capacity: Approximately 1 GWh thermal energy, enabling realistic data center load-following without external battery hardware.

Advantage over SMRs without storage:

  • Oklo Aurora (75 MW) and Kairos Hermes (50 MW) cannot follow variable data center loads; they run flat baseload
  • Natrium with storage is truly flexible—can match 24/7 varying compute demand without grid backup
  • Eliminates need for 500+ MWh battery systems (cost: $100–300M for equivalent storage)

Data Center Partnerships

Meta: 8-Unit Natrium Fleet (2.76 GW)

Signed: Binding agreement (spring 2025) Capacity: 8 × 345 MW = 2.76 GW baseload + storage-enabled flexibility Locations: Specific Meta data center sites TBD (likely Wyoming + additional locations) Timeline: First unit (Kemmerer) ~2030; subsequent units phased through 2035

Significance:

  • Largest SMR data center order to date (by GWh)
  • Meta’s diversification strategy: balances Oklo (15 GW LOI), TerraPower (2.76 GW binding), geothermal (Sage, 150 MW), and solar
  • Storage feature aligns with Meta’s variable load profile (model training batches + inference peaks)

Sabey Data Centers: Multi-Site MOU

Announced: 2025 Type: Memorandum of Understanding (nonbinding framework) Target: Wide-scale deployment at Sabey facilities nationwide Status: Early-stage; specific site selection and commitment timelines TBD

Context:

  • Sabey is a mid-tier colocation/hyperscale data center provider (not Meta/Google/Amazon scale)
  • MOU suggests TerraPower exploring sub-hyperscaler customer segment for Natrium units
  • Sabey partnership may focus on 1–2 unit deployments at regional data center clusters

First Deployment: Wyoming Kemmerer Site

Site & Regulatory Status

Location: Kemmerer, Wyoming (Sweetwater County) — formerly a coal plant site Site Preparation: Underway (site survey, environmental baseline, infrastructure planning) Expected Permit Status: Construction permit application expected 2026; NRC review 24–36 months nominal Target First Power: ~2030

Advantages of Wyoming:

  • Permissive regulatory environment (no state-level SMR size limits or restrictions)
  • Industrial precedent (coal generation → power plant workforce transition narrative)
  • Proximity to Meta hyperscale demand centers
  • Adequate water supply for cooling (Kemmerer site has industrial water access)

Construction & Manufacturing

TerraPower partners with:

  • BWXT (Babcock & Wilcox): Reactor vessel fabrication, ASME N stamped components
  • Kiewit Corporation: Heavy civil construction (concrete, structural steel, site infrastructure)
  • Burnham Industrial: Molten-salt storage tank fabrication and installation

Lead time expectations:

  • Permit-to-first-pour: 12–18 months post-license
  • First-of-a-kind construction: 36–48 months (Kemmerer likely on higher end due to prototype complexities)
  • Subsequent units (units 2–8): 24–36 months per unit (learning curve, pre-fabrication maturity)

Technology Advantages for Data Centers

Load-Following Without External Storage

Unlike Oklo or Kairos (which run constant baseload), Natrium’s integrated storage enables:

  • Matched load response: If data center demand drops 30%, reactor can divert heat to storage rather than curtailing power
  • Demand peaks: Molten-salt discharge can boost output 45% above baseload for 5+ hours (e.g., training run, model deployment surge)
  • Cost savings: No separate 500 MWh battery pack ($100–300M capex) needed for daily load-following

Fuel Supply Advantage

Natrium uses low-enriched uranium (LEU) metallic fuel, not HALEU:

  • LEU enrichment (<20% U-235) widely available; no supply bottleneck
  • Competing with Oklo (HALEU-dependent), Kairos (TRISO/HALEU-dependent), X-energy (TRISO-dependent)
  • TerraPower’s fuel pathway more resilient to supply constraints

Caveat: No commercial metallic-fuel processing in the U.S. yet; terawatt would need pilot-scale fab (currently BWXT can support ~4 GW/year capacity if funded). This is a soft constraint, not a binding limitation like HALEU enrichment.


Competitive Position

vs. Oklo Aurora (75 MW, nonbinding LOIs)

Factor TerraPower Natrium Oklo Aurora
Capacity 345 MW + storage 75 MW baseload only
Load-following Yes (integrated storage) No (constant output)
Fuel supply risk Low (LEU available) High (HALEU bottleneck)
Data center contracts 2.76 GW binding (Meta) 15 GW LOI (~90% nonbinding)
First deployment 2030 (Kemmerer) 2027–2028 (INL, demo phase)

vs. Kairos Hermes (50 MW, molten-salt cooled)

Factor TerraPower Natrium Kairos Hermes
Fuel LEU metallic TRISO (HALEU in some contexts)
Cooling Sodium + molten-salt storage Molten-salt direct coolant
Temperature 545°C 700+°C (higher efficiency potential)
Load flexibility Integrated thermal storage None (constant output)
Data center contracts 2.76 GW (Meta, 8 units) 500 MW (Google, 6–7 units)

Advantage: TerraPower’s storage + LEU fuel is more flexible and supply-resilient. Disadvantage: Later first deployment (~2030 vs. Kairos 2030).


Regulatory & Permitting Status

NRC Licensing

TerraPower’s Natrium design is undergoing pre-application phase reviews with NRC:

  • Design Certification Application (DCA): Expected to be filed 2026–2027 (ahead of specific project COLAs)
  • Kemmerer Combined License Application (COLA): Expected 2027–2028
  • NRC review timeline: 18–24 months nominal; first-of-a-kind designs often slip to 24–36 months
  • Operating License: Earliest issuance ~2029; construction then 36–48 months → first power ~2030–2031

Wyoming State Regulatory

  • Air Quality: No major expected opposition; Wyoming Office of Air Quality (air regulator) is pro-business
  • Water permits: Adequate industrial water available at Kemmerer site; no expected conflict
  • Decommissioning / Financial Assurance: Wyoming requires standard financial assurance for reactor decommissioning; TerraPower is securing DOE backing and commercial financing

Financial & Supply Chain

Capex Estimates

Component Estimate (2026$)
Reactor vessel + core internals $200–300M
Molten-salt storage system $100–150M
Steam turbine + generator $80–120M
Site, BOP, engineering $300–500M
Contingency (first-of-a-kind) +30–50%
Total capex (unit 1, Kemmerer) ~$1.0–1.5B
Subsequent units (learning curve) ~$0.8–1.2B

**$/MW:** ~$2.9–4.3M/MW all-in capex (Kemmerer); ~$2.3–3.5M/MW (units 2–8, learning curve).

Financing

  • Bill Gates (TerraPower founder/investor): Committed $1B+ via Breakthrough Energy Ventures
  • U.S. DOE: Path to federal loan guarantee or direct investment (advanced reactor program)
  • Meta partnership: Potential direct investment or structured PPA financing
  • Private equity: Discussions ongoing with infrastructure funds

Supply Chain Partnerships

Component Supplier Lead Time Risk
Reactor vessel (ASME N) BWXT 24–30 months Medium (first-of-a-kind)
Molten-salt storage tank Burnham Industrial / ITM 18–24 months Medium (specialty fabrication)
Metallic fuel manufacturing TBD (BWXT capable) 12–18 months Medium (scale-up needed)
Turbine/generator GE or Siemens 18–24 months Low (standard LWR heritage)
Cooling loops, piping, controls Multiple (Sensormatic, Emerson, etc.) 12–18 months Low (industrial sourcing)

Key constraint: Metallic-fuel fabrication is not yet commercial in the U.S. BWXT is the leading candidate but would require pilot-scale production facility startup (2026–2027). This is a soft constraint (solvable with capex + time), not a binding supply bottleneck like HALEU enrichment.


Risks & Challenges

⚠️ First-of-a-Kind Construction Complexity

Natrium has never been built at commercial scale:

  • Molten-salt storage system: Proven in academic + pilot settings; no full commercial deployment as primary BTM storage
  • Sodium reactor construction: Proven design lineage (EBR-II, FFTF), but U.S. regulatory environment for new sodium reactors is untested at commercial scale
  • Likelihood: Kemmerer unit 1 is likely 10–30% more expensive and 12–24 months behind schedule relative to plan

Metallic Fuel Supply Ramp

Natrium uses metallic U-Zr fuel, which offers:

  • ✅ Higher density than oxide (more burnup per batch)
  • ✅ No HALEU enrichment bottleneck
  • ⚠️ Manufacturing not yet operational in U.S.; BWXT capable but unfunded

Status: TerraPower is working with DOE on pilot-scale fuel fab funding (expected decisions 2026–2027). If funded, production facility operational by 2028; if delayed, first units may source from foreign facilities (Russia, France, UK have limited capacity) → potential 12–18 month supply delay.

Regulatory Uncertainty

  • Molten-salt systems in nuclear plants: NRC has limited precedent for integrated thermal energy storage in reactor designs; review could uncover new safety questions → schedule slippage
  • Decommissioning of molten-salt equipment: No established decommissioning precedent for large molten-salt systems; NRC may impose costly monitoring / long-term stewardship requirements

Market Execution Risk

  • Meta commitment: 8 units is a binding order, but contract likely includes “development milestones” gates. If Kemmerer unit 1 experiences major cost overruns or schedule delays >24 months, Meta may renegotiate or withdraw
  • Sabey MOU: Nonbinding; heavily dependent on Kemmerer success and cost trajectory

2026–2030 Milestones

Date Milestone Probability Impact
Q2–Q4 2026 Design Certification Application filing (NRC) High Clears regulatory pathway for any COLA
2027 Kemmerer COLA filed; construction prep begins High Site ready for first concrete
2027–2028 Metallic fuel fab pilot phase funded (DOE decision) Medium Resolves fuel supply path for units 2+
Late 2028 First concrete pour (Kemmerer) Medium ~18-month construction phase begins
2029 Fuel loading (first core assembly) Medium Final NRC operating license imminent
~2030 Critical and first power (Kemmerer unit 1) Medium–Low Demonstrates Natrium commercial viability
2031–2032 Meta unit 2 online (site TBD) Low–Medium Parallel construction begins
2033–2035 Meta units 3–8 online (phased) Low If schedule holds and costs acceptable

Outlook: Why Natrium Matters for Data Centers

Natrium’s integrated molten-salt storage is a genuine competitive advantage over constant-baseload SMRs (Oklo, Kairos, X-energy). Data center power demands are not flat—they vary 2–5x between idle and peak model-training loads. Most SMR developers are ignoring this; Natrium is solving for it.

If Kemmerer unit 1 succeeds (on-time, on-budget), Natrium could capture 5–10% of SMR data center market by 2035. If Kemmerer faces major cost overruns or delays, Meta’s 8-unit commitment likely dissolves or shrinks to 2–3 units; Natrium becomes niche player.

Supply chain and fuel: Advantage over Oklo/Kairos (no HALEU bottleneck), but dependent on metallic-fuel fab capitalization in 2026–2027.

Competitive pressure: Watch for Kairos molten-salt cooled design to offer storage options (different approach from Natrium’s separate storage tank). If Kairos adds storage, it could erode Natrium’s differentiation.

Key watch: Kemmerer site prep and COLA timeline 2026–2027 will determine whether 2030 first-power target is realistic or slips to 2032+.


Key People

Bill Gates — Co-Founder, Principal Investor

  • Co-founded TerraPower in 2006 alongside John Deutch; primary private backer
  • LinkedIn: public profile available
  • Background: Microsoft co-founder; global health/energy philanthropist via Bill & Melinda Gates Foundation; nuclear energy advocate

Chris Levesque — President & CEO

  • LinkedIn: TODO: verify slug — note this is the same name as Kairos CEO; these are different people. TODO: find TerraPower’s Chris Levesque LinkedIn.
  • Prior: TerraPower COO (pre-CEO); deep advanced reactor operations experience. ⚑ Overlap: TerraPower and Kairos Power are often discussed together in data center nuclear context; the coincidence of the name “Chris Levesque” at both companies requires verification — confirm these are distinct individuals.

People — Last Reviewed: 2026-04-25


  • /research/datacenters/behind-meter-power/oklo-btm/ — Oklo Aurora microreactor (75 MW, nonbinding LOIs)
  • /research/datacenters/behind-meter-power/kairos-power-btm/ — Kairos Hermes (50 MW molten-salt cooled SMR, Google partnership)
  • /research/energy/nuclear/x-energy/ — X-energy Xe-100 (80 MW HTGR, Amazon partnership)
  • /research/energy/nuclear/blue-energy/ — Blue Energy SMR (emerging, Crusoe partnership)