nuclear partnerships for data centers

Overview

Nuclear energy has emerged as the strategic solution to data centers’ insatiable power demands and carbon-free commitments. In 2024-2025, $12+ billion has been committed to data center-dedicated nuclear projects, representing the largest private investment in U.S. nuclear infrastructure since the 1980s.

Two Models Emerging:

  1. Small Modular Reactors (SMRs): New construction, 50-300 MW units
  2. Reactor Restarts: Bringing recently-closed nuclear plants back online
MetricValue
Total Committed Investment$12+ billion
Number of Partnerships9 major agreements
Total Planned Capacity6.5+ GW
SMR Projects5 facilities
Reactor Restart Projects3 facilities
Hyperscalers InvolvedGoogle, Microsoft, Amazon

Major Partnerships

Google-TVA-Kairos Power (Tennessee)

Announced: October 2025 Capacity: 50 MW (initial), expandable to 500 MW Technology: Kairos Power Hermes SMR (molten salt-cooled) Location: Oak Ridge, Tennessee area Timeline: First reactor operational by 2030

ComponentDetails
Investment$1.5B (Google contribution undisclosed)
Reactor TypeFluoride salt-cooled high temperature
Google Commitment25-year power purchase agreement
TVA RoleSite host, operator, grid integration
Kairos RoleReactor design, construction, licensing

Significance:

  • First hyperscaler-utility-vendor SMR partnership in U.S.
  • Google’s carbon-free energy commitment (24/7 by 2030)
  • TVA returning to nuclear construction after decades

Technical Innovation:

  • Molten salt cooling: Higher efficiency than water-cooled
  • Walk-away safe: Passive safety systems (no active cooling required)
  • Modular construction: Factory-built, site-assembled
  • Fuel flexibility: Can use HALEU (high-assay low-enriched uranium)

Development Milestones:

  • 2025 Q4: NRC license application submission
  • 2026-2028: NRC review process
  • 2028: Construction commencement
  • 2030: First unit operational
  • 2032-2035: Expansion to 500 MW (10 units)

Google Data Center Integration:

  • Location: Tennessee data center cluster (existing + planned)
  • Dedicated offtake: 100% of power to Google facilities
  • Grid backup: TVA grid provides redundancy
  • Renewable complement: Solar + storage augmentation

Microsoft-Constellation Three Mile Island Restart

Announced: September 2024 (advancing through 2025) Capacity: 835 MW (Unit 1) Location: Londonderry Township, Pennsylvania Timeline: Operational 2028

ComponentDetails
Investment$1.6B restart costs
Microsoft PPA20 years, 100% offtake
Original Closure2019 (economic reasons)
Restart Timeline4 years (2024-2028)
RenamedCrane Clean Energy Center

Why Three Mile Island?

  • Recently closed (2019) - infrastructure intact
  • Strong safety record since 1979 incident (Unit 1 unaffected)
  • PJM interconnection - existing transmission
  • Pennsylvania location - proximity to East Coast data centers

Restart Process:

  • 2024: Constellation announces intent, Microsoft PPA signed
  • 2025: NRC restart review initiated
    • Safety systems inspection
    • Equipment certification
    • Environmental review
  • 2026-2027: Refurbishment work
    • Steam generators replacement
    • Control systems upgrade
    • Cooling systems restoration
  • 2028: Operational restart

Microsoft Data Center Integration:

  • Pennsylvania data center cluster: Multiple facilities powered
  • Carbon-free goal: Eliminates need for renewable energy certificates (RECs)
  • 24/7 matching: Nuclear provides constant baseload (vs. intermittent renewables)
  • Grid independence: Reduces reliance on constrained PJM grid

Economic Impact:

  • Jobs: 3,400 during refurbishment, 600 permanent operations
  • Local economy: $3B economic impact over 20 years
  • Tax revenue: $1B+ to Pennsylvania over PPA term

Amazon-Energy Northwest (Washington)

Announced: October 2024 Capacity: 4 SMRs, 320 MW total (80 MW each) Technology: X-energy Xe-100 SMR (high temperature gas-cooled) Location: Near Columbia Generating Station, Washington Timeline: First unit 2029, full deployment 2032

ComponentDetails
Investment$2.0B (Amazon portion)
Reactor TypeXe-100 high temperature gas-cooled
FuelTRISO (tristructural isotropic) pebbles
Amazon CommitmentRight to purchase up to 960 MW (12 units)
Energy NorthwestOperator, existing nuclear site

Strategic Rationale:

  • AWS US-West region: Proximity to Oregon/Washington data centers
  • Baseload carbon-free: Complements existing hydropower (Columbia River)
  • Scalability: Modular deployment (4 units initially, up to 12 potential)

X-energy Xe-100 Technology:

  • TRISO fuel: Inherently safe (cannot melt down)
  • High temperature: 750°C outlet temp (vs. 300°C light water reactors)
  • Small footprint: 80 MW each, factory-assembled
  • Passive cooling: No active systems required for shutdown

Deployment Plan:

  • Phase 1 (2029): 2 units, 160 MW
  • Phase 2 (2031): 2 additional units, 160 MW
  • Phase 3 (2032+): Up to 8 more units (640 MW)

Amazon-Dominion Energy (Virginia)

Announced: November 2024 Capacity: Exploring “multiple gigawatts” near existing nuclear plants Location: Louisa County, Virginia (near North Anna nuclear station) Timeline: Feasibility study 2025-2026

Focus:

  • Data center cluster: Louisa County emerging as Virginia’s next major hub
  • Existing infrastructure: Leverage North Anna transmission
  • SMR deployment: Likely 6-10 SMR units (500-1000 MW total)

Status: Feasibility + site studies underway


Pennsylvania Nuclear Corridor

Multiple projects exploring nuclear integration in Pennsylvania, creating a “nuclear corridor” for data centers.

Susquehanna Nuclear Station Proximity Projects

Homer City Energy Campus (4.5 GW planned):

  • Location: 45 miles from Susquehanna nuclear plant
  • Strategy: Grid connection to nuclear baseload
  • Status: Planned, permitting stage

TECfusions Keystone Connect (3 GW):

  • Location: Northeast Pennsylvania
  • Nuclear proximity: Susquehanna + Limerick
  • Status: Under construction

Project Jupiter (New Mexico) Nuclear Component

Planned Addition: 500 MW SMR Timeline: 2027-2030 feasibility, 2030+ deployment Technology: Vendor selection in progress (NuScale, X-energy, or Kairos finalists) Rationale: On-site baseload for AI training workloads


Small Modular Reactor (SMR) Technology

Leading Vendors

1. NuScale Power

  • Design: Light water SMR, 77 MW per module
  • Status: First NRC design certification (January 2023)
  • Projects: Utah UAMPS project (delayed), evaluating data center market
  • Advantages: Proven light water technology, certified design
  • Challenges: Higher cost than competitors, smaller capacity per unit

2. X-energy

  • Design: Xe-100, high temperature gas-cooled, 80 MW
  • Status: NRC design certification review underway
  • Projects: Amazon-Energy Northwest (Washington)
  • Advantages: TRISO fuel safety, high temperature efficiency
  • Challenges: New fuel form factor (supply chain development)

3. Kairos Power

  • Design: Hermes, molten salt-cooled, 50 MW (scalable)
  • Status: NRC application submitted October 2025
  • Projects: Google-TVA (Tennessee)
  • Advantages: Highest temperature (efficiency), walk-away safety
  • Challenges: Novel design (longer NRC review expected)

4. TerraPower (Bill Gates)

  • Design: Natrium, sodium-cooled fast reactor, 345 MW
  • Status: Demonstration project in Wyoming
  • Projects: Not yet data center-specific, but evaluating
  • Advantages: Large capacity per unit, molten salt energy storage integration
  • Challenges: Sodium cooling complexity

SMR Economics

Capital Costs:

  • Per unit: $400M - $1.2B (depending on design, 50-350 MW)
  • Per kW: $5,000 - $7,000/kW (higher than conventional nuclear)
  • Learning curve: Expected 30-40% cost reduction by 10th unit (factory production)

Operational Costs:

  • Fuel: $15-20/MWh (vs. $40-60/MWh natural gas)
  • O&M: $10-15/MWh
  • Total LCOE: $60-80/MWh (first units), declining to $40-55/MWh (at scale)

Data Center Value Proposition:

  • 24/7 carbon-free: Eliminates REC purchases ($20-40/MWh)
  • Price certainty: 40-60 year reactor life (vs. volatile gas prices)
  • Grid independence: Reduces interconnection queue risk

Reactor Restart Economics

Three Mile Island Model

Restart Costs: $1.6B total

  • Equipment replacement: $800M (steam generators, turbines)
  • Systems upgrades: $400M (control systems, safety)
  • Licensing/permitting: $200M (NRC fees, environmental)
  • Contingency: $200M

Cost Comparison:

  • Per kW: $1,900/kW (vs. $5,000-7,000 for new SMR)
  • Timeline: 4 years (vs. 7-10 years new construction)
  • Economics: Highly favorable for recently-closed plants

Restart Candidates (closed within 10 years):

PlantStateCapacityClosedRestart Feasibility
Three Mile IslandPA835 MW2019In progress (Microsoft)
Duane ArnoldIA615 MW2020Study underway (Google/QTS)
Indian Point 2NY1,028 MW2020Low (political opposition)
Indian Point 3NY1,041 MW2021Low (political opposition)
PilgrimMA685 MW2019Low (decommissioning started)

Duane Arnold (Iowa) - Under Evaluation

Status: Feasibility study initiated Q4 2025 Drivers: Google Iowa ($7B), QTS Cedar Rapids ($10B), Microsoft West Des Moines Combined data center load: 2+ GW by 2030

Restart Economics:

  • Cost estimate: $1.2B
  • Timeline: 2026-2029 (if approved)
  • PPAs: Google, Microsoft, QTS negotiating 25-year agreements
  • Viability: High (strong PPA interest, intact infrastructure)

Industry Implications

Carbon-Free Energy Achievement

Hyperscaler Carbon-Free Goals:

  • Google: 24/7 carbon-free by 2030
  • Microsoft: Carbon negative by 2030
  • Amazon: Net-zero carbon by 2040

Nuclear’s Role:

  • 24/7 generation: Unlike wind/solar (intermittent)
  • No RECs needed: Direct carbon-free power (vs. market-based credits)
  • Baseload match: Constant data center load requires constant generation

Impact:

  • Renewable displacement? No - nuclear complements renewables
  • Grid stability: Nuclear provides baseload, renewables provide marginal

Grid Independence

Interconnection Queue Crisis:

  • PJM: 3-5 year queue for large loads
  • ERCOT: 70 GW backlog
  • CAISO: Similar constraints

Nuclear Solution:

  • On-site generation: Bypasses queue entirely
  • Behind-the-meter: Direct connection (vs. grid interconnection)
  • Timeline certainty: 4-7 years (vs. 5-10 years grid-dependent)

Investment Scale

Total 2024-2030 Nuclear for Data Centers:

  • SMR deployments: $8.0B (5 projects)
  • Reactor restarts: $4.0B (2-3 projects)
  • Total committed: $12.0B+
  • Potential pipeline: $25B+ (additional announced studies)

Comparison:

  • Utility grid expansions for data centers: $52B (2024-2030)
  • Nuclear share: 23% of total infrastructure investment

Regulatory Pathway

NRC Licensing Process

SMR Design Certification:

  1. Pre-application (1-2 years): Vendor engagement with NRC
  2. Formal review (3-5 years): Design certification application
  3. Certification (upon completion): 15-year design approval

Construction&Operating License (COL):

  1. Site-specific review (2-3 years): Environmental, safety analysis
  2. Combined license issued: Allows construction + eventual operation
  3. ITAAC completion: Inspections, tests, analyses, and acceptance criteria

Total Timeline: 7-10 years from concept to operation (first unit)

Accelerated pathways:

  • Fleet deployment: 2nd+ unit 4-5 years (vs. 7-10 first)
  • Reactor restarts: 4-5 years (existing license modification)

Environmental Review

NEPA Process:

  • Environmental Impact Statement (EIS) required
  • Public comment period: 60-90 days
  • Common concerns: Water usage, thermal discharge, waste storage

Data Center-Nuclear Integration:

  • Co-location advantage: Single EIS for facility + power (vs. separate reviews)
  • Cooling synergy: Data center waste heat + nuclear cooling optimization

Challenges & Risks

Technical Challenges

SMR Deployment:

  • First-of-a-kind risk: New designs have construction/operational unknowns
  • Supply chain: HALEU fuel production limited in U.S.
  • Workforce: Limited SMR construction/operation experience

Reactor Restarts:

  • Equipment degradation: Extended closure may require more extensive refurbishment
  • Regulatory: NRC scrutiny on safety systems after extended shutdown
  • Public perception: Historical incident concerns (e.g., TMI name recognition)

Economic Risks

Cost Overruns:

  • Nuclear history: Conventional reactors typically 2-3x over budget
  • SMR promise: Factory construction should reduce, but unproven
  • Mitigation: Fixed-price PPAs shift risk to generator

Competitive Pressure:

  • Renewable costs declining: Solar + storage approaching $40/MWh
  • Nuclear LCOE: $60-80/MWh (SMR first units)
  • Counter: 24/7 matching value + grid independence premium

Political/Social Risks

Nuclear Opposition:

  • Environmental groups: Some oppose nuclear expansion
  • Local resistance: Waste storage concerns, safety fears
  • Mitigation: Data center community engagement lessons apply to nuclear component

Waste Management:

  • No permanent repository: Yucca Mountain cancelled
  • On-site storage: SMRs produce less waste (smaller cores)
  • Solution horizon: 40-60 year reactor life outlasts current waste storage debate

Forward Outlook

2026-2030 Projections

SMR Deployments:

  • Confirmed: 5 projects, 1.2 GW, $8B
  • Probable: 8-10 additional projects, 2-3 GW, $12-15B
  • Possible: 20+ projects by 2030 (if early projects succeed)

Reactor Restarts:

  • Confirmed: TMI Unit 1, 835 MW, $1.6B
  • Probable: Duane Arnold, 615 MW, $1.2B
  • Possible: 2-3 additional (if market develops)

Total 2030 Data Center Nuclear Capacity:

  • Conservative: 3 GW ($12B invested)
  • Base case: 6 GW ($25B invested)
  • Bullish: 10+ GW ($40B+ invested)

Technology Evolution

Next-Generation SMRs (2030+):

  • Microreactors: 1-20 MW (single data center facility)
  • Advanced fuels: Accident-tolerant fuels (longer refueling cycles)
  • Hybrid systems: Nuclear + hydrogen production + data center loads


Last Updated: December 30, 2025 Nuclear partnership data current as of December 2025 Next Update: Q2 2026 (following Kairos/Google NRC application milestone)

on this page