The Infinite Unknown

Summary

This profile focuses on Oklo’s behind-the-meter (BTM) power value proposition for data centers, separate from the full Oklo company profile (see /research/energy/nuclear/oklo/ for detailed company history, financing, and technical background).

BTM angle: Oklo has the largest single order book for data center power among all nuclear vendors (~18 GW of nonbinding LOIs/framework agreements with Meta, Switch, Equinix, Wyoming Hyperscale, and others). The Aurora reactor’s compact form factor, passive safety, and 75 MWe output make it uniquely suited for co-located or near-site deployment at large hyperscale facilities. First commercial unit at Idaho National Lab (groundbreaking Sept 2025) is the critical proof-of-concept for broader datacenter deployments (2028+).

Key BTM advantages:

• Form factor: Compact, below-grade installation; ~1–2 acre footprint vs. 50+ acres for conventional plants
• Walk-away safe: Passive cooling; no active systems required for core cooling post-shutdown
• Time-to-grid: Could provide first power to datacenter in 3–4 years post-site selection (vs. 5–10 for grid plants)
• Certainty: Oklo owns and operates reactor; datacenter customer has long-term PPA with single provider (no utility intermediary risk)

Key risks:

• HALEU fuel supply: First unit gets INL inventory; commercial supply beyond 2–3 units unresolved
• NRC licensing: First-of-a-kind; COLA resubmission 2025, approval uncertain, could slip 12–24 months
• Order book reality: ~18 GW is 90% nonbinding LOIs; binding PPAs likely <5 GW as of April 2026
• Execution risk: Single first project (INL); if delays >12 months, entire timeline cascades backward

BTM Value Proposition vs. Grid-Connected Nuclear

Why Datacenter Operators Choose Oklo Aurora BTM

1. Grid interconnection bypass: Data center operator avoids:

   • 5–10 year queue for grid connection
   • Utility transmission pricing (eliminates $50–100M+ transmission cost over 20-year PPA)
   • Capacity reserve mandates (operator self-provides via on-site generation)
   • Regulatory review by state utility commission (Oklo negotiates directly with datacenter)

2. Time-to-power: Aurora can achieve first power 3–4 years post-site selection (if NRC licensing path accelerates), vs. 10+ for grid-connected plant

3. Certainty & resilience: Datacenter has guaranteed power supply independent of grid; enables 99.99%+ uptime commitment to customers (critical for hyperscaler SLAs)

4. Sustainability narrative: 24/7 carbon-free power (genuinely zero-carbon, unlike gas BTM); enables “powered by nuclear” marketing

5. Scale: 75 MW per unit is suitable for large hyperscale campuses (datacenter draw: 50–200 MW typical); multiple units can be deployed at same site for 150–300 MW+ clusters

Oklo Aurora: BTM-Specific Technical Fit

Compact Footprint

Aurora is designed for below-grade siting:

• Reactor vessel in underground pool; containment building ~50 ft x 50 ft
• Total footprint: <2 acres (vs. 50+ for traditional 1,100 MW reactor)
• Enables siting near or adjacent to datacenter (reduces transmission loss)

Datacenter campus integration:

• Multiple Aurora units can be deployed on same site
• Modular power addition (75 MW at a time); scalable to 150–300 MW
• Integrated cooling (Oklo + Vertiv partnership provides both power and thermal management)

Passive Safety (No Active Cooling Required)

Aurora uses liquid sodium coolant in a pool configuration:

• Gravity-fed natural circulation (no pumps during normal operation)
• After shutdown, core continues cooling passively via thermosiphon action
• No operator action, backup power, or external cooling water required
• Walk-away safe: even if all active systems fail, reactor remains safe

Datacenter implication:

• Exceptionally low risk of emergency shutdown (rare for stationary power plants)
• Minimal environmental permitting friction (exceptional safety record enables easier local approval)

Flexible Output

Aurora was originally designed at 15 MWe, but uprated to 75 MWe in 2024–2025 specifically to meet datacenter power demand. This flexibility:

• Signals Oklo’s responsiveness to market pull (Meta, Switch wanted bigger units)
• Enables 75 MW single-unit deployments (more economical than clusters of small units)
• Future uprates to 100+ MWe possible (design margin exists)

Data Center Customer Order Book

Meta (Facebook’s parent company)

Announcement: Master Power Agreement (~15 GW implied capacity)

• Largest single customer commitment to Oklo
• Spans multiple Aurora units (200–300 reactors if fully executed)
• Timeline: Not specified; likely phased 2029–2035

Reality check:

• 15 GW represents all of Meta’s forecast datacenter power demand through 2035
• However, agreement is a framework (nonbinding LOI); actual unit orders likely <1 GW binding as of April 2026
• Meta is hedging across multiple nuclear vendors (TerraPower, Vistra, Constellation Energy); Oklo is largest but not exclusive

Switch Communications (Las Vegas colocation)

Announcement: 12 GW supply agreement (2024)

• Second-largest Oklo customer
• Switch operates large data center campuses (Las Vegas, other U.S. locations)
• Timeline: 2028+

Reality check:

• Switch is smaller than Meta/Google/Amazon
• 12 GW is ~160 Aurora units; implies $160B+ total project cost
• More likely a framework; actual binding orders <2 GW

Equinix (Global colocation)

Announcement: Letter of Intent for 500 MW initial; potential for more

• Equinix is largest global colocation provider (>250 data centers)
• LOI does not specify timeline or location
• Part of Equinix’s energy diversification strategy (also contracts with Radiant, Bloom, Kongu, others)

Wyoming Hyperscale

Announcement: 100 MW power purchase agreement (20-year PPA)

• First Oklo customer to announce binding PPA (not just LOI)
• Smaller scale but contractually binding
• Target commissioning: 2030s

Other LOI Holders

• Prometheus Hyperscale (Wyoming): LOI, scope TBD
• Diamondback Energy (oil+gas company hedging): LOI, scope TBD

First Commercial Deployment: Idaho National Lab (INL) Aurora Unit 1

Project Timeline

September 2025: Oklo broke ground on Aurora Unit 1 at INL

• Constructor: Kiewit (lead contractor)
• Target criticality: Q3–Q4 2027 (original target July 4, 2026 now slipped)
• Commercial operation: Early 2028

Current status (April 2026):

• Early construction underway (site prep, building foundation)
• NRC pre-application review active
• Formal COLA Phase 1 submission expected 2025 (per Oklo guidance), but no public announcement of submission as of April 2026
• Fuel supply: DOE has committed to provide HALEU from INL inventory for this unit

Why INL?

Advantages:

• Federal site (no state regulatory approval required for nuclear construction)
• Existing nuclear infrastructure (NRF, advanced reactor testing facilities)
• HALEU fuel available at INL (eliminates commercial fuel supply bottleneck for first unit)
• Willing host community (nuclear expertise, favorable history)

Disadvantages:

• Remote location (not adjacent to major datacenter; power must be transmitted via grid or dedicated transmission line)
• INL project is a demonstration, not pure datacenter application (but proves commercial viability)
• Regulatory pathway less precedent-setting for commercial datacenter-sited SMRs

Licensing Path

NRC Combined License Application (COLA):

• Phase 1: Design-basis information (Oklo submitted pre-app; NRC completed readiness review 2025)
• Phase 2: Site-specific safety analysis, emergency planning, security (18–24 months)
• Phase 3: Construction and operational review (12 months)
• Total time to operating license: 3–4 years from COLA Phase 1 submission (if no major deficiencies)

Current estimate:

• COLA Phase 1 submission: 2025 (reported, but not publicly confirmed)
• COL issuance: 2027–2028
• Critical heat-up: 2028
• Full power operation: 2028–2029

Risk: First-of-a-kind licensing delays are common (+12–24 months vs. schedule).

HALEU Fuel: The Real Bottleneck

Supply Chain Reality (April 2026)

U.S. HALEU enrichment capacity:

• Current: INL has ~200–300 MT of historical HALEU inventory (mostly from EBR-II fuel)
• Commercial production: No U.S. commercial enrichment facility currently operating at HALEU levels (>19.75% U-235)
• Centrifuge capacity: U.S. has limited enrichment capacity; primary civilian enrichment provider (Urenco) operates at Kentucky facility but is licensed for LEU only (<5% U-235)

DOE’s 2024 actions:

• Committed to establish domestic HALEU production capacity
• Funded grants to Centrus Energy for uranium enrichment expansion (up to 20% enriched)
• Timeline: First commercial HALEU production ~2027–2028 (unproven; may slip)

Impact on Oklo Scaling

Scenario 1: HALEU supply remains constrained

• INL covers Oklo’s first 1–2 units
• Commercial supply beyond 2027 is uncertain
• Oklo deployments capped at ~5–10 units (375–750 MW) by 2030
• This significantly reduces Oklo’s addressable market vs. 18 GW order book claim

Scenario 2: Commercial HALEU production ramps (2027–2029)

• Sufficient supply for 50–100 MT/year production
• Enables Oklo + other HALEU users (X-energy, Kairos, USNC) to collectively deploy ~10–20 units/year
• Oklo could achieve 2–3 GW deployed capacity by 2035

Realistic scenario (as of April 2026): HALEU supply will likely be constrained through 2028–2030; first-wave SMR deployments (Oklo, X-energy, Kairos) will be limited to 5–10 units each. This is a major headwind against the 18 GW order book narrative.

Vertiv Partnership: Integrated Power + Cooling

July 2025 announcement: Oklo and Vertiv announced strategic collaboration

• Scope: Co-develop integrated power and thermal management systems for hyperscale/colo datacenters
• Value add: Oklo provides electricity + high-temperature steam; Vertiv provides power distribution and cooling solutions

Implication:

• Aurora’s excess heat can be used for space heating, hot water, or process heat (hydrogen production, desalination)
• Integrated design reduces total capital cost vs. separate nuclear plant + cooling plant
• Vertiv’s customer relationships (major colo/hyperscaler accounts) accelerate Oklo market entry

Status: Partnership in early stages; no announced joint projects yet (as of April 2026)

Regulatory Friction Points

⚑ State-Level Nuclear Siting Opposition

Some states (California, New York, Vermont) have statutory bans or restrictions on new reactor construction. If Oklo targets siting in:

• California: Regulatory path highly constrained (state law restricts new reactors)
• New York: Political opposition from anti-nuclear groups; state PSC would need to approve
• Texas, Wyoming, Louisiana: Favorable permitting environment; faster siting approval

Oklo’s strategy: Deploy first units in permissive states (Wyoming, Texas, Idaho). Success at INL and Wyoming Hyperscale will create political cover for future coastal siting.

⚑ Interconnection vs. BTM Labeling

If Oklo-powered datacenters are configured for grid-parallel operation (backup to grid or load-sharing), utility regulators may require:

• Formal interconnection review (adds 12+ months)
• Capacity reserve contribution (utility demand on operator)
• Transmission cost allocation (operator pays for grid support)

Oklo’s mitigation: Position Aurora as pure BTM (fully islanded, no grid interconnection). This avoids regulatory review but requires datacenter to manage load independently (no grid backup).

Competitive Position vs. Other SMR Vendors

Oklo’s advantage: Smallest, most compact, first to INL. Oklo’s disadvantage: Smallest capacity per unit (75 MW vs. 345 MW for TerraPower); higher per-MW capex.

Financial Implications for Datacenter Customers

Capex Estimate

Oklo Aurora is estimated at:

• $5–7B per unit (first-of-a-kind cost, including licensing, security, grid connection)
• $70–90M/MW (higher than mature nuclear, lower than emerging SMRs like NuScale early units)
• This compares to $2–3M/MW for gas turbines, $1–1.5M/MW for fuel cells

Cost curve: Future units (post-first-commercial) could drop to $3–5B/unit if licensing and supply chains mature (2030+).

PPA Pricing

Oklo is likely to offer 20–30 year PPAs at:

• Base price: $50–80/MWh (all-in energy cost, including amortized capex, O&M, fuel)
• Escalation: 2–3% annually
• Certainty premium: Customers pay premium for guaranteed 24/7 carbon-free power (vs. renewable + grid)

Datacenter economics:

• Hyperscaler willing to pay $50–80/MWh for 24/7 carbon-free baseload (vs. $30–40/MWh grid power + $10–20/MWh renewable premium)
• Payback: If Oklo eliminates 5–10 year grid interconnection delay, capex is worth $1–2B in avoided opportunity cost per GW

Key People (Oklo-specific roles)

• Jacob DeWitte (CEO, Co-founder) — MIT nuclear engineering; LinkedIn: TODO: verify slug. Prior: MIT research; co-founded Oklo 2013 with Caroline Cochran.
• Caroline Cochran (COO, Co-founder) — LinkedIn: TODO: verify slug. Prior: MIT aerospace engineering; co-founded Oklo with DeWitte.
• Sam Altman (Chairman) — LinkedIn: TODO: verify slug. OpenAI CEO; invested in Oklo via SPAC (2023). ⚑ Overlap: Altman is CEO of OpenAI (major Stargate initiative customer) and Chairman of Oklo — a potential conflict of interest worth noting in any deal between OpenAI infrastructure and Oklo power.
• Kiewit: Construction partner for Aurora Unit 1 at INL. No named individual confirmed as executive sponsor on public record as of April 2026.

People — Last Reviewed: 2026-04-25

2026–2027 Critical Milestones (BTM-specific)

1. Q2–Q3 2026: COLA Phase 1 submission to NRC (if on schedule)

   • Public announcement confirms licensing momentum
   • Triggers NRC technical review (18–24 months)

2. Q4 2026 – Q2 2027: First binding datacenter PPA announcement (beyond Wyoming Hyperscale LOI)

   • Signals market pull for commercial deployment
   • Likely candidate: Meta, Switch, or new enterprise datacenter operator

3. Q3–Q4 2027: Aurora Unit 1 achieves criticality

   • Proves sodium-cooled fast reactor commercial viability
   • Reduces technical risk for subsequent units
   • Triggers Meta, Switch, and other LOI holders to accelerate to binding commitments

4. 2027–2028: Second datacenter-sited Aurora project breaks ground

   • If INL success drives this, Oklo’s BTM datacenter strategy is validated
   • Likely location: Wyoming, Texas, or strategic location chosen by Meta or Switch

Risks to BTM Oklo Narrative (2026–2027 Watch)

1. INL licensing delay: If COLA approval slips >12 months beyond 2027 target, entire timeline cascades. Subsequent datacenter projects (Meta, Switch) will defer commitments.

2. HALEU supply bottleneck becomes public: If DOE/Centrus struggle to produce commercial-scale HALEU by 2028, Oklo will be forced to publicly acknowledge limitation on deployments. This will crimp near-term order book enthusiasm.

3. Meta/Switch backing away: If hyperscalers pivot more aggressively to nuclear-existing-plants (Constellation Energy) or other SMR vendors (Kairos, X-energy), Oklo’s order book becomes less credible.

4. Cost escalation: If INL Aurora Unit 1 costs balloon >20% above initial estimates, capex per MW rises further; datacenter customer ROI deteriorates, making long-term PPAs less attractive.

Summary: Oklo BTM Strategy

Oklo is the leading SMR vendor targeting data center BTM power, with the largest single order book (~18 GW LOIs) and first commercial deployment (INL Aurora Unit 1, targeting 2027–2028). The company’s value proposition is compelling: compact, safe, 24/7 carbon-free power with potential time-to-market advantage over grid-connected nuclear.

However, the narrative faces three critical tests:

1. INL Aurora Unit 1 must achieve criticality on or near schedule (slips >18 months are show-stoppers)
2. HALEU fuel supply must transition to commercial production by 2028–2029 (constrained supply caps deployments to ~5–10 units/developer by 2030)
3. Meta, Switch, and other large LOI holders must move to binding commitments (LOI-to-PPA conversion rate will determine real addressable market)

If all three succeed: Oklo could deploy 2–3 GW by 2035, and become a major player in datacenter power. If any fails significantly: Oklo becomes a niche provider, with <500 MW deployed by 2035.

For detailed company background, financing, and technology, see /research/energy/nuclear/oklo/.