Crusoe Energy
Started flaring gas into Bitcoin miners — now the OpenAI Stargate prime data center developer with 4.5 GW of contracted gas power.
1. Core Product / Service
Crusoe pivoted from "Digital Flare Mitigation" (2018-2022: capturing waste flare gas from oil wells to power Bitcoin miners) to integrated AI data center developer: it sources stranded power, builds the gas turbines, builds the data center shell, and operates the GPU compute on top.
Today the stack:
- Power generation — secured 4.5 GW of natural-gas turbine capacity for AI data centers, expected online 2027 [3][4]
- Data center development — Abilene, TX flagship campus fully operational at 1.2 GW, with the final building topped out Q1 2026 [4]
- AI cloud (Crusoe Cloud) — H100/H200 + Blackwell GPUs sold by the hour, similar profile to coreweave, nebius, lambda-labs
The big shift in 2024-2026: Crusoe became the prime real-estate developer for OpenAI's Stargate — the Abilene campus is the anchor site for the $500B Stargate program, with Oracle as the cloud operator and OpenAI as the workload owner [5][6].
2. Target Users & Pain Points
- OpenAI / Stargate — anchor tenant, multi-year, multi-GW commitment
- Hyperscalers and AI labs needing immediate power-secured capacity
- Crusoe Cloud customers — AI training startups buying GPU-hours
Pain solved: power, fast. The constraint in 2025-2026 isn't GPUs (NVIDIA is shipping); it isn't data center buildings (any colo can build); it's megawatts on the grid. Crusoe's strategy of bringing its own gas-turbine generation lets it bypass utility interconnect queues that block conventional builds for 3-5+ years.
3. Competitive Landscape
| Company | Approach | Positioning vs Crusoe |
|---|---|---|
| coreweave | Lease colo, finance GPUs via debt | Doesn't own power generation |
| nebius | Build own DCs (mostly EU) | Smaller scale, slower power thesis |
| lambda-labs | Mostly leased capacity | Pure GPU cloud, no power play |
| Stargate (Oracle/OpenAI/SoftBank) | Massive multi-site program | Crusoe is inside this — the prime developer for the anchor site |
| Vantage / Aligned / QTS | Hyperscale wholesale | Don't generate own power; depend on grid |
Differentiation: vertical integration from gas wellhead to GPU. No competitor combines power generation + data center development + GPU cloud at this scale.
4. Unique Observations
- From flare gas to Stargate prime: Crusoe's origin (2018, mining Bitcoin on stranded flare gas from Bakken oil wells) is the genesis of its current edge — it has 7 years of operational experience integrating gas turbines with compute load. Most hyperscalers are now scrambling to do behind-the-meter gas; Crusoe has been doing it from day one.
- Stargate Abilene is the anchor: 1.2 GW operational at Abilene as of Q1 2026 [4], running on Oracle Cloud Infrastructure with OpenAI as primary workload. The total Stargate program is reportedly ~7 GW planned with $400B+ in committed investment, $500B announced [5][6].
- Customer concentration is extreme — heavily skewed to OpenAI/Stargate (with Oracle as the cloud-layer counterparty). Similar concentration risk to coreweave's Microsoft/Meta exposure but arguably worse: Stargate cancellation/restructuring risk has been publicly discussed (some Stargate sites are reportedly being renegotiated due to Oracle terms).
- $/MW build economics — Crusoe + partners' Abilene campus, mixed with on-site gas generation, runs ~$15-25M/MW all-in (data center + power generation + GPU fit-out estimated separately). Without power generation, hyperscale wholesale is ~$10-12M/MW; the gas turbine adds ~$1-2M/MW; AI-density retrofit is the big swing factor.
- Depreciation profile — gas turbines: 25-30 years. DC shell: 20-30 years. GPU fit-out: 4-6 years on a chip refresh cycle. Crusoe's vertical integration helps spread the chip-refresh risk across longer-lived infrastructure assets, structurally similar to coreweave's "buy GPU, refresh GPU, keep DC" playbook.
- Capital intensity is the bear case: 4.5 GW of gas + AI data centers = $50-100B+ of capex over 3-5 years. Crusoe's $1.375B Series E at $10B valuation barely scratches the funding need; the company is increasingly debt-financed (Goldman Sachs led a recent conventional debt round, Feb 2026 [7]).
- Cross-link: Crusoe sits at the L1+L2 boundary in framework.md terms — owns physical infrastructure (L1) and operates as a GPU cloud (L2 New Cloud), with Stargate vertical-integrating L1+L2+L3 (OpenAI's training workload).
5. Financials / Funding
- Series E: October 2025, $1.375B at >$10B valuation — led by Valor Equity Partners and Mubadala Capital; participation from NVIDIA, Fidelity, T. Rowe Price, Tiger Global, Salesforce Ventures, and Founders Fund [1]
- Total funding: ~$2.77B across 15 rounds from 69 investors [4]
- Prior valuation: $2.8B (Dec 2024) → >$10B (Oct 2025) — ~3.5× valuation step in 10 months [2]
- Goldman Sachs first investment: Feb 20, 2026, conventional debt round [7]
- Power capacity contracted: 4.5 GW natural gas, online by 2027 [3]
- Operational capacity: 1.2 GW at Abilene flagship (topped out Q1 2026) [4]
6. People & Relationships
- Co-founders: Chase Lochmiller (CEO), Cully Cavness (President), Jason Sloop
- Lead investors: Valor Equity Partners, Mubadala Capital
- Strategic investors: NVIDIA, Founders Fund, Fidelity, T. Rowe Price, Tiger Global, Salesforce Ventures
- Anchor partner: Stargate (OpenAI / Oracle / SoftBank / MGX) — Crusoe builds, Oracle operates the cloud, OpenAI is the workload
- Cross-stack: NVIDIA equity stake mirrors NVIDIA's strategy with coreweave — "we own a piece of the customer that buys our GPUs"