$30 billion. Eleven months after its last round, Crusoe has tripled in value. The deal says more about energy than about chips.
Crusoe closed a round of more than $3 billion at a $30 billion valuation, co-led by Atreides Management and Valor Equity Partners, with Mubadala Capital participating.
A five-year, $13 billion contract to supply Jane Street with GPUs and AI infrastructure anchored the raise.
The 3× valuation step-up in 11 months prices in a model where power sourcing, not silicon, decides who wins the AI infrastructure build-out.
Crusoe started in 2018 as a crypto mining operation powered by flared natural gas. Today it builds gigawatt-scale AI data centers and sells compute to Meta, Microsoft, OpenAI and Oracle.
The company reports 4.9 gigawatts of contracted compute capacity and a project pipeline above 40 gigawatts. Reuters carried both figures in June.
Round size, Series F
The round was finalized on September 3. Atreides Management and Valor Equity Partners co-led, and Mubadala Capital joined as a participant. · Bloomberg, 2026
The round that repriced the neocloud
The previous valuation was $10 billion, set last October when it raised $1.38 billion in a Series E. Eleven months later the same company is worth $30 billion. Nothing in the core technology changed in that window. The power contracts did.
Crusoe now sits between three categories. It develops hyperscale data centers like a real-estate builder. It operates an AI cloud platform called Crusoe Cloud. And it sources its own energy, a vertical that the market is pricing as the scarce resource in AI infrastructure.
Hyperscalers have done the math on this too. Meta is under contract to buy capacity at Crusoe sites in Texas and Missouri. Oracle and OpenAI anchor the Abilene campus in Texas, where the first phase of a 1.2 gigawatt site went live a year after construction began.
Valuation step-up
Bloomberg first reported the tripling in July while the round was still in talks. The close confirmed it. · Bloomberg, 2026
Energy is the constraint, not the chip
AI demand stopped being a silicon story somewhere in 2025. Compute supply is now gated by power, water and interconnection timelines. A data center that can secure 500 megawatts of dispatchable power can sell every megawatt of GPU capacity it builds. The company's model attacks the top of that chain first.
Flared gas was the origin. Its earliest rigs ran on methane that oil producers could not sell. The insight survives in the company's current positioning: find energy nobody else is using, convert it into compute, and own the margin between a stranded commodity and a rented GPU.
The pace of progress is constrained by bottlenecks in energy and compute. Crusoe is in the business of activating energy for intelligence.— Chase Lochmiller, CEO and co-founder of Crusoe
Lochmiller said that in October, at the Series E close. The $3 billion follow-on says the market now believes him. A five-year, $13 billion contract with Jane Street, the quantitative trading firm, is the clearest proof yet that this model attracts buyers who care about speed and reliability over everything else.
What the Jane Street contract actually anchors
The contract does double duty. It gives the company contracted revenue that de-risks the build-out, and it signals to other buyers that the platform runs at the reliability standard a trading firm demands. Bloomberg reported that the deal helped pull in the new round.
Sovereign capital meets algorithmic demand
Look at who put money in. Atreides Management is an investment firm with a track record of concentrated tech bets. Valor Equity Partners co-led both this round and the October Series E. Mubadala Capital, the Abu Dhabi sovereign fund's asset manager, participated again.
Sovereign capital stacking into energy-first compute is a recurring pattern this year. Countries that control energy are betting they can host AI infrastructure. A data center developer that locks up power early becomes the entry ticket.
Crusoe also met with Goldman Sachs and Morgan Stanley about a potential near-term IPO, Axios reported last month. A public listing would test whether public markets assign the same premium to energy ownership that private investors do.
Five-year GPU supply deal
The cloud agreement covers clusters of GPUs and AI infrastructure for training and inference over five years. · Reuters, 2026
The risk side of the trade deserves equal weight. Its pipeline of more than 40 gigawatts is mostly unrealized. Interconnection queues, permitting and turbine or gas supply all sit between a signed term sheet and a live data center. As we wrote in September, the AI infrastructure build-out is on track to swallow trillions of dollars of capex through 2050, and not every gigawatt announced will be built.
Can energy-first AI infrastructure keep compounding at this rate?
Probability: 55% — Goldman Sachs and Morgan Stanley are already engaged, and a $13 billion contracted base gives the public markets a revenue story to underwrite.
✅ Arguments for
Jane Street renews, and Meta and Oracle expand, on a five-year horizon.
Public markets are hungry for AI infrastructure exposure with real revenue.
Confirmation criteria: a publicly confirmed S-1 filing, or a multi-gigawatt campus announcement in a new region.
❌ Arguments against
Interconnection delays push pipeline projects past their target dates.
Hyperscalers building their own energy positions could compress neocloud margins.
Disconfirmation criteria: a down round at the next private raise, or a major customer switching capacity to an in-house build.
Jane Street expanding beyond the five-year commitment or doubling capacity.
New gigawatt-scale campuses: Abilene (1.2 GW), Wyoming (1.8 GW) and further sites clearing interconnection.
Sovereign wealth funds beyond Mubadala joining later closings of the round.
A named hyperscaler signing a multi-gigawatt take-or-pay contract with Crusoe.
Development scenarios
🟢 Optimistic scenario (30%)
Implications: energy ownership becomes the reference case for every neocloud, and power assets in AI portfolios re-rate upward.
🟡 Base-case scenario (50%)
Implications: the energy-first model stays differentiated, but the pipeline grows at the speed of grid connections, not at the speed of capital.
🔴 Pessimistic scenario (20%)
Implications: a re-rating toward the $15–$20 billion range, and a longer wait for any public listing.
