In October 2025, Crusoe raised $1.38 billion at roughly a $10 billion valuation. Ten months later, on September 3, the Denver company finalized more than $3 billion at approximately $30 billion post-money — a tripling in under a year, first reported by Bloomberg. Atreides Management and Valor Equity Partners co-led. Mubadala Capital, the asset-management arm of Abu Dhabi's sovereign wealth fund and an existing backer, joined again.

The proximate cause is not mysterious. Days earlier, Crusoe locked in a five-year cloud contract with Jane Street worth roughly $13 billion, supplying the quantitative trading firm with GPU clusters and AI infrastructure through Crusoe Cloud. It is the company's highest-profile cloud customer to date, and by Bloomberg's account it is what pulled additional investors into the round. A single contract worth more than four times the equity being raised changes the underwriting question from "will demand materialize" to "can you build fast enough."

What makes the arc unusual is where it started. Chase Lochmiller and Cully Cavness founded Crusoe in 2018 to capture natural gas that oil producers were flaring off at the wellhead — gas with no pipeline to reach and no buyer to sell to — and burn it in generators powering containerized bitcoin miners parked in the oilfield. The company sold that mining unit to NYDIG in early 2025. What survived the pivot was the part that turned out to matter: the ability to source and stand up electrical power in places where no one had bothered to build.

From flare stacks to five gigawatts

Crusoe now describes itself as "the AI factory company," and the scale is no longer speculative. As of June 9, it had 4.9 gigawatts of AI infrastructure under contract across its data centers and cloud platform, with a total development pipeline — contracted projects, sites in tenant negotiation, and sites in advanced development — exceeding 40 GW.

"The demand from the world's leading technology companies for AI infrastructure — quickly and at scale — has never been greater, and Crusoe is uniquely positioned to meet it," Lochmiller said in announcing that milestone. "We work shoulder to shoulder with our customers to move from vision to operational infrastructure at the speed AI demands."

The physical portfolio spans five hyperscale campuses. The flagship is 1.2 GW in Abilene, Texas, purpose-built for Oracle and the first phase of the Stargate project; two buildings are operational and six more are under construction. Crusoe broke ground on a second Abilene campus, 900 MW for Microsoft. Two more Texas sites and one in Missouri are in various stages of site work. TechCrunch reports the company counts Meta, Microsoft and OpenAI among its customers.

The differentiator Crusoe sells is sequencing. Conventional developers treat power procurement and construction as serial steps; Crusoe co-develops them, manufactures long-lead electrical components at its own plants in Colorado, Oklahoma and Louisiana, and ships prefabricated gear ready to install. It has stacked energy partnerships accordingly — 5 GW of battery-based "AI UPS" with ON.energy in July, a nuclear partnership with Aalo Atomics the same month, and a $300 million round at Boom Supersonic last December to build natural gas turbines for Crusoe sites.

The neocloud math, and its fault lines

Synergy Research Group puts total neocloud revenue above $25 billion for 2025 and projects the market approaching $400 billion by 2031 — roughly 58% compound annual growth. That is the number underwriting a 3x markup on a private company in ten months, and it is why CoreWeave, Nebius and Lambda are all either public or mid-raise.

But the economics are unforgiving in a specific way. Neoclouds must commit billions to chips, land and substations before they have contracts that conventional lenders will finance against. That gap is precisely what Nvidia's AI Compute Partnership Program was built to close: launched in July with roughly $36 billion in commitments and Sharon AI and Firmus Technologies as initial partners, it offered credit support and a guarantee to lease unused GPU capacity if a provider couldn't fill it — in exchange for as much as 50% of revenue above a threshold.

In late August, Nvidia suspended parts of it. Reporting indicates partners balked at restrictions on how they could rent out capacity, and some Nvidia employees flagged antitrust exposure in a structure where one company sells the chips, finances the purchase, and shares the revenue. Neocloud shares fell on the news. Nvidia says the model remains in place and is evolving.

That retreat matters to Crusoe twice over. Nvidia was an investor in its Series E, and the circularity critique — capital flowing from chipmaker to cloud to chip purchase — is exactly what public-market investors will price if Crusoe lists. Axios reported last month that the company has met with JPMorgan, Goldman Sachs, Morgan Stanley and Bank of America about a near-term IPO.

Power remains the binding constraint, and Crusoe's own partnerships concede it. "The next generation of AI infrastructure demands power that is reliable, abundant, clean and at scale," Cavness said in July. Reliable and abundant is the easy half. At scale, on the timeline contracted, is the hard half.

What to watch

Three things. Whether the Jane Street contract is a template — financial firms buying frontier compute directly rather than through hyperscalers — or a one-off. Whether Crusoe's 40 GW pipeline converts to contracted capacity at anything like the pace implied by a $30 billion mark. And whether an IPO filing arrives, because that is the moment the company must disclose what its GPU depreciation schedules, customer concentration and capital commitments actually look like. Until then, the valuation rests on contracts, not on concrete.

“The demand from the world's leading technology companies for AI infrastructure, quickly and at scale, has never been greater, and Crusoe is uniquely positioned to meet it.”
— Chase Lochmiller, Co-founder and CEO, Crusoe
$30B
Post-money valuation
$13B
Jane Street cloud contract
4.9 GW
Contracted AI capacity
3x
Valuation gain in ten months