The Story
Crusoe has raised more than $3 billion at a valuation of about $30 billion, co-led by Atreides Management and Valor Equity Partners, with Mubadala Capital participating. The valuation is on a post-money basis, so it includes the capital just raised.
The round follows a five-year cloud contract with the quantitative trading firm Jane Street, reported at around $13 billion, under which Crusoe supplies GPU clusters and the supporting infrastructure for AI training and inference. Jane Street is the company's highest-profile cloud customer to date, and the contract is understood to have drawn additional investors into the raise.
That headline figure needs qualifying. The reported $13 billion depends on two separate contract extensions plus an option for additional compute capacity, which makes it a ceiling if everything proceeds as planned rather than revenue Crusoe has booked.
Crusoe was launched in 2018 by Chase Lochmiller and Cully Cavness to capture flared natural gas at oil wells and convert it into power for portable data centres, initially running bitcoin mining. It sold that mining unit to NYDIG in early 2025 and now designs, builds and operates full AI campuses rather than leasing space in other people's buildings. Its customers include Meta, Microsoft, Oracle and OpenAI, and its Abilene, Texas campus, developed for OpenAI and Oracle, is the first phase of the Stargate project.
The company raised $1.38 billion at a valuation above $10 billion in October, ten months ago. It has also been meeting investment banks, among them Goldman Sachs and Morgan Stanley, to discuss a possible near-term listing.
Why It Matters
Jane Street is the part of this worth thinking about, and it has little to do with the size of the cheque.
Almost every large AI infrastructure contract so far has come from one of three places: a hyperscaler, a model lab, or an enterprise buying capacity it does not yet know how to use. Jane Street is none of those. It is a proprietary trading firm buying hyperscaler-scale compute to train and run its own models, and that is a category of customer nobody was underwriting a year ago.
The reason it makes sense is attribution. Most enterprises cannot say what a better model is worth to them, which is why enterprise AI spending has been slow and hesitant. A trading firm can. A model that improves execution by a basis point produces a number that shows up in the profit and loss that quarter. Firms with proprietary data and direct financial attribution are the buyers who can justify compute at this scale without a strategy deck, and if Jane Street has done the arithmetic, its peers are doing it too.
The other thing worth naming is that Crusoe never really pivoted. Flare gas to bitcoin to AI training looks like two reinventions, but the underlying business has been constant: put compute where stranded energy is, because both workloads are energy-hungry and tolerant of latency. The mining unit went to NYDIG in early 2025. The competence stayed exactly where it was.
The Strategic Read
The valuation is doing something specific, and it is worth being precise about what.
Ten months ago Crusoe was worth $10 billion. It is now worth $30 billion, and the thing that changed in between is a contract whose full value depends on extensions nobody has exercised. Strip out the extensions and the option and the committed portion of the Jane Street deal is a fraction of $13 billion. The round was raised on the headline. That is neither unusual nor improper, since infrastructure is routinely financed against contracted revenue, but the market is pricing a ceiling as though it were a floor.
The composition of the round says what it is. Atreides Management is a hedge fund, and Crusoe has been meeting Goldman Sachs, Morgan Stanley and others about a listing. This is a crossover round, and crossover rounds are priced with an IPO in view rather than against the current business. Thirty billion dollars is a number designed to be defensible at listing, not a considered assessment of today's revenue.
Customer concentration is the structural exposure. Meta, Microsoft, Oracle, OpenAI, Jane Street: a handful of names, each enormous, and several of them building their own capacity. Crusoe constructed the first phase of Stargate for OpenAI, a customer whose own compute commitments are financed against revenue it projects rather than earns. The neocloud model works while hyperscalers are capacity-constrained. It compresses the moment they are not, and every one of those customers is spending heavily to stop being constrained.
Energy remains the genuine moat, and it is why Crusoe is not simply a GPU reseller. Siting, generation and power procurement are far slower to replicate than buying chips, which is what lets the company build campuses rather than rent racks. Its purchase of roughly $400 million of AMD accelerators points the same way: a company hedging its dependence on a single chip supplier is one planning for the next cycle rather than this one.
The question this round does not answer is what a $30 billion valuation looks like if Jane Street declines its extensions.
For daily, sharp analysis of the biggest moves in the Indian business and startup ecosystem, follow StartupFox.
