Crusoe Neocloud, September 2026: $3.9B Series F, $13B Jane Street Deal, and a Nuclear Power Bet

Crusoe closed a $3.9 billion Series F at roughly a $30 billion valuation in September 2026, backed by a reported $13 billion Jane Street compute agreement and a new $65 million Thinking Machines Lab inference contract, as it pursues a gigawatt-scale buildout underpinned by nuclear power ambitions.

Thinking Machines Lab's $65 million inference contract with Crusoe would, in other circumstances, register as a typical neocloud commercial win. It arrives instead as the latest in a sequence of deals that, taken together, has repositioned Crusoe from a startup-tier GPU rental service into one of the most heavily capitalized AI infrastructure platforms in the world. The announcement comes alongside the close of Crusoe's Series F — a round of more than $3.9 billion that set the company's valuation at roughly $30 billion, up from approximately $10 billion reported in August 2025, with NVIDIA among the participants. The speed of that rerating — nearly a tripling in around ten months, according to contemporaneous reporting — reflects both a market premium on contracted AI compute capacity and the specific profile of customer agreements Crusoe has assembled.

The company's origins are worth recalling as context. Crusoe began as an operator that monetized stranded natural gas — originally through Bitcoin mining — and built an early, practical expertise in the economics of cheap, remote, and otherwise-wasted power. That thesis carried directly into its AI infrastructure pivot. The $13 billion, five-year compute agreement with Jane Street, reported in September 2026, and the 1.6 gigawatt capacity agreement with Meta covering Texas and Missouri sites give the company a demand anchor that most neoclouds its size do not hold. In the context of an August 2026 sector analysis that placed CoreWeave and Nebius ahead of Crusoe on contracted capacity leadership — and noted spot pricing for NVIDIA B200s at $6.69 per GPU-hour from Lambda — these long-duration contracts provide a margin buffer that spot-market competitors cannot easily replicate.

On the supply side, Crusoe's differentiation is most visible in its approach to power. The Lancium partnership announced in July 2026 targets a one-gigawatt AI campus in Childress, Texas. Blue Energy has filed for a nuclear reactor at a Texas power plant specifically to supply Crusoe's data centers, and the Aalo microreactor — backed by Crusoe — achieved criticality on July 4, 2026, with a 2027 nuclear AI factory demonstration on the roadmap. Heron Power, a transformer manufacturer in which Crusoe holds a stake, has selected Morgan Hill, California, for its first large-scale factory targeting 10,000 units per year — an acknowledgment that transformer supply is as binding a constraint as GPU availability in a multi-gigawatt buildout. ON.energy's deployment of five gigawatts of AI uninterruptible power technology across Crusoe's hyperscale campuses rounds out an unusually complete infrastructure stack for a private company. Gartner placed Crusoe as a Visionary in its 2026 Magic Quadrant, while separately predicting that neocloud providers as a class will capture 20 percent of a $267 billion AI cloud market by 2030.

The scale of the ambition carries commensurate execution risk. A reported $10 billion commitment to Israeli data centers over ten to fifteen years layers geopolitical and regulatory complexity onto an already demanding multi-site Texas program. Nuclear power timelines have historically run well past initial projections; the Aalo criticality milestone is a genuine technical achievement, but converting it into reliable commercial grid supply by 2027 is a different order of undertaking. The neocloud competitive environment remains unforgiving: pricing pressure in B200 spot markets signals that capacity is still growing faster than near-term demand in some segments, and CoreWeave is the category benchmark against which Crusoe will inevitably be measured. Reports of IPO preparation, first surfacing in August 2026, introduce public-market scrutiny that private investors did not apply, and the revenue recognition structure of long-duration compute contracts like the Jane Street agreement will face detailed examination at any roadshow.

Three signals are worth watching. The first is whether Crusoe files formally for an IPO and what its prospectus discloses about revenue recognition on multi-year contracts — whether the contracted backlog translates into a cash flow profile that justifies a $30 billion public-market entry for incoming shareholders. The second is the nuclear timeline: a successful 2027 Aalo demonstration at a Crusoe facility would validate the energy-differentiation thesis in a way that no battery or renewable arrangement can replicate, while delay would expose a continued dependency on Texas grid capacity already under pressure from competing data center loads. The third is the evolution of the inference customer base: the Thinking Machines Lab agreement establishes Crusoe as a managed inference vendor for open-model workloads, a segment structurally distinct from the long-duration compute leases that have driven its valuation to date. Whether inference revenue scales alongside the capital-intensive GPU and power buildout — or remains a smaller complement to it — will shape the revenue mix at any IPO and determine the company's positioning in a market that has not yet sorted its long-term winners.