Crusoe Raises $3.9 Billion at a $30.9 Billion Valuation to Build AI Factories at Any Scale

Crusoe said on September 17 that it raised $3.9 billion in a Series F round at a $30.9 billion post-money valuation, a step that turns an eight-year-old company which began by mining bitcoin on flared natural gas into one of the most valuable private AI infrastructure businesses in the world. The company described the round as oversubscribed and said it had closed an initial tranche.

Related reading: OpenAI Nears a $30 Billion Round at $1.25 Trillion, $150 Billion Below Its Earlier Target · DeepSeek Reportedly Raises $15 Billion at a $74 Billion Valuation, Two Days After a US Federal Order.

The round was co-led by Atreides Management, Mubadala Capital and Valor Equity Partners, with Founders Fund, GIC, Nvidia, the Qatar Investment Authority, Radical Ventures and TPG among the participants, according to Crusoe and TechCrunch. Those lead names had circulated for weeks before the announcement. Bloomberg reported on September 3 that Crusoe had raised more than $3 billion at a valuation of roughly $30 billion, citing people familiar with the deal, and Reuters confirmed the final figure on September 17. Crusoe says the capital will fund its existing campus program and the build-out of its own AI factories, from large vertically integrated sites to modular Crusoe Spark units, plus growth at Crusoe Cloud.

The pitch: own the electrons and the tokens

The strategic case Crusoe makes is vertical integration. Most data center developers buy land, chase grid connections and hire contractors, treating power as a constraint to solve after a site is chosen. Crusoe starts with power. It originates and manages energy directly, develops its own plants and pairs them with grid, battery, nuclear, thermal and renewable providers. It also manufactures long-lead electrical components at its own facilities in Colorado, Oklahoma and Louisiana, which the company says compresses construction timelines.

Crusoe describes the model as controlling infrastructure from electrons to tokens. That phrase, from co-founder and chief executive Chase Lochmiller, is the investment thesis in four words: if the cost of intelligence keeps falling, the margin accrues to whoever owns the cheapest path from a power source to a running model.

The framing is why the round reads less like a typical venture deal and more like project finance. A $140 billion total contracted value and gigawatt-scale power commitments are the language of utilities and independent power producers, not of software startups. AI compute is being financed like energy infrastructure, with long-duration contracts standing behind the capital.

What the $140 billion number means

Crusoe says it holds more than $140 billion in total contracted value across its vertically integrated platform, serving AI-native, hyperscaler, frontier-model and enterprise customers. It also reports more than 6 GW of gross contracted capacity across data centers and cloud, including 1 GW delivered and operational, and more than 20 times year-over-year growth in Crusoe Cloud bookings so far in 2026. Its managed inference product, launched late last year, has contracted more than $100 million in annual recurring revenue, the company says.

Those are company-reported figures, and the distinction matters. Total contracted value is not revenue. It is the aggregate value of signed agreements over their terms, which can run for many years, and it depends on customers taking and paying for capacity on schedule. Crusoe has not published contract-by-contract detail, audited financials or a breakdown of how much of the contracted value is tied to firm take-or-pay commitments rather than more flexible arrangements. The $140 billion is best read as the company's own measure of its order book, and it is large enough that the number alone drove much of the coverage.

The most recent independent benchmark came in June. Crusoe said then that it had contracted 4.9 GW of AI infrastructure across its data centers and Crusoe Cloud, with a total development pipeline exceeding 40 GW. That pipeline includes contracted projects, sites under active tenant negotiation and sites in advanced development, so it measures ambition and option value rather than committed demand. The jump from 4.9 GW contracted in June to more than 6 GW now, against the same 1 GW of operational capacity, suggests the growth sits mostly in newly signed contracts rather than newly energized sites. That gap between contract and construction is the operational problem the Series F is meant to close.

Energy first, buildings second

Crusoe's flagship is the Abilene, Texas campus, built over two years and used by OpenAI, whose Astra model was trained there, according to Crusoe. Dealroom describes the site as roughly 2.1 GW of planned capacity. Crusoe has also said it broke ground on a second, 900 MW Abilene campus for Microsoft, which would take its Abilene footprint above 2 GW. The company's June disclosure named a 1.2 GW Abilene campus purpose-built for Oracle and a fifth campus in Missouri, alongside additional Texas sites.

Microsoft, Meta and OpenAI are the customer names most consistently attached to Crusoe. TechCrunch lists Meta, Microsoft and Oracle, while Bloomberg has reported the company does business with OpenAI, Microsoft and Meta. Crusoe Cloud counts Cognition, Figure and Perplexity among its customers, and Perplexity signed a multi-year partnership across the model lifecycle in September.

Power is the binding constraint across the industry, and sites are increasingly chosen for megawatts rather than fiber routes. Crusoe's pitch is that it can find stranded or underused energy, build generation or storage alongside it and move faster than developers who wait in interconnection queues.

Truck-ready AI factories

The newer half of the story is modular. Crusoe manufactures Crusoe Spark, a compact data center it builds in U.S. factories, loads onto flatbed trucks and connects to large power sources, sometimes without building a conventional campus at all. The company says Spark shortens field construction from years to weeks, lowers cost relative to conventional methods and lets customers add capacity incrementally as demand arrives.

The modular approach reduces exposure to the two scarcest inputs in data center construction, skilled labor and long-lead electrical gear. It also offers a partial answer to local opposition, which has become a real constraint as multi-gigawatt projects draw protests over power prices, water use and land. A factory-built unit that arrives by truck is easier to site and easier to remove than a sprawling campus.

It is also a hedge on the demand mix. Training the largest models still favors enormous, tightly coupled clusters. Serving those models to users, known as inference, can run on far fewer chips in more distributed locations. Lochmiller has argued that customers do not need an Abilene to run inference, according to Dealroom. If inference becomes the dominant share of AI compute spending, modular capacity aimed at it could be a large market.

The neocloud race

Crusoe is one of a group of neoclouds that rent GPU capacity and AI cloud services, competing with CoreWeave, Nebius and others, and increasingly with hyperscalers building their own capacity. The comparison set matters for how investors judge the round. CoreWeave is public and trades at a large multiple of revenue, while Nebius is building aggressively, including roughly 120 MW of leased capacity in Israel, according to Globes. Crusoe's argument is that it earns at several layers: it develops and leases data centers, rents the GPUs inside them and sells compute and inference to run models. Lochmiller has put it plainly, saying the company sells data centers, GPUs and tokens.

That three-layer model is a margin story and a risk story. Owning more of the stack can raise returns per dollar of capital, but it also makes Crusoe an operator of power plants, factories, data centers and a cloud platform at once. Each layer carries its own execution risk.

Long-duration contracts and the Israel footprint

The clearest evidence of the infrastructure-finance turn is contract structure. Crusoe has signed long agreements that look like offtake deals. Bloomberg reported in September that the company signed a roughly $13 billion, five-year cloud contract to supply quantitative trading firm Jane Street with GPUs and AI infrastructure. In Israel, Globes reported that Crusoe's leases across three sites would bring its total planned investment there to about $10 billion over 10 to 15 years, with total leased capacity of 150 MW. A Crusoe executive framed Israel as potentially one of its largest markets.

Crusoe has also been preparing the ground for public markets. Axios reported in August that the company had met with investment banks, including Goldman Sachs and Morgan Stanley, to discuss a potential IPO. The Series F is private for now, but its size and structure point toward an eventual listing or large-scale debt financing, the way utilities and infrastructure operators raise capital.

The round's lineage supports that reading. In October 2025, Crusoe raised a $1.375 billion Series E at a valuation above $10 billion. The new mark roughly triples that in about ten months, and the disclosed use of proceeds is physical build-out rather than product expansion.

What to watch

Three things will test the bull case. First, the conversion rate from contracted capacity to energized, revenue-producing capacity. The company reports 1 GW operational against more than 6 GW contracted, and closing that gap is the whole game. Second, whether the $140 billion in contracted value converts to cash on schedule, since long-duration agreements can be renegotiated, delayed or cancelled. Third, the capital intensity treadmill: every new gigawatt requires more power, more equipment and more financing, and Crusoe has now committed to being a developer, a manufacturer, a landlord and a cloud provider at the same time.

For the wider market, the Crusoe round is a signal about how the AI buildout will be paid for. The largest rounds are no longer funding software. They are funding power, buildings, chips and multi-year contracts, underwritten on the assumption that demand for compute stays scarce for years. From Series E to Series F, Crusoe's valuation tripled in about ten months. Whether that holds depends less on the next model release than on whether the megawatts show up.

Cover: The Crusoe logo. Image: Crusoe (company brand asset).