On September 17, 2026, AI infrastructure company Crusoe announced the completion of a $3.9 billion Series F financing, with a post-money valuation of $30.9 billion. The round was co-led by Atreides Management, Mubadala Capital, and Valor Equity Partners, with participation from Founders Fund, GIC, Nvidia, the Qatar Investment Authority (QIA), Radical Ventures, and TPG. Dozens of other institutions, including Altimeter, ARK Invest, Baillie Gifford, Fidelity, Salesforce Ventures, T. Rowe Price, and Tiger Global, also joined, making it a clearly oversubscribed financing.
This figure needs a time coordinate to understand its weight: about 11 months earlier, in October 2025, Crusoe completed its Series E financing, raising $1.38 billion at a $10 billion valuation. From $10 billion to $30.9 billion, the valuation rose by more than 200% in less than a year, while the Series F financing amount itself was nearly three times that of Series E. During the same period, valuations across the AI infrastructure sector underwent dramatic repricing, but according to valueaddvc.com, this pace “even exceeded the aggressive revaluations of other companies in this AI infrastructure cycle.”
A Data Center Empire Built from Flared Gas
Crusoe’s starting point was far from its current image as an AI compute provider. In 2018, Chase Lochmiller and Cully Cavness co-founded the company, whose initial business was converting associated natural gas that would otherwise be flared at oilfields into electricity, powering bitcoin mining containers on site. This was an arbitrage logic that both solved oil and gas companies’ waste gas disposal problem and monetized excess energy.
After the generative AI wave arrived, Crusoe transferred this energy-first strategy of “first find cheap electricity, then build compute around it” to the data center sector, and ultimately exited cryptocurrency mining entirely. Today, Crusoe builds its own power plants while also partnering with grid, battery storage, nuclear, thermal, and renewable energy suppliers, and only begins deploying compute assets after securing power supply. This logic inverts the priority of traditional data center site selection—the traditional approach is to first choose a city or talent hub, then solve the power problem; Crusoe’s route is to lock in power first, then decide where to build the facility.
Vertical Integration: A Complete Chain from Electrons to Tokens
Crusoe CEO Chase Lochmiller stated the company’s core proposition in the announcement: “Getting there means controlling the entire infrastructure from electrons to tokens.” This sentence describes a relatively rare vertical integration model in the AI infrastructure field.
Specifically, Crusoe’s value chain contains three layers of revenue sources: first, leasing data center space to customers, with customers bringing their own GPUs; second, renting out Crusoe’s own GPU compute; and third, directly selling inference compute services through the Crusoe Managed Inference product. These three layers cover the complete stack from physical space to computing resources to AI services, so the company does not have to depend on any single segment. Officially disclosed data shows that Crusoe platform’s total contract value (TCV) has exceeded $140 billion, with contracted capacity over 6 GW, of which about 1 GW is already in operation.
The use of this Series F financing points to two types of assets: one is large campus-level, vertically integrated data centers, with the Abilene, Texas campus as its flagship project—this is the first Stargate site jointly used by OpenAI and Oracle, designed and built by Crusoe; the other is modular AI factory units named “Spark,” which can be transported by truck for deployment rather than depending on construction timelines often measured in years. The emergence of Spark represents a direct response to the speed of compute expansion: AI application deployment does not wait for construction cycles, and modular deployment gives compute supply greater elasticity.
Breaking Down the Stakeholders Layer by Layer
Among the disclosed customer list, Crusoe’s coverage has extended to different types of buyers: Meta, Microsoft, and Oracle are tenants of its data center campuses; Cognition, Figure, and Perplexity run on Crusoe Cloud; and in September 2026, it signed a five-year AI cloud services contract worth about $13 billion with quantitative trading giant Jane Street—this contract will provide Jane Street with GPU clusters and AI inference infrastructure, according to Bloomberg. The significance of Jane Street’s appearance is that it breaks the narrative that “AI infrastructure buyers are only AI labs,” drawing financial institutions into the compute consumption side and, with the scale of a single contract, directly contributing a considerable proportion of Crusoe’s total contract value.
For the three cloud giants AWS, Azure, and Google Cloud, the “neocloud” track represented by Crusoe is eroding compute demand that originally belonged to hyperscale cloud providers. Companies such as OpenAI and Meta that already have hyperscale cloud partnership frameworks are also buying compute from vertical players like Crusoe, showing that compute demand from leading AI workloads has exceeded the absorption capacity of a single supplier, and procurement diversification is becoming an engineering decision rather than a negotiating tactic.
For developers and small and medium-sized AI application companies, Crusoe Cloud provides a compute option outside mainstream hyperscale cloud providers, especially against the backdrop of long-term GPU shortages, where one more reliable supply source has direct engineering value. Crusoe Cloud bookings reportedly achieved more than 20-fold year-over-year growth, indicating the authenticity of this demand.
Nvidia’s appearance as an investor in this round is not accidental. As analyzed by valueaddvc.com, Nvidia investing in multiple companies that mainly profit by reselling its chips has become a common pattern in this track, not an isolated case. Critics argue that this amounts to Nvidia partly funding demand for its own hardware, creating a self-reinforcing loop; from another perspective, it is also Nvidia’s way of ensuring its chips are allocated first to high-priority infrastructure players.
Side-by-Side Comparison with CoreWeave
In the neocloud category, Crusoe’s most direct competitors are CoreWeave and Nebius. CoreWeave completed its IPO on Nasdaq in March 2025 at an issue price of $40, corresponding to a valuation of about $23 billion; subsequently, in January 2026, Nvidia completed a $2 billion private placement follow-on investment at $87.2 per share. CoreWeave’s path—the compressed timeline from private financing to listing—has been cited by many as a reference point for AI infrastructure valuation repricing, and Crusoe’s valuation climb in the private market during the same period is highly similar.
The difference between the two lies in the depth of their strategic positioning: CoreWeave’s business model centers on GPU leasing; Crusoe’s integration chain extends further upstream to power assets, which in theory provides lower marginal costs and is also the logic emphasized by Gavin Baker, CIO of lead investor Atreides Management. He said in the announcement: “As AI develops, economic benefits flow to the lowest-cost producer of intelligence. Through a vertically integrated model, Crusoe owns the entire value chain—a structural advantage that compounds as it scales.”
After this financing round, Crusoe’s board simultaneously added three new members: Cloudflare CFO Thomas Seifert, Bill Stein of Primary Digital Infrastructure, and Redwood Materials CEO JB Straubel (co-founder of Tesla). These three names respectively represent public market financial discipline, digital infrastructure operations experience, and materials/energy supply chain expertise, a combination pointing to the capability gaps a company preparing for a public listing needs to fill.
© 2026 Winzheng.com 赢政天下 | 转载请注明来源并附原文链接