Crusoe confirmed on Friday that it has sold an unfinished data center project in southeastern Wyoming to Google. The Denver developer had announced the site last year with Tallgrass, the Blackstone-backed infrastructure company, and had intended to grow it to 10 gigawatts of power capacity. Google, which Crusoe had spent months courting as the anchor tenant, is now taking over construction itself and says it will put $17 billion into the campus.
Neither company disclosed the purchase price. The reversal is the clearest illustration yet of how hard it has become for independent developers to carry a multi-gigawatt project from land deal to live servers, even ones with blue-chip backers and a roster of AI customers.
The customer became the owner
Google did not walk away from the site so much as take it. According to reporting on the transaction, the company raised concerns about cost and the construction timeline under Crusoe's management before agreeing to buy the project outright. TechCrunch first confirmed the sale, which Bloomberg had reported was under discussion in June.
The two companies are still working together. Crusoe is developing a separate data center for Google near Amarillo, Texas, so the Wyoming exit reads less like a rupture than a reassignment of risk. Google keeps the shell and the interconnection queue; Crusoe keeps a customer relationship it can bill against.
Crusoe was founded eight years ago by Chase Lochmiller, now chief executive, and Cully Cavness, as a cryptocurrency miner running on natural gas that would otherwise have been flared. It was valued at $30.9 billion in its most recent round, a $3.9 billion raise in September that also added three board members, among them Redwood Materials founder and Tesla board member JB Straubel.
Ten gigawatts is a power problem
A 10-gigawatt campus is measured in years of interconnection work, not in construction schedules. The constraint is rarely concrete. It is transformers, switchgear, turbines and the queue to connect. Local opposition adds a second clock: several jurisdictions have already capped or delayed projects that were supposed to be energised in 2027, a squeeze we covered when local bans held up part of that pipeline.
Crusoe's other commitments show what a developer has to juggle at once. Its campus in Abilene, Texas, is being built in phases for Oracle and OpenAI. Next door it is putting up a 900-megawatt site for Microsoft, with the first building expected online in mid-2027. Meta is a customer, and Jane Street signed a $13 billion, five-year cloud deal with the company.
Those contracts are the reason Crusoe could absorb the Wyoming loss. They are also the reason the loss matters: a company that leases shell space, rents out its own GPUs and sells inference is only as good as its delivery dates. Handing a site to the customer and keeping the Amarillo build keeps the revenue line and removes the schedule risk from its own balance sheet.
What Google gets for $17 billion
Google is not short of capacity plans. It has committed tens of billions to data center capital spending, and the same race for power has pushed buyers toward projects that already have a grid connection, as with SpaceX's plan for ten gigawatts of its own. Buying a partially built campus in Wyoming shortens the permitting timeline in a way that a greenfield announcement cannot.
The open question is who fills it. Google's own TPU fleets and its cloud customers will compete for the same megawatts, and a site that was engineered around a tenant that never signed may need rework before either moves in. Crusoe, for its part, has now shown that the fastest way to solve a stalled project is to sell it to the company it was built for.