Sunday, November 8, 2026
AI 인프라 · 뉴스 & 분석
홈 › 데이터센터 › 리포트
데이터센터 · 리포트

Google가 Crusoe 단독으로는 완공하지 못한 와이오밍 데이터센터 캠퍼스를 인수했다.

하이퍼스케일러가 중단된 대규모 AI 캠퍼스를 인수한 것은, 완공 리스크와 자본 수요로 인해 생산능력이 최대 구매자들의 손으로 넘어가고 있음을 보여준다.
업계 전문지Slicast · 2026년 11월 7일 09:59 UTC · 미국 · 출처: Startup Fortune
중요도 92

Crusoe spent two years lining up permits, power, and a site in Cheyenne for a data center campus that could scale past a gigawatt. Google has now bought the whole thing out from under it.

On November 6, Crusoe said Google had acquired its southeast Wyoming data center site, taking over a project the AI infrastructure company had championed as one of the biggest in the country. Bloomberg reported the same day that Google is taking control of the Cheyenne-area campus, with plans to invest roughly $17 billion to build it out. Crusoe's own newsroom post described the deal as the conclusion of its "pre-development" work: picking the site, securing government approvals, and lining up a power strategy. Other developers typically skip straight past this groundwork when a hyperscaler takes over.

The change was not sudden. It is the final step in a process that has been unfolding publicly since June. Crusoe and energy partner Tallgrass announced the Wyoming campus in July 2025. They proposed a 600-acre site in the Switchgrass Industrial Park that could eventually scale to 1.8 gigawatts, tied to Tallgrass's roughly $7 billion natural gas power hub next door. By June 2026, Crusoe was calling the project a "pause." Bloomberg had already reported the underlying story: Google, the customer Crusoe needed to anchor the project, raised concerns about cost and construction timelines and pushed Crusoe out. Google's involvement remained officially unconfirmed until July. Then Wyoming utility filings and local reporting from the Cowboy State Daily revealed that Jupiter Star Holdings, a Google subsidiary, was behind what had by then become a 2.7-gigawatt campus renamed Project Tembo. November's announcement makes the ownership change contractual.

Crusoe built its business on vertical integration: generating power, building data centers, and running the cloud layer on top. That is why Wyoming was such a blow. The company told Bloomberg it remains a partner with Google elsewhere, developing a separate campus for Google in Armstrong County, Texas, which has not run into the same friction. In Wyoming, however, Crusoe was the developer and Google was the tenant. When the tenant objected to the price or the schedule, the developer lost the project. Crusoe has been raising capital as if such a outcome were unlikely elsewhere. This year it closed a $3.9 billion Series F at a $30.9 billion valuation, backed by Nvidia, Mubadala Capital, and the Qatar Investment Authority. That valuation was up from $10 billion just ten months earlier. A company that roughly tripled its valuation in under a year still could not keep its anchor tenant satisfied on a single deal.

The timing places the deal within a broader debate about whether AI infrastructure spending has outpaced demand. Months earlier, Microsoft walked away from a roughly $3 billion cloud lease with Oracle, reportedly because Oracle's infrastructure lacked the FedRAMP certification needed for government workloads, a characterization Oracle disputed. That collapse is tied to the same data center buildout frenzy that has made Wall Street nervous about overbuilt capacity. Even so, Microsoft still plans close to $190 billion in capital spending for fiscal 2026, so the industry is not pulling back on dollars. It is becoming more selective about who builds the capacity it pays for.

The Wyoming deal itself is less revealing than what it suggests about the market. Google did not walk away from an AI data center because it had too much capacity. It parted ways with a developer and then bought the site directly, which is the opposite of retreat. Hyperscalers increasingly want land, permits, and power contracts in-house rather than leased through a middleman, even a well-funded one like Crusoe. For the rest of the AI infrastructure sector, the lesson is that companies that raised billions on the assumption that hyperscalers would keep relying on third-party developers should take note. Google did not decide it needed less compute. It decided it did not need Crusoe to get it.

원문 보기
Google가 Crusoe 단독으로는 완공하지 못한 와이오밍 데이터센터 캠퍼스를… · Slicast