Crusoe sells troubled Wyoming data center project to Google

The cost of building large-scale AI infrastructure has just been revealed, and it's a lot higher than anticipated.

Crusoe, an eight-year-old Denver-based data center developer, has sold an unfinished data center project in southeastern Wyoming to Google. The sale comes after Google reportedly raised concerns about the project's cost and construction timeline under Crusoe's management. Crusoe had initially announced the project last year with Blackstone-backed Tallgrass, with plans for up to 10 gigawatts of power capacity. While Google did not disclose the purchase price, it announced a separate $17 billion investment to develop the data center campus. This deal comes after Crusoe secured a $3.9 billion funding round in September, despite this setback. Crusoe continues to develop another data center for Google in Amarillo, Texas, and has significant ongoing projects for Oracle, OpenAI, Microsoft, and Meta, including a recently reported $13 billion cloud deal with Jane Street.

The sale of the Wyoming project to Google underscores the immense capital and execution challenges faced by independent data center developers in scaling massive AI infrastructure projects. Despite Crusoe's recent $3.9 billion funding round and its continued work with Google on other sites, this transaction highlights the risks associated with large-scale development, especially when facing cost overruns and timeline complexities. It suggests that established hyperscalers like Google may be reassessing their reliance on third-party developers for flagship projects, potentially favoring direct investment and control.

This event may signal a broader trend where major cloud providers are internalizing more of their critical infrastructure development to ensure tighter control over costs, timelines, and technological integration. For companies like Crusoe, it means a heightened need to demonstrate robust project management and cost-efficiency, particularly to attract and retain partnerships with these demanding clients. The future of independent AI infrastructure development might depend on a more focused approach, perhaps on specialized solutions or smaller-scale deployments, rather than ambitious, multi-gigawatt projects.