Google Takes the Wyoming Bid — and Reveals the Real Price of AI Capex
Two days after voters in multiple U.S. states rejected the deregulated expansion of data centers at the polls, Google announced on Thursday, November 5, a $17 billion investment in its first data center in Wyoming. The confirmation came on Friday with the news that Google had officially acquired the land and all the pre-development infrastructure built by Crusoe, the AI infrastructure developer that failed to find customers for the project before being replaced by the search engine.
What the Google announcement reveals is much more than another data center. With a budget that exceeds the annual GDP of 150 countries, the Tembo project — as it was internally named — is a clear symptom of where the artificial intelligence infrastructure race is headed. And the short answer, confirmed by the size of the bet, is: to a place with no limits.
From Crusoe to Google: the handover
The story of Project Tembo began almost a year before the official announcement. In July 2025, Crusoe — the Denver company founded in 2018 by Chase Lochmiller and Cully Cavness, which started as cryptocurrency mining powered by wasted natural gas — announced alongside Blackstone-backed energy company Tallgrass Energy a data center campus that would span two parcels: 600 to 716 acres for the data center and 659 acres for the power plant. The ambition was ambitious even by industry standards: scale up to 10 gigawatts of power capacity.
However, throughout the first half of 2026, things began to go wrong. Bloomberg reported in June that Google had raised concerns about costs and timelines under Crusoe's management. The search engine tried unsuccessfully to become a customer of the project rather than a partner — and ultimately took control of the land. Google did not disclose how much it paid Crusoe for the unfinished site.
Crusoe exited the situation financially unscathed: it raised $3.9 billion in a September round — with three new board members, including JB Straubel, founder of Redwood Materials and Tesla board member — and continues building another campus for Google outside Amarillo, Texas, in Armstrong County. The Wyoming project was only one of the company's nineteen operations. But the exit was revealing: independent developers can select sites, obtain permits, and design energy strategies. They just can't build the data centers that really matter.
Project Tembo: four buildings, hybrid power, and carbon capture
According to Google's official announcement and a presentation made at a community meeting in Cheyenne in September, the Tembo campus includes four data center buildings, effluent evaporation lagoons, a logistics center, two network structures, and a two-story office building in the style of its San Francisco headquarters. Google's projected timeline anticipates mass grading starting late 2026, vertical construction in 2027, and full completion in the first half of 2031.
During the construction phase, the project is expected to support more than 8,800 jobs statewide, including 2,500 specialized jobs on the construction site, and generate approximately $610 million in average annual wage income. Once operational, Google estimates creating 250 permanent direct jobs.
The campus energy mix is particularly revealing. Unlike data centers from other companies that bet on renewable energy or nuclear power, the Tembo project will be powered by natural gas turbines operated by Tallgrass Energy and Black Hills Energy, combined with fuel cells and grid connection. Google also contracted carbon capture and storage (CCS) capacity equivalent to nearly 1 gigawatt — the second CCS project contracted by Google in the United States, following the first announced last fall. The company plans to offset emissions through a regional CO2 sequestration hub developed by Tallgrass.
The political and economic context
The timing of the announcement is not innocent. The voting on November 4, in the U.S. midterm elections, resulted in express rejection of data centers in several states, including Wisconsin and Ohio — traditionally business-friendly states. Local resistance has been growing as communities discover the hidden cost of artificial intelligence: massive water consumption, electrical demand that overloads the local grid, and changes in municipal revenue bases.
Wyoming, on the other hand, offers a politically favorable environment. The state has a tradition of supporting energy and mining projects, and Google explicitly highlighted that the project will generate hundreds of millions of dollars in property taxes for Laramie County and School District 1. The company also announced community investments in K-12 education, workforce development, and water replenishment initiatives — 165 active stewardship projects across 97 watersheds globally.
What this means for the AI ecosystem
The $17 billion investment in Wyoming represents more than 1% of Google's total annual revenue. Put in perspective, it is approximately the same annual training budget as the world's largest AI models, and five times larger than the infrastructure spending by any single competitor in the same state. The Tembo project is, without exaggeration, one of the largest single investments ever made in computing hardware at a single location.
But the most important message may lie in what Google decided not to do. By taking over Crusoe's project instead of developing a greenfield site, Google showed that it prefers to absorb pre-developed infrastructure rather than risk long timelines. Crusoe had completed the site selection, government approvals, and the hybrid energy strategy — and Google bought that prior work without spending months on entitlements.
This is the logic of a search engine that is investing billions to ensure its AI — Gemini — can process more tokens, faster, for more people, without relying on partners who can't deliver. Crusoe can raise $3.9 billion and have Jane Street signing a $13 billion contract, but if you can't build the data center the customer wants on the timeline they need, the customer simply changes contractors.
The question that remains is: how many more 17-billion-dollar investments does Google need before it has enough processing capacity for whatever comes next? Because the current answer — four buildings, 10 gigawatts, completion in 2031 — is already obsolete before the first brick is laid.
Sources: Google, TechCrunch, Crusoe, Bloomberg
✓ Independent sources cross-checked and verified before publishing
