Wyoming’s upcoming AI campus could draw five times more power than its homes

Mayor Patrick Collins has backed a proposal from infrastructure specialist Tallgrass and AI-compute developer Crusoe for a new data center campus. The campus would open at 1.8 GW and could scale to 10 GW of IT load. This would surpass the combined electricity consumption of every home in the state.

At the initial stage, the facility is expected to draw an estimated 15.8 TWh per year. This is five times Wyoming’s residential demand. When fully built, it could use 87.6 TWh, more than twice the state’s annual generation. Calculations cited by Ars Technica reach similar conclusions. They note that a single gigawatt can supply roughly one million U.S. homes.

Because the public grid cannot handle such a draw, Tallgrass plans to pair dedicated gas-fired generation with renewable assets. This marks a major shift for Wyoming. Currently, the state exports nearly 60 percent of the electricity it produces. Governor Mark Gordon has welcomed the project for the boost it could give to local natural-gas producers.

The site is located just south of Cheyenne, off U.S. Route 85. State and county regulators still need to grant approvals. Cheyenne already hosts Microsoft data centers and an almost-finished $800 million Meta campus. These companies are drawn by the region’s cool, dry climate and ample energy infrastructure.

Tallgrass and Crusoe have not named an anchor tenant. This has prompted speculation that the facility could become part of OpenAI’s “Stargate” build-out. A Crusoe spokesperson declined to confirm or deny that link. The developer already partners with OpenAI on a one-gigawatt campus in Abilene, Texas. The company describes it as the world’s largest data center site. Crusoe also plans to secure an additional 4.5 GW of capacity with Oracle.

If the Cheyenne project proceeds on its promised “sooner rather than later” schedule, Wyoming could become one of the world’s largest single locations for AI computation. At the same time, the state would be testing its ability to balance energy exports with rising domestic demand for machine-learning workloads.