Crusoe and Lancium pick Childress, Texas for a 1GW AI campus

Crusoe and Lancium will develop a 1.0 gigawatt AI data center campus on 270 acres in Childress, Texas, a panhandle county whose entire population is 6,743. Nothing about the labor pool or the land basis explains that choice. An ERCOT-approved gigawatt interconnect does.

Why it matters

Childress is the clearest statement yet of what actually decides data center site selection in 2026. The developer economics no longer start with a metro, a workforce, or a land price. They start with an energization date, and everything downstream is negotiable. Lancium already owns the ground and, by its own account, holds a 1GW grid interconnect that has been fully approved by ERCOT after formal review with transmission service providers. That approval is the asset. The 270 acres is the cheap part.

The second variable is friction. A rural Texas county with no meaningful opposition infrastructure processes a project of this scale on a timeline that a contested suburban jurisdiction cannot match, which is the same lesson developers took from Northern Virginia’s rejections. The campus is grid-connected rather than islanded, with behind-the-meter solar and energy storage layered on site, and a closed-loop, non-evaporative liquid cooling system that Lancium says leaves local water supplies untouched. Water and neighbors are the two arguments that kill these projects. Both are pre-answered here.

The numbers

Construction is expected to begin in the third quarter of 2026. Crusoe says the campus will create thousands of construction jobs and more than 100 long-term positions in Childress County. The end user is described only as a leading hyperscale technology company; neither firm named it, and no total investment figure was disclosed.

This is the second Crusoe-Lancium site under the same division of labor established in Abilene: Lancium takes the land and the energy infrastructure, Crusoe designs, builds, and operates the data centers.

What’s next

Watch whether the interconnect holds its schedule, because that is the only date that matters. For developers, the transferable lesson is the one we mapped last week in speed to power, not land: the site-selection map is being redrawn around approved interconnects, and counties nobody had on a list are now competitive. The downside case is concentration risk. A county of under 7,000 people building its tax base around one unnamed tenant is a single-customer bet.