Texas

~350 facilities21% of facilities tracked here

The headline number

~10,000 gal

used in 2024 at Google's Pflugerville site, the least of any Google data center: proof that siting and cooling design, not AI itself, set the water bill

Water stress

High and uneven: West Texas sites sit in chronic drought zones while Gulf-adjacent sites don't; ERCOT's largely gas-fired grid adds indirect water either way.

The local story

Texas is where the AI build-out is fastest and the water answers diverge hardest. The same company demonstrates both ends: Google's Pflugerville site used roughly ten thousand gallons in 2024 (about what one Texas home uses in two months) while evaporatively cooled campuses elsewhere drink millions of gallons a day.

The Stargate-scale projects around Abilene anchor the biggest single bets: gigawatt-class campuses in semi-arid counties, mostly planning dry or hybrid cooling precisely because local groundwater couldn't carry evaporative loads. The trade is physical: dry cooling spends more electricity to save water, shifting the footprint from the Low scenario into the grid's Mid-scenario water.

Texas regulators now field data center water requests alongside municipal and agricultural ones, and several panhandle districts have started publishing data-center-specific groundwater accounting, the first state where that's routine.

Cite this pagev1.3.1 · 2026-09

Your number

What does your AI actually use?

Pick your model, set your usage, get the number, with sources.

Try the calculator

Sources

Cited on this page, in order of appearance.