How much water do Texas data centers and chip fabs use?

What Austin Water measured

Austin’s five-year schedule reports billed water in thousands of gallons. For 2025, Samsung, NXP and Cypress, formerly Spansion, add up to about 3.132 billion gallons on their three named accounts. The number is specific to the Austin Water accounts listed. A site using its own well or a different water supplier would require a separate record. Austin Water five-year comparative data.

A separate Austin city memo found that its identified potential data-center customers represented 0.15% of the utility’s total use in 2023 and 0.28% in 2024. The city did not identify those customers individually. That percentage answers a useful local question about the utility’s screen. It cannot identify the use of a particular campus or establish a statewide share. Austin staff response on data-center water use, page 6.

Google reported 333.6 million gallons withdrawn and 271.0 million gallons consumed in calendar 2025 across its facilities in Midlothian, Red Oak, Wilmer and Pflugerville. These totals add the four Texas locations in Google’s disclosure. The company’s methodology allows estimates where metered or invoiced figures are unavailable. Google’s 2026 Environmental Report, location table, p. 97. Water methodology, p. 87.

Why the cooling tower matters

A chip fab needs water for manufacturing, including rinsing wafers. A data center’s on-site water demand depends heavily on how it sheds heat. A liquid loop can circulate water through servers and still use an evaporative tower at the other end. Dry cooling avoids routine evaporation at that point, though the facility may still buy water for other uses. That is why knowing a site is “liquid cooled” or “closed loop” tells you less than knowing the equipment that rejects heat outdoors.

Fort Worth’s November 2026 infrastructure commission modeled cooling-water consumption for a hypothetical 1,000 MW plant. Its examples ranged from near zero with dry cooling to roughly 625 million gallons a year at a fleet-average intensity, and about 5.65 billion gallons with open evaporative towers. The same stated power rating produced very different water outcomes. The city labeled them examples, and they do not describe the measured use of any specific site. Fort Worth Data Center & Infrastructure Commission report, section 4.2.

Hot days test delivery capacity

A utility may have ample annual supply and still need a larger main, a new treatment plant or peak-day storage to serve an industrial campus. An annual purchase figure divided by 365 gives an average day. It cannot tell a city how much the campus will take on its hottest day or whether that peak coincides with household irrigation.

A capacity reservation in a contract is another distinct number. A four-million-gallon-per-day maximum describes what a supplier agrees to make available under the stated terms. It becomes a measured withdrawal only when supported by meter records. Drought restrictions, alternate supplies and the funding for dedicated lines matter as much as the reservation when judging the local cost. Fort Worth commission findings on infrastructure and drought. Tarrant Regional Water District’s supply overview.

Method and limits

Austin purchases come from the city’s published customer accounts. Fort Worth’s examples are engineering scenarios. Subtracting measured discharge from measured intake over the same period would give a better estimate of a fab’s net use, with adjustments for other sources or transfers where applicable. Without comparable facility-level records, there is no defensible statewide measured total here.