Texas has built a process to decide which data centers get connected to the power grid. It has nothing equivalent for water. Two research groups have now put numbers on that gap, and their estimates suggest that the water footprint of the state's data center boom could grow several times over by 2030.
The latest is a white paper from the COMPASS research consortium at the University of Texas Bureau of Economic Geology, first published in December 2025 and updated in August 2026. It estimates that Texas data centers used about 33 billion gallons of water in 2025, counting both on-site cooling and the water consumed by power plants that supply them. That is about 0.6% of statewide water use.
Direct and indirect
The UT estimate splits water use into two parts. Direct use, mainly evaporative cooling at the data center itself, was about 13 billion gallons in 2025, under an assumed 8 GW of data center power capacity. The rest, about 20 billion gallons, is indirect: water consumed at the power plants generating the electricity.
The figures are consistent with an independent estimate from the Houston Advanced Research Center. In a white paper released on January 21, HARC put existing Texas data center water use at about 25 billion gallons a year for electricity generation and cooling, or 0.4% of state water use.
The 2030 range
The forecasts diverge widely, because they depend on how much of the proposed capacity gets built. HARC estimates that data center water use could rise to between 29 and 161 billion gallons a year by 2030, as much as 2.7% of the state's total water use. The Texas Tribune noted that the upper end equals the consumption of 1.3 million average U.S. households.
The UT scenarios run higher. Averaged across its modeled cases, direct use rises to 90 billion gallons in 2030 (1.5% of projected statewide demand) and 139 billion gallons in 2040 (2.3%). Including power plant water, total use reaches 225 billion gallons in 2030 (3.7%) and 348 billion gallons in 2040 (5.8%).
The paper tests that range against grid forecasts. Under an ERCOT-adjusted estimate of 22 GW of data center capacity in 2030, direct use is about 36 billion gallons and total use about 91 billion gallons, or 1.5% of statewide demand. Under the much higher 78 GW figure built from transmission service provider requests, direct use reaches about 127 billion gallons a year.
The size of the project pipeline explains the spread. The UT paper cites a directory that listed 479 data centers in Texas in late 2025 and 766 by August 2026, a count that includes planned and proposed projects that may never be built. A year earlier, The Texas Tribune counted more than 400 data center facilities in the state, with about 70 more on the way.
Why cooling choices matter
How a data center is cooled decides most of its direct water use. The UT paper notes that evaporative cooling is highly energy efficient but consumes significant amounts of water, while dry cooling and heat pump systems can nearly eliminate direct evaporative losses at the cost of higher power use. Its comparison table puts water use for direct evaporative cooling at about 0.03 to 0.26 gallons per kWh and for indirect evaporative systems at 0.46 to 0.66 gallons per kWh.
That creates a tradeoff. A data center that saves water by running dry coolers uses more electricity, which can mean more water consumed at the thermal power plants that supply it. The UT estimate counts both sides, which is why its totals are higher than direct-use figures alone.
Rack density is rising fast too. The paper says a typical rack drew about 4 to 5 kW in 2010, while GPU racks in 2025 can exceed 80 to 100 kW, and heat loads are expected to approach 1 MW per rack by 2028. Removing that heat with air alone becomes impractical, which pushes operators toward liquid cooling systems whose water use depends on design.
The planning blind spot
The core problem identified by both groups is that Texas does not plan for this demand. The Texas State Water Plan, the main tool for funding water infrastructure, relies largely on historical data, so it does not account for forward-looking data center growth, HARC said.
"Texas' water planning process has a structural blind spot," said Margaret Cook, HARC's vice president of water and community resilience and the paper's primary author. "ERCOT has begun developing processes to manage large load interconnections for energy, but no parallel process exists for water planning. This leaves utilities and city managers to negotiate individually with multinational technology firms, often without the data or leverage necessary to protect local resources."
Statewide percentages can understate local strain. "[It] may not feel like it's a whole lot at the state level," Cook told The Texas Tribune last year, but if large data centers locate in small communities, those communities may not be able to handle the jump in demand. In Amarillo, residents organized against five planned data centers over risks to the Ogallala Aquifer, which is being drained faster than it is replenished, the Tribune reported.
What researchers recommend
HARC recommends requiring large industrial users to report expected water and electricity use, building forward-looking forecasts into the State Water Plan and offering incentives for water-lean technology such as dry cooling and brackish water reuse.
"The resource footprint of data centers is not a future challenge; it is already here," Cook said. "By requiring transparency, incentivizing efficiency, and prioritizing alternative water supplies, Texas can accommodate the digital economy without compromising the reliability and affordability of resources for its residents."
What to watch
The next State Water Plan cycle is the obvious venue for change. If regional planning groups begin to use the same large-load data that ERCOT now collects for power, Texas would have its first forward-looking estimate of data center water demand. The other signal is local. City councils negotiating water contracts with hyperscale campuses, especially in West Texas, will show whether the industry adopts dry and hybrid cooling at scale or keeps relying on evaporation in a drought-prone state.