On September 14, 2026, Texas Governor Greg Abbott issued a directive that would have been unthinkable two years earlier. The Texas Water Development Board was ordered to "use its enforcement authority to penalize past failures and to impose legal consequences for any future failure to comply" — specifically targeting data centers that had failed to report their water usage.
The trigger was a compliance rate that bordered on defiance. In 2025, the state sent mandatory water-use surveys to more than 300 data centers. Fewer than 17 percent responded. In the three years the survey has been distributed, only 28 percent of facilities have ever completed it.
Texas is now the first state to penalize data centers for failing to disclose how much water they consume. The directive marks a shift from encouragement to enforcement — and it raises a question that the industry has spent years avoiding: how much water does AI actually use?
The Numbers Texas Wanted
The state had good reason to want the data. According to the Houston Advanced Research Center, existing data centers in Texas consume an estimated 25 billion gallons of water annually for cooling and electricity generation. By 2030, that figure could rise to between 29 and 161 billion gallons per year — potentially representing up to 2.7 percent of the state's total water use.
A single medium-sized data center can consume roughly 110 million gallons annually — comparable to the yearly water needs of a town of 50,000 residents. Large hyperscale facilities can draw up to 5 million gallons per day during peak conditions.
Those figures are not hypothetical. In Hays County, water advocates have identified five potential data center developments and are preparing to protest them. In Henderson County, the West Cedar Creek Municipal Utility District suspended new commitments to high-intensity water customers above 250,000 gallons per day in May 2026 — a threshold that one proposed data center project exceeded by roughly twenty times.
In Corpus Christi, leaders believe data center plans may be behind delays to an emergency water supply. The city, which serves half a million people, is fighting in court for groundwater beneath the small town of Sinton — and officials suspect Sinton is hoarding that water for a new data center complex.
"I have a neighbor whose well is down," said Roger Browning, a lifelong cattle ranger from Quanah, testifying before the Texas House Natural Resources Committee. "I have another neighbor whose well volume went from one inch stream, running 24 hours, and now it pumps down after 30 minutes. The only variable is they're pulling water out to build the data center."
Why Data Centers Use So Much Water
The water consumption is a direct consequence of heat. Training and running large AI models requires dense clusters of GPUs that run far hotter than traditional servers. Most large facilities still rely on evaporative cooling because it is efficient and relatively cheap on power. Water is sprayed or circulated over cooling towers, and as it evaporates, it absorbs heat and cools the equipment.
The problem is that the water does not return. Typically 70 to 85 percent of withdrawn water is fully consumed through evaporation. It does not go back into the aquifer or the reservoir — it leaves as humidity.
The indirect footprint is even larger. Data centers consume water not just on-site, but also at the power plants that supply their electricity. The Congressional Research Service found that U.S. data centers directly consumed roughly 17.4 billion gallons of water in 2023 — more than triple the 5.6 billion gallons consumed in 2014. When the water used to generate their electricity is included, the total footprint is 12.4 times larger.
What China Does Differently
China faces the same physics. It does not face the same political constraints.
China's data center water consumption market was estimated at 228.74 billion liters in 2025, with projections reaching 374.75 billion liters by 2031. Like the United States, China has concentrated much of its data center capacity in water-stressed regions — approximately 60 percent of its facilities are located in areas with high water scarcity, compared to 38 percent in the U.S.
The difference is in how China has responded.
Rather than fighting local communities over scarce water, China's "East Data, West Computing" strategy routes computation to where the resources are. Inner Mongolia, with its cold climate and abundant renewable energy, has become a major hub. The region's total computing power reached 345,000 petaflops as of July 2026, with more than 85 percent of data center electricity coming from green power.
The cooling technology follows the climate. In Zhangjiakou, where the average annual temperature is 7°C and parts of the region drop below 2.6°C, data centers universally adopt natural cooling combined with air-cooling systems, reducing reliance on traditional water cooling. The city's average PUE is approximately 1.21, with some advanced projects achieving below 1.1.
In Datong, Qineng Technology and Envicool implemented a high-density free-cooling air-cooled solution at the Guangling Super Energy Complex that operates with zero water consumption. Qinhuai Data's "Xuanbing" waterless cooling technology achieves WUE=0 with a CLF below 0.1, and in Lingqiu, it has recorded 8,600 hours per year without needing a chiller.
China has also experimented with underwater data centers. A facility off the coast of Shanghai uses seawater as a passive cooling mechanism, with a closed loop where back-end air handlers absorb hot air generated by servers.
The Regulatory Gap
The contrast between Texas and China is not about technology. It is about governance.
Texas has the technology to reduce data center water consumption. Closed-loop cooling systems use a one-time water fill — one to two million gallons — and then recirculate it for years. Amazon and Google both presented examples of non-potable water use and closed cooling systems at the June 2026 legislative hearing.
But the technology is optional. Nothing in Texas law requires data centers to use it. Nothing requires them to report how much water they use. And until September 14, 2026, nothing gave the state the authority to penalize them for not reporting.
The Texas Water Code requires data centers to submit survey responses or face a Class C misdemeanor. But the Water Development Board has said it does not have authority to prosecute anyone because it is a non-regulatory agency — a gap that Abbott's directive now orders it to address by referring violators to local authorities. Civil penalties could include automatic denial of permits from the Texas Commission on Environmental Quality.
What the Industry Says
The Data Center Coalition, which represents Amazon, Microsoft, Google, Meta, Anthropic, and OpenAI, has argued against restrictions. Dan Diorio, the coalition's vice president of state policy, told lawmakers that the industry takes water issues seriously and is exploring cooling techniques that do not use surface or groundwater. But he advised against mandating specific technologies. "Mandating the technology of today fails to create an environment that fosters the innovations of tomorrow," he said.
The industry's position is that disclosure requirements are unnecessary because data centers use less than 1 percent of Texas water. That figure is accurate today. The question is what happens when data center water demand grows from 25 billion gallons to 161 billion gallons — and the state still does not know how much each facility is using.
The Real Question
The Texas directive is less about water use itself than about the information gap that makes water planning impossible.
The state cannot plan its water future if it does not know who is using how much. The Water Development Board's projections for the next 50 years depend on data that more than 80 percent of data centers have refused to provide. The agency is building a plan on a foundation of missing numbers.
Abbott's directive is a first step. It does not mandate water-efficient cooling. It does not cap data center water consumption. It does not require closed-loop systems. It simply requires data centers to answer the survey they were already legally required to answer.
The industry has spent three years refusing to clear that low bar. The next step — the harder step — is deciding what to do with the numbers once they arrive. Texas has not yet had that conversation. Neither has the rest of the country. And the data centers keep coming.
Sources:News from the States (September 14, 2026); Texas Tribune (September 14, 2026); Dallas Morning News (September 15, 2026); Courthouse News (June 23, 2026); Inside Climate News (May 19, 2026); Houston Advanced Research Center (January 2026); Congressional Research Service via Data Vault (July 2026); Zhangjiakou Municipal Government (September 2025); Envicool (October 2025); Qinhuai Data (2023); Space Group (July 2026); Mordor Intelligence China Data Center Water Consumption Report (2026); Hong Kong Economic Times (August 2026); Thansettakij (July 2026).
Disclaimer
The information provided in this article is for general informational and educational purposes only. It does not constitute legal, financial, or professional advice. The author and publisher are not responsible for any actions taken based on the content of this article. Readers should consult qualified professionals for advice specific to their situation. All trademarks and references to third-party products, services, or organizations are the property of their respective owners. The performance data and benchmarks discussed are based on specific research studies and may not generalize to all use cases or environments. As of the publication date, the AI landscape continues to evolve rapidly, and readers should verify current information independently.
Limitations
This analysis is based on reporting and public data available as of the article date; figures may be revised as sources update.
Forecasts from third-party analysts can change with market conditions.
Cost and pricing examples are point-in-time estimates; actual rates vary.
Country and company comparisons rely on public reporting, not operational data.
This sector moves fast; timelines and deal terms may be updated later.
Company deals and regulatory rulings may evolve; verify current status.
AI infrastructure is changing quickly; claims can become outdated soon.
Sources
- News from the States (September 14, 2026)
- Texas Tribune (September 14, 2026)
- Dallas Morning News (September 15, 2026)
- Courthouse News (June 23, 2026)
- Inside Climate News (May 19, 2026)
- Houston Advanced Research Center (January 2026)
- Congressional Research Service via Data Vault (July 2026)
- Zhangjiakou Municipal Government (September 2025)
- Envicool (October 2025)
- Qinhuai Data (2023)
- Space Group (July 2026)
- Mordor Intelligence China Data Center Water Consumption Report (2026)
- Hong Kong Economic Times (August 2026)
- Thansettakij (July 2026).
The information provided in this article is for general informational and educational purposes only. It does not constitute legal, financial, or professional advice. The author and publisher are not responsible for any actions taken based on the content of this article. Readers should consult qualified professionals for advice specific to their situation. All trademarks and references to third-party products, services, or organizations are the property of their respective owners. The performance data and benchmarks discussed are based on specific research studies and may not generalize to all use cases or environments. As of the publication date, the AI landscape continues to evolve rapidly, and readers should verify current information independently.
Limitations: This analysis is based on reporting and public data available as of the article date; figures may be revised as sources update.; Forecasts from third-party analysts can change with market conditions.; Cost and pricing examples are point-in-time estimates; actual rates vary.; Country and company comparisons rely on public reporting, not operational data.; This sector moves fast; timelines and deal terms may be updated later.; Company deals and regulatory rulings may evolve; verify current status.; AI infrastructure is changing quickly; claims can become outdated soon.