AI Data Centers and Texas Water: Responsible Growth Needs Local Facts

Texas should welcome serious AI infrastructure. Cheap land, energy access, fiber, tax incentives, and a friendly business climate make the state a natural place for data centers to grow. The water question is not a reason to stop that buildout. It is a reason to plan it with better local facts.

The strongest version of the issue is not that every data center will drain an aquifer. Some will use little water on site. Some will use recycled water. Some will use closed-loop systems. Some will shift the water burden into the power system instead. That is exactly why the story matters. Responsible growth depends on the local design, source, and planning question, not one statewide scare number.

For Texas property owners, the practical question is even narrower. If a major new water user enters the local conversation, what does that mean for nearby wells, groundwater district rules, drilling expectations, and rural land diligence? That is where this becomes a TurnKey Wells story.

The signal: Texas wants the boom, and the water math needs visibility

The University of Texas at Austin reported in May 2026 that data centers currently account for less than 1% of Texas water use, but could potentially account for 3% to 9% by 2040. That estimate includes both direct cooling water and indirect water used to produce the power that data centers need.

That second part matters. A data center can reduce water used on site by changing cooling technology, but if the facility drives more electricity demand from thermal power generation, the water footprint does not disappear. It moves. It may show up in a power plant, a water supply plan, a municipal decision, or a regional groundwater conversation.

The same UT summary noted that manufacturing accounts for about 7% of the state’s water use under the current state water plan. So the high end of the data-center projection is not background noise. It would put the sector in the range of an existing major water-use category.

HARC’s January 2026 white paper put a nearer-term frame around the issue. It estimated existing Texas data centers consume roughly 25 billion gallons of water in 2025, or about 77,000 acre-feet. If growth projections hold, HARC estimated data-center demand could reach 29 to 161 billion gallons by 2030.

The spread is wide because the public data is still thin. That is the point. If Texas is going to lead the AI buildout, planners, utilities, developers, and landowners need clearer visibility into cooling design, power source, local water source, and the scale of projects that actually get built.

The planning challenge: growth is moving faster than water data

Texas water planning was not designed for a fast-moving industry that can arrive in large blocks. HARC warned that the 2027 State Water Plan would be based on regional plans that do not include forecasted data-center water use. In plain English: the formal planning process can lag the buildout.

E&E News reported in May 2026 that the draft 2027 plan did not allude to data centers, even as Texas had hundreds of data-center locations operating or in development. The same report noted that historic water figures used by regional planning groups and the Texas Water Development Board do not capture growing data-center water use.

The Public Utility Commission of Texas has started collecting information. Its Energy and Water Use Survey for Data Center and Virtual Currency Mining Facilities is meant to help regulators understand direct water use, electricity demand, and sources. Texas Living Waters and HARC have argued that transparent reporting and planning should become stronger and more enforceable, not just a one-time information request.

That gap between development speed and water-planning visibility is why local property owners should pay attention. The first good questions are not ideological. They are factual questions that help good projects explain themselves clearly and help nearby landowners understand the local water context.

San Antonio shows what responsible water planning can look like

San Antonio is a useful case because it shows how data-center growth can fit better when a city has a diversified water portfolio and mature recycled-water infrastructure. SAWS reported 269,848 acre-feet of potable water distributed in 2025, plus 60,503 acre-feet of recycled water delivered through its recycled-water system. SAWS also reported 1,623 acre-feet of Edwards Aquifer water stored in the H2Oaks Aquifer Storage and Recovery facility in 2025.

The San Antonio Report wrote in March 2026 that local data centers used around 0.1% of SAWS drinking water, and that recycled-water use by data centers had climbed from 48% of their water usage in 2023 to 75% in 2025. It also reported that new data centers were planning to request about 600 acre-feet of drinking water and 3,000 acre-feet of recycled water.

That is the more nuanced version of the issue. A data center using recycled water in a city with a mature purple-pipe system is not the same as a rural project asking for groundwater in a dry county. A project using closed-loop or dry cooling is not the same as a project relying heavily on evaporative cooling. A facility connected to a diversified municipal portfolio is not the same as one depending on a small system with little spare capacity.

San Antonio should be treated as a model, not a loophole. The San Antonio Report noted that SAWS’ recycled-water system has limits and would need upgrades if new demand keeps growing. Even a strong water portfolio needs capacity planning when enough large users arrive at the same time.

The cooling counterpoint is real and important

There is a legitimate counterargument to the simplest water-panic version of this story: many new data centers are not built like older evaporative-cooling facilities. Some use air cooling. Some use closed-loop systems. Some use direct-to-chip liquid cooling that can sharply reduce on-site water use.

Nvidia’s June 2026 announcement around its Rubin-generation reference design is the cleanest example of the industry argument. The company says high-temperature liquid loops can reject heat through dry coolers and reduce water use from conventional cooling-tower levels to near zero for that design.

That matters, and it is one reason the conversation should not collapse into anti-data-center talking points. Technology choices vary by project. Power demand remains. Construction and local infrastructure still matter. Public water planning still needs to know what is actually being built, where it is being built, which water source it uses, and whether the local community will have enough visibility before commitments are made.

What this means for private well owners

Private wells are property-specific. A statewide data-center estimate does not tell you whether one tract can support a domestic well. A city press release does not tell you what a driller will encounter at 420 feet. A county average does not tell you whether nearby wells are shallow, deep, domestic, public-supply, irrigation, monitor wells, or plugged.

When a large water user appears near a property or inside the same water-planning conversation, the property owner should ask five practical questions:

• What water source is the project expected to use: potable municipal water, recycled water, groundwater, surface water, brackish groundwater, or a combination?
• What cooling design is proposed: evaporative, air cooled, closed loop, direct liquid cooling, or something else?
• Is the property inside a Groundwater Conservation District, and what permitting, spacing, or reporting rules apply?
• What do nearby water well records show by distance, depth, date, use type, aquifer, driller, pump setting, and yield where reported?
• Are there older, plugged, or undocumented wells that could change the property history?

That is the reason TurnKey Wells treats water news as property intelligence. The headline might be about AI. The useful question for a land buyer is still local: what evidence should change the due diligence?

What TurnKey Wells would check first

For a property-level read, we would start with the same evidence stack we use for Pre-Drill intelligence. That includes TWDB Submitted Driller Reports, GWDB records, plugging records, Groundwater Conservation District boundaries, nearby driller lithology where available, aquifer signals, and local context from public water planning sources.

The work is not to claim certainty. No public record can guarantee exact depth, yield, water quality, or formation under one parcel. The work is to narrow the uncertainty before a buyer, seller, or landowner makes an expensive decision.

In a county where nearby domestic wells cluster around a narrow depth range, the planning conversation is different from a county where records swing from shallow stock wells to deep public-supply wells. In a GCD with strict permitting, the conversation is different from a non-district area. In a city with recycled water and aquifer storage, the conversation is different from a small system facing one new large user.

Bottom line

AI data centers are not just server buildings. In Texas, they are becoming water-planning events, and Texas should want the responsible ones. Some projects will be water-lean, transparent, and well matched to local supply. Some may be poorly understood until late in the process. The difference matters.

For property owners, the right reaction is not panic and it is not reflexive opposition. It is diligence. The goal is not to slow the AI buildout; it is to make sure developers, utilities, local officials, buyers, and landowners are working from real water facts. When a new large water user enters the local conversation, check the nearby wells, the aquifer context, the GCD rules, and the water-source details before relying on broad averages or public talking points.