Austin has 13 data center sites under construction right now. Thirteen. The city's entire electric utility, Austin Energy, peaks at just over 3 gigawatts serving roughly 570,000 customers. Some of the individual facilities requesting grid connections want 5 to 11 GW. Read that again. A single campus asking for more power than every house, hospital, and office building in Austin combined.
This is a cooling story masquerading as an energy story.
Because every megawatt of compute needs to be cooled. And the way Texas cools its data centers, overwhelmingly through evaporative systems that chew through freshwater like it grows on trees, is about to collide with a geography that spent most of the last decade in some stage of drought.
Typical data center power requests in the Austin market used to land around 30 MW. That number has jumped tenfold, to 300 MW or more. Tract, the land and infrastructure developer, nearly doubled its Caldwell Valley Technology Park to nearly 3,000 acres capable of supporting up to 4 GW of data center capacity. The first 360 MW is targeted for energization in 2028 through a facility design agreement with Bluebonnet Electric Cooperative. That campus sits 30 minutes from the Austin airport, between Austin and San Antonio, in a corridor that barely had municipal water infrastructure a decade ago.
ERCOT, the grid operator for most of Texas, projects systemwide load could nearly double by 2028. More than 220 GW of large projects have applied for grid interconnection by 2030, and over 70% of those are data centers. Forty gigawatts of upcoming gas power capacity in Texas is earmarked specifically for compute.
Chevron is building a 2.5 GW natural gas plant in West Texas, expandable to 5 GW, to feed data centers directly. ExxonMobil, under CEO Darren Woods, is developing gas plants with carbon capture to power AI campuses. Diamondback Energy is doing the same. Oil and gas companies are becoming data center utilities. That transition creates new thermal loads on both ends: generating the power produces waste heat, and consuming the power produces more.
The Houston Advanced Research Center published a white paper in January 2026 titled "Thirsty Data and the Lone Star State." The headline number: Texas data centers consumed an estimated 25 billion gallons of water in 2025 when you account for both direct cooling use and indirect use from power generation. About 0.4% of total statewide consumption.
By 2030, that figure lands somewhere between 29 billion and 161 billion gallons annually. The range is absurd. It spans from a modest uptick to a sixfold increase. HARC's researchers attribute the spread to uncertainty in construction pace, electricity demand forecasts, and cooling technology choices. That last variable is the one cooling vendors and thermal engineers should be staring at, because it represents the delta they can actually influence.
A single 100 MW data center uses as much water as 2,600 households. The facilities coming online in central Texas are not 100 MW. They are multiples of that. When Tract fills out its Caldwell Valley campus at 4 GW, the cooling water demand from that single park could rival a small city's entire municipal supply, depending entirely on what cooling architecture the tenants deploy.
Here is the part that should make thermal infrastructure planners lose sleep. No central entity in Texas collects or compiles forward-looking water demand data from data centers. None. The Texas Water Development Board relies on self-reported historical consumption surveys. Compliance is inconsistent. There is no centralized registry of data center operations. The State Water Plan does not include projected demand growth for the industry.
ERCOT has built processes to manage large load interconnections on the power side. There is no parallel mechanism for water. City managers and local utility districts are left negotiating individually with multinational tech companies that have more lawyers than some of these towns have residents. The Texas Observer reported that the AI boom is outpacing water regulations entirely, and HARC's recommendations include requiring large industrial users to report expected water and electricity consumption in advance, integrating forward-looking forecasting into the State Water Plan, and incentivizing water-lean cooling technologies.
That last recommendation is where the cooling industry enters the frame.
The gap between 29 billion and 161 billion gallons is not a statistical curiosity. It represents a direct question to the cooling industry: are you going to build the next wave of Texas data centers on evaporative cooling, or are you going to shift the architecture?
Evaporative cooling dominates in Texas because the economics are favorable. Hot, dry climate. Cheap land. Historically cheap water. Evaporative systems reject heat using a fraction of the electricity that air-cooled alternatives require. The fan power penalty for dry cooling follows a cube law curve. Double the fan speed, draw eight times the power. In a Texas summer pushing 105 degrees, that math gets ugly fast.
But the water math is getting uglier faster. Hays County, just south of Austin, is drought-prone and rapidly growing. The aquifer systems feeding the I-35 corridor between Austin and San Antonio are already contested between residential development, agriculture, and industrial users. Adding gigawatts of water-cooled compute to that equation without a statewide planning framework is a gamble that assumes the water will always be there. Texas ranchers and municipal water authorities would disagree.
Direct liquid cooling, both cold plate and immersion, changes the equation. These architectures move heat rejection from the building level to the chip level, reducing or eliminating the need for evaporative towers. Closed-loop systems can cut freshwater consumption by up to 70%. Some operators are exploring brackish water reuse and geothermal rejection loops that return water to its source without consumption. The technology exists. The question is deployment speed.
Senate Bill 6, passed by the Texas Legislature, imposes new standards on customers requesting 75 MW or more of grid connection. The bill targets electricity reliability, not water. But the regulatory attention is moving toward resource constraints broadly, and water scarcity has a way of creating political urgency that kilowatt-hours alone do not.
Thirteen sites under construction in Austin alone. Nearly 3,000 acres at Caldwell Valley. Forty gigawatts of dedicated gas generation being built across the state. The Texas data center pipeline is the largest concentration of new cooling infrastructure demand anywhere in North America right now.
Every thermal system vendor, every cooling tower manufacturer, every direct liquid cooling startup should be asking the same question: what happens when Texas realizes it cannot supply 161 billion gallons of cooling water annually and still keep the taps running in San Marcos?
The vendors who show up with water-lean architectures that don't tank PUE will own this market. The ones still pitching evaporative towers as the default will find themselves on the wrong side of a regulatory shift that is coming whether Austin's boosters want it or not. The 29-to-161-billion-gallon spread is the cooling industry's addressable market, measured in water it can save Texas from wasting.