Phoenix-area data center water consumption is projected to climb from 385 million gallons per year to 3.7 billion gallons, a nearly ninefold increase, as AI infrastructure buildout accelerates across the Sonoran Desert. Lake Mead is approximately one-third full. Lake Powell sits below one-quarter capacity. Arizona has already absorbed mandatory Colorado River allocation cuts. There is no buffer in this system.
The regional picture is worse than Phoenix alone. A study covering Arizona, Colorado, Nevada, New Mexico, and Utah projects combined data center water use reaching 7 billion gallons per year by 2035, and that figure excludes the water consumed by electricity generation that powers the facilities. Nevada data centers already account for roughly 22 percent of state electricity load. Projections show Nevada needing nearly three times the output of Hoover Dam in additional generation capacity to support its data center pipeline.
Hoover Dam's installed capacity is 2,080 megawatts. Three times that is 6,240 megawatts of new generation for Nevada data centers alone, built in a state drawing from a river that is already overallocated and falling.
Phoenix suburbs have adopted industrial water caps. Marana, Arizona has barred data centers from using potable water entirely. These are not policy proposals or advocacy positions. They are permit conditions in effect now.
Tucson's move is the most instructive. The city rejected a proposed data center until the developer pledged zero-water cooling technology. Not reduced water use. Not efficiency targets. Zero. The developer made the pledge and the project moved forward. Tucson did not ban data centers. Tucson changed the technology requirement as a condition of approval.
That is the distinction the industry needs to track. The question for any developer bringing evaporative cooling architecture to a western permit application is no longer whether a community might object. The question is whether the project can survive a Tucson-style condition being attached to the permit.
A medium-size data center can consume as much water annually as roughly 1,000 households. That figure is the comparison that lands in a city council hearing. One facility, the water footprint of a thousand homes, in a region where new residential development is already constrained by water availability.
Steven Wadsworth, chairman of the Pyramid Lake Paiute Tribe, which holds senior water rights on the Truckee River system, put the temporal problem plainly: "Everybody's only focused on what's going to happen this year, and the next two, three years. And that's just not far enough to think."
The hyperscale development pipeline runs on five-to-ten-year horizons. The water planning horizon for the Colorado River basin runs on multidecade drought cycles. Those two timelines are not aligned, and the operators committing to facilities today will be operating them during the decade when the Colorado River compact states finish working through their allocation disputes.
Dry cooling and closed-loop adiabatic systems are the architectures that get permits in Marana and Tucson. They carry a higher capital cost than evaporative towers. They run at a higher PUE during peak summer heat in the Sonoran Desert, where ambient temperatures exceed 110 degrees Fahrenheit and the thermodynamic advantage of evaporative cooling is at its maximum. Those trade-offs are real.
The trade-off math looks different when the alternative is a denied permit, a Tucson-style conditional approval that requires retrofitting the cooling architecture anyway, or a project that breaks ground and then faces community opposition severe enough to trigger litigation that delays commissioning by 18 months. The cost of zero-water cooling, priced against the cost of permit failure or construction delay, is the calculation the operators in the Arizona and Nevada pipelines need to be running now.
Zero-water cooling pilots are already running in Phoenix and Mt. Pleasant. The technology is not theoretical. The operators running those pilots are gathering the operational data that will support permit applications in the next build cycle. The ones who are not running pilots will be citing those operators' data in their own permit hearings, which is a weaker position.
The 7 billion gallon regional projection excludes water consumed by electricity generation. That exclusion matters because thermoelectric generation, the kind that powers most of the grid, is among the largest water users in the western United States. Coal and natural gas plants use water for steam generation and cooling. As Nevada adds generation capacity to feed its data center load, that generation draws additional water from the same overallocated Colorado River basin.
The data center water story in the West is not just the cooling tower draw. It is the full stack: facility cooling water plus generation water plus the allocation that agriculture and municipalities have already been told to cut. The national framing on AI water use is frequently inflated. The western basin framing is not. The numbers are localized, verifiable, and already driving permit conditions in Arizona cities that are not known for aggressive environmental regulation.
Tucson requiring a zero-water pledge before approving a project is a permit condition, not activist pressure. The difference matters to the cooling industry because permit conditions travel. Marana's potable water ban, Tucson's zero-water requirement, Phoenix suburb industrial caps, these are model conditions that other municipalities can adopt without writing new legislation. A city attorney in Albuquerque, Las Vegas, or El Paso can cite Tucson's precedent in tomorrow's planning commission meeting.
The Box Elder County opposition in Utah showed that community resistance to data center water use can stop a nine-gigawatt build in a state with a historically permissive development environment. The Tucson case shows that municipal governments do not need community opposition to act. They can attach the condition at the permit stage and let the developer solve the engineering problem.
The facilities that get built in the western U.S. over the next five years will be the ones that arrive with dry cooling or closed-loop architectures already designed in. The projects that show up with evaporative tower layouts and rely on the utility agreement to close the water budget will face the Tucson outcome or worse. The window for negotiating evaporative cooling approvals in the Arizona market is closing from both directions: from the water supply side as reservoir levels fall, and from the regulatory side as municipalities standardize on the conditions Tucson and Marana have already written.
The cooling industry has the technology to solve this. The question is whether operators spec it before the permit hearing or after the project gets conditioned.
Source: Newser, reporting on Quartz original, "The West Is Parched and AI Adds to the Strain," May 2026