The Data Center Water Panic Has a Better Answer: Water Markets

Data centers are booming. As companies race to build the infrastructure behind artificial intelligence, communities across the country are sounding alarms about how much water these facilities consume. Some worry the massive complexes will strain water supplies that are already stretched thin.

The debate has quickly turned toward regulatory fixes, such as disclosure mandates, environmental reviews, and limits on how facilities operate. But the problem isn’t how much water data centers use—it’s how water gets allocated among competing demands. And as with any resource-allocation problem, the best answer isn’t mandates. It’s well-functioning markets.

Data centers use water in two main ways. First, they use it directly to cool servers through evaporative cooling systems or closed-loop recirculating systems. Second, they use water indirectly through the electricity they consume, since the power plants that generate that electricity often require water for cooling.

By some estimates, data centers account for roughly 0.2 percent of U.S. freshwater consumption, most of it indirectly through electricity generation. Compared with other uses, the numbers are relatively modest. Data centers in Arizona—home to major facilities from Google, Microsoft, and Meta—used roughly 905 million gallons of water in 2025. By comparison, Phoenix-area golf courses consume about 29 billion gallons annually. Agriculture uses far more still: irrigating a single acre of crops can require hundreds of thousands of gallons of water each year.

But simply comparing which sector uses more water misses the point. The key question isn’t whether one user consumes more water than another. It’s whether competing demands for a scarce resource can be resolved voluntarily, at mutually agreeable prices, so that water flows to its highest-valued use. Otherwise, disputes over water devolve into political and legal battles over who gets to use how much.

Markets offer a better path. In a market setting with secure water rights, water isn’t simply pumped from a river or aquifer because a user happens to need it. If water rights are clearly defined and transferable, users must acquire water through voluntary exchange. A data center that needs water might purchase rights from a farmer, a city, or another user. If the buyer values the water more than the seller, the trade benefits both parties.

When markets work in this way, many of the questions driving today’s AI water debate become far less contentious. Are data centers using “too much” water? Should regulators impose limits or mandate particular cooling technologies? Do we need sweeping environmental reviews to determine whether these facilities deserve access to scarce supplies? Markets resolve those questions through voluntary exchange rather than regulatory decree.

For that system to function, however, water markets must actually work—and in many places they face significant obstacles. 

In the Western United States, water is governed by the doctrine of prior appropriation, often summarized as “first in time, first in right.” Users who established water rights earlier hold priority over junior users in times of shortage. These rights are generally separable from land and transferable to other users. In principle, that allows water to move from lower-valued uses to higher-valued ones.

Trading does occur through Western water markets. Cities frequently purchase rights from agricultural users as they grow, and other projects—from housing developments to golf courses—must secure water rights before proceeding.

But in practice, these markets are plagued by high transaction costs. Transferring a water right typically requires regulatory approval, hydrological studies, and lengthy review processes. Third-party objections can delay or derail transactions entirely. The result is a system where mutually beneficial trades are often too costly or uncertain to pursue. 

In the Eastern United States, water rights are less clearly defined and less easily transferable, so formal water markets are less common. The rapid buildout of data centers is exposing the importance of water rights and well-functioning water markets, in both the East and the West.

Today, most data centers obtain water not directly through voluntary water markets but through municipal utilities. Those utilities typically price water based on infrastructure and operating costs rather than scarcity, meaning large data centers may not face the true marginal cost of the water they consume. Municipal utilities may then need to acquire additional water rights to meet this growing demand.

When companies obtain water through subsidized municipal rates rather than open markets, they have little incentive to conserve, and existing users are not compensated for increased demand. 

A better approach would rely on open water markets. If data centers had to acquire water rights through voluntary transactions—or pay prices that reflect the true scarcity value of municipal supplies—they would face the same incentives as any other user. Existing rights holders would be compensated, new demand would face accurate prices, and water would flow to the uses that value it most.

The potential gains from such trades are enormous. Agriculture accounts for roughly 85 percent of water use in the American West. If agricultural consumption fell by just nine percentage points, the water saved would roughly equal the entire amount currently used by all residential, commercial, and industrial users combined, according to one estimate.

Realizing those gains will require reform. Western states should streamline approval processes for water transfers, clarify existing rights, and expand water banks that make trading easier. Eastern states should use the data-center boom as an opportunity to develop clearer and more transferable water-rights systems. Municipal utilities can help as well by pricing water closer to its scarcity value and requiring large projects to cover the cost of the new infrastructure or additional water sources their demand requires.

The real question isn’t whether data centers use “too much” water. It’s whether water is allocated through voluntary exchange or through politics. Where markets function, users pay for what they consume and those who give up water are compensated. Where they don’t, scarcity turns into conflict.

As demand grows in the age of AI, the best way to manage water isn’t more regulation. It’s better water markets.