Will Data Centers Drain Our Water Supply?

A single proposed data center in Farmington, Colorado, could consume more than 800 million gallons of water per year. Let that number sink in for a moment. That's enough water to supply roughly 7,500 American households annually — drawn from one facility, in one town, in a state that already treats water like the finite resource it is.

The developer behind this project is Tract, a Colorado-based infrastructure company betting big on the data center boom. They're not alone. Across the American West — in Nevada, Arizona, New Mexico, and Colorado — data centers are staking claims in communities where aquifers don't recharge fast enough and rivers are already over-allocated. The timing couldn't be more fraught. The questions being asked in Farmington today will be asked in dozens of other towns before this decade is out.

Understanding Data Center Water Consumption

Data centers don't just burn electricity — they sweat. The dominant cooling method for large-scale facilities is evaporative cooling, which works by passing hot air over water to dissipate heat into the atmosphere. It's effective, relatively cheap, and extraordinarily thirsty.

To put 800 million gallons in industrial context: that's comparable to the annual water consumption of a mid-sized semiconductor fabrication plant or roughly the irrigation demand of 2,500 acres of corn in a dry year. The difference is that chip fabs produce physical goods, and farms feed people. Data centers produce compute — something essential, yes, but in a fundamentally different category when communities are asked to trade water rights for server racks.

The water intensity of a data center scales directly with its power density and cooling architecture — and as AI workloads drive up both heat output and rack density, the water problem is getting worse, not better.

Industry benchmarks measure this with a metric called Water Usage Effectiveness (WUE) — liters of water consumed per kilowatt-hour of IT equipment energy. Hyperscale operators like Google and Microsoft have published WUE figures ranging from 0.5 to over 1.7 L/kWh, depending on the facility and climate. At scale, those fractions add up to billions of gallons per year across a single company's global footprint.

The Environmental Implications of High Water Usage

The Farmington case isn't happening in a vacuum. Colorado's water law operates under the prior appropriation doctrine — first in time, first in right. When Tract or any developer secures water rights for an 800-million-gallon annual draw, that water doesn't come from nowhere. It comes from the same basin that farmers, municipalities, and ecosystems have been negotiating over for generations.

The San Juan River basin, which serves much of the Farmington area, is already subject to interstate compact obligations and tribal water rights settlements. Adding a massive industrial consumer into that system isn't just a local issue — it reverberates across a hydrological network that crosses state lines and sovereign boundaries.

When a data center draws down local groundwater or surface rights in an arid region, the downstream consequences aren't abstract: reduced streamflows mean less habitat for native fish, lower water tables threaten residential wells, and agricultural users face a powerful new competitor with deeper pockets.

What makes this particularly thorny from an environmental standpoint is that data center water consumption is largely consumptive — meaning it doesn't return to the watershed. Evaporated water is gone. Unlike industrial users who discharge treated water back into a river or municipal system, cooling towers release that water into the atmosphere, where it may fall as precipitation somewhere else entirely. For drought-stressed Western basins, that's a permanent loss.

Investment Considerations: Risks and Opportunities

For infrastructure investors, data center water usage has moved from footnote to front-page risk. ESG frameworks increasingly require disclosure of water consumption at the asset level, and institutional investors are starting to apply real scrutiny to facilities sited in water-stressed regions.

The risk isn't purely reputational. It's operational. A data center that secures water rights today isn't guaranteed those rights will be politically or legally defensible in ten years — especially as climate change tightens supply and communities grow more resistant to large industrial draws. Regulatory rollback is a genuine exposure, not a tail risk.

On the flip side, developers who get water strategy right early have a durable competitive advantage. Sites with access to reclaimed water, deep wells with strong recharge rates, or proximity to bodies of water that support non-consumptive cooling create infrastructure assets that will hold value as water scarcity tightens across the Sun Belt. The data centers that will command premium valuations in 2035 are the ones being built today with water resilience baked into the site selection, not retrofitted after the fact.

There's also a first-mover opportunity in markets that other developers are avoiding precisely because of water constraints. Pairing a data center development with water recycling infrastructure or co-locating with municipal water treatment facilities creates defensible moats and community goodwill that purely extractive projects never achieve.

Innovative Solutions for Sustainable Water Use

The industry isn't standing still, and it's worth separating genuine innovation from greenwashing. A few approaches are showing real traction.

Closed-loop cooling systems recirculate water rather than evaporating it, dramatically reducing consumptive loss. The tradeoff is higher capital cost and, in some configurations, lower cooling efficiency — but the technology is mature and deployable at hyperscale.

Air-side economization leverages ambient outdoor air for cooling during temperate months, reducing or eliminating water use for portions of the year. In northern climates, facilities can run water-free for the majority of the year. In the American Southwest, the math is harder, which raises an obvious question: why are so many data centers being built there in the first place?

Proximity to geography matters enormously here. The Pacific Northwest — with its cool climate, hydroelectric power, and less acute water stress — is a fundamentally better thermal environment for data centers than Arizona or New Mexico. The reason developers keep choosing desert markets comes down to land cost, permitting speed, and tax incentives. Those are solvable policy problems, not immutable facts.

Some operators are experimenting with direct liquid cooling (DLC), which routes chilled water or dielectric fluid directly to chip-level heat exchangers. DLC can reduce facility-level water consumption by 90% or more compared to traditional evaporative systems. As GPU clusters for AI inference become the dominant workload — and as those GPUs get hotter — DLC is transitioning from boutique solution to mainstream requirement.

The best-in-class facilities are also pursuing water recycling partnerships with municipalities, treating wastewater to data center standards and offsetting freshwater draws. Microsoft has made public commitments to be "water positive" by 2030 — replenishing more water than it consumes globally. Whether those commitments survive the pressure of AI infrastructure buildout is a genuine open question.

Future Trends in Data Center Development

Regulatory pressure on data center water consumption is building, and it will arrive unevenly — which creates both risk and opportunity depending on where assets are sited.

California has already implemented reporting requirements for large water users that capture hyperscale data centers. Arizona's legislature has passed restrictions affecting data center development in the Phoenix area tied directly to groundwater depletion concerns. These aren't isolated policy experiments — they're early signals of a regulatory wave that will eventually reach every water-stressed market in the country.

The developers and investors who treat water permitting as a cost of doing business rather than a strategic constraint are accumulating risk they don't fully see yet.

On the demand side, enterprise customers — particularly those with their own ESG commitments — are beginning to ask harder questions about where their cloud compute actually lives. A Fortune 500 company reporting Scope 3 emissions and water consumption increasingly needs to account for the infrastructure footprint of its cloud providers. That's a procurement signal, and data center operators are paying attention.

The Farmington proposal is a useful lens precisely because it concentrates all of these tensions into a single, concrete case. A developer, a community, a water system, and a technology industry all colliding in a town most Americans couldn't find on a map. The outcome there won't resolve the broader question, but it will set a precedent — for regulatory posture, community expectations, and investor calculus — that will echo across the dozens of similar projects in development across the West.

Water has always been the limiting factor in arid-land development. Data centers didn't create that constraint, but they've arrived into it with an industrial-scale thirst and a business model premised on cheap, abundant resources. The developers who survive the next decade will be the ones who figured out that water isn't just a utility cost — it's a strategic asset. The ones who didn't will be left holding permits in watersheds that can't support them.

Explore more about sustainable data center solutions in our marketplace.

[INTERNAL LINK: data center sustainability]

[INTERNAL LINK: water usage effectiveness]

[INTERNAL LINK: regulatory trends in data centers]