Are Data Centers Harming Water-Stressed Communities?

California is running out of water. That's not a forecast — it's been the operational reality for much of the state for over a decade. Reservoirs drop. Farmers fallow fields. Cities mandate shorter showers. And quietly, in the background, a different kind of consumer has moved in: the data center, drawing millions of gallons of water to keep servers cool while facing almost none of the scrutiny applied to agriculture or municipalities.

The collision between data center expansion and water-stressed California communities isn't a future problem. It's happening now — and the public largely doesn't know it.

The Scale of Data Center Growth in California

California hosts one of the largest concentrations of data center infrastructure in the world. Silicon Valley alone has historically anchored that density, but growth is spreading to secondary markets — the Inland Empire, Sacramento suburbs, and Central Valley corridors — precisely because land is cheaper and power connectivity exists. Nationwide, data center capacity is projected to more than double by 2030, driven by cloud computing, AI workloads, and streaming infrastructure that demands always-on, always-cool hardware.

What most people don't realize is that a single hyperscale data center can consume between 1 and 5 million gallons of water per day — comparable to the daily water needs of a small city.

That water doesn't disappear into a product. It evaporates. Cooling towers pull in municipal or well water, use it to dissipate heat, and release it into the atmosphere as vapor. The facility consumes it completely, classified under what the industry calls "consumptive use" — a technical phrase that does a lot of quiet work obscuring a simple reality: the water is gone.

AI has accelerated this problem faster than most regulators anticipated. Training large language models is extraordinarily compute-intensive, and inference — actually running those models at scale — maintains that heat load continuously. Microsoft acknowledged in its 2023 sustainability report that its global water consumption jumped 34% in a single year, largely attributable to AI infrastructure. Google reported similar trends. Both companies have major California footprints.

What Water Scarcity Actually Means at the Community Level

Abstract statistics about water stress become viscerally real when you look at specific communities. Parts of the San Joaquin Valley already rely on water sources that are legally and hydrologically overcommitted — meaning more water has been promised to existing users than actually flows through the system in dry years. Groundwater basins in the region are being depleted faster than natural recharge can replace them, a crisis the state is attempting to address through the Sustainable Groundwater Management Act (SGMA), which requires basins to reach sustainability by 2040.

Into this environment, data centers arrive as a new class of major water consumer — often without the public scrutiny that a new agricultural operation or housing development would face.

For low-income and rural communities, the stakes are particularly acute. These are places where residents may already be on domestic wells that tap the same aquifers a nearby facility is drawing from. When water tables drop, those wells go dry. The cost of deepening or replacing a domestic well runs from $15,000 to $50,000 or more — an impossible expense for many families. The data center, meanwhile, is drawing from the same basin under permits that were issued with minimal public process.

This isn't hypothetical. The dynamic has played out repeatedly in communities across the American Southwest, and California's regulatory gaps make it structurally likely to repeat.

A Regulatory Framework With Significant Blind Spots

California has some of the most sophisticated environmental regulations in the world. The California Environmental Quality Act (CEQA) requires environmental review for major projects. Water districts have permitting authority. Air quality management districts oversee cooling tower emissions. On paper, the oversight architecture exists.

In practice, disclosure of actual water consumption by data centers is remarkably thin.

Unlike electric utilities — which report consumption data that feeds into public planning documents — data centers are not required to publicly disclose real-time or even annual water use figures in a standardized, accessible format. The information exists somewhere in permit applications and water purchase agreements, but it's fragmented across dozens of jurisdictions and agencies, inaccessible without significant legal and technical resources.

The gap between what regulators know and what the public can access creates conditions where communities are making decisions about land use and growth without the full picture.

Water agencies approving new data center service connections are often making those decisions based on projections provided by the applicant — projections that, historically across the industry, have underestimated actual consumption. There's no standard methodology for calculating data center water use, and no California agency currently requires facilities to meter and report consumptive water use with the same rigor applied to, say, an agricultural irrigation district.

The result: community members attending planning meetings don't have the numbers they need to ask meaningful questions.

The Economic Trade-off Nobody Wants to Calculate Honestly

Data center proponents make a straightforward economic case: facilities bring construction jobs, permanent employment, and a substantial property tax base without the school enrollment costs of residential development. That argument is real. A large data center can generate tens of millions of dollars in annual tax revenue for a county that might otherwise be struggling.

But the economic trade-off calculation almost never includes the cost of water.

Water in California isn't priced at anything close to its scarcity value. Agricultural water rights, some of which date to the 19th century, can cost a fraction of what municipal or industrial water costs — and industrial water is still dramatically underpriced relative to the economic damage caused by aquifer depletion. When a data center negotiates a water service agreement with a municipal utility, neither party is internalizing the long-run cost of drawing down a basin that may take centuries to recharge.

The community bearing the long-term cost of aquifer depletion is rarely the same community receiving the tax revenue from the data center. Rural residential and agricultural users downstream or in the same basin absorb the externality — reduced water availability, lower well yields, higher pumping costs — while the fiscal benefits flow primarily to county governments and utility ratepayers in the service area.

This isn't an argument against data centers. It's an argument that the economic case for hosting them needs to account for costs that are currently invisible in the ledger.

What Better Looks Like: Practices Worth Adopting

The technology exists to dramatically reduce data center water consumption. The industry knows this. The question is whether market pressure and regulation will accelerate adoption.

Air-Side and Closed-Loop Cooling

Some of the most advanced facilities have moved away from evaporative cooling entirely. Microsoft's underwater datacenter experiment (Project Natick) and its land-based facilities in cooler Nordic climates use outside air for cooling — a technique called air-side economization — that requires essentially no water. In climates like California's, air-side cooling is harder to deploy year-round, but hybrid approaches that use water only during peak temperature events can reduce consumption by 80% or more compared to conventional cooling towers.

Closed-loop cooling systems — where water circulates without evaporative loss — represent another meaningful step. The water is used as a heat transfer medium but isn't consumptively lost. These systems cost more to build and operate, but they're deployable at scale.

Mandatory Metering and Disclosure

Singapore, facing its own severe water constraints, has implemented mandatory water efficiency reporting for data centers as part of its broader facility licensing framework. The European Union's Energy Efficiency Directive, updated in 2023, requires large data centers to report water usage metrics — establishing a baseline for accountability that doesn't yet exist in California.

California could require any data center above a certain capacity threshold — say, 1 megawatt of IT load — to file annual water consumption reports with the State Water Resources Control Board in a publicly accessible format. That single change would allow researchers, community organizations, and planning agencies to understand cumulative basin impacts before they become irreversible.

Siting Policy as a First Line of Defense

The most powerful intervention is also the most upstream: don't site water-intensive facilities in water-stressed basins without rigorous analysis of cumulative impact. Several California counties have begun developing specific data center siting criteria within their general plans, but the practice is inconsistent and often reactive — triggered by a specific proposal rather than proactive regional planning.

Water is the infrastructure that all other infrastructure depends on. A region that loses reliable water supply doesn't attract data centers or anything else. The tech industry's expansion into California communities is generating real economic value — but sustainable growth requires accounting for what's being consumed in the process.

The data centers that will anchor this industry for the next 30 years are being sited and permitted right now. Getting the disclosure rules, the siting criteria, and the water pricing right isn't an obstacle to that growth. It's the precondition for it lasting.

**Explore how InfraSale Marketplace can help you navigate these challenges.**

[INTERNAL LINK: data center growth]

[INTERNAL LINK: water scarcity impact]

[INTERNAL LINK: regulatory challenges]