U.S. Cities Consider Limits on AI Data Centers’ Water Use

Cities across the United States are pausing or rethinking approvals for new artificial-intelligence data centers after a study published earlier this year found that server farms optimized for AI workloads can substantially increase local water demand for cooling.

Local planning departments in multiple states have placed applications on hold while elected officials and staff review zoning, environmental impacts and utility capacity. Western and southwestern jurisdictions with strained groundwater and reservoir levels are the most active in debating new controls.

Proposals under discussion include temporary moratoria on new data-center construction, limits on the volume of potable water a facility may draw, requirements to use nonpotable or recycled water, mandatory water-impact studies tied to permitting, and conditions that link approvals to investments in local water infrastructure or conservation programs.

Developers and major cloud operators point to technologies that reduce freshwater intake, including closed-loop cooling systems, recycled water use and designs that rely on nonpotable sources. Industry representatives note that some large cloud projects plan regional water sourcing in coordination with utilities and that data-center investment can bring jobs and tax revenue to host communities.

Water utilities and state regulators are reviewing how to include data centers in long-term planning. Actions under consideration include stricter accounting for large industrial users, permitting thresholds for high-volume customers, new rate structures that reflect peak-season demand, and requirements that operators submit multi-year water-use forecasts tied to expected server growth and cooling strategies before final service commitments.

Environmental groups and local residents have pushed for tighter limits and greater transparency. Community organizations are seeking access to water contracts, public review of impact assessments and clearer information on how new facilities might affect household supply, agriculture and ecosystems.

Researchers who study data-center infrastructure list mitigation options: air-cooled servers where climate allows, liquid cooling that reuses coolant in closed loops without evaporation, expanded use of recycled or reclaimed wastewater, siting near industrial users with surplus nonpotable supply, and district cooling systems that serve multiple buildings. Each option involves trade-offs in cost, energy use and technical complexity, and feasibility varies by location.

Municipal leaders are weighing short-term steps to manage pending applications while updating codes and utility agreements that did not anticipate AI-scale water demand. Some councils are considering conditional approvals with defined conservation benchmarks, while others seek state model ordinances or guidance to coordinate rules across jurisdictions that share aquifers and regional water systems.

Officials are also evaluating the link between cooling choices and electricity demand, noting that higher power use can affect grid planning and emissions projections. City hearings and regulatory reviews are expected to continue in the coming months as local governments incorporate the study’s findings into permitting and infrastructure planning.