Two-Thirds of Planned US AI Datacenters Sited in Drought-Affected Regions, Guardian Analysis Finds
Summary
- • Guardian analysis finds 517 of 809 planned US AI datacenters (~64%) are located in drought-affected regions
- • US datacenter water demand is projected to surge from 17 billion to 73 billion gallons per year by 2028
- • Each 100-word AI prompt consumes roughly one 500ml bottle of water due to datacenter cooling requirements
- • Experts warn an inevitable 'crunch point' looms as AI water demand collides with record-setting US drought
Details
64% of planned datacenters in drought zones
517 of 809 planned US AI datacenters are in locations under drought conditions over the past year, per Guardian analysis of Cleanview and federal government drought data.
Water demand: 17bn → 73bn gallons/year by 2028
US datacenter water consumption is set to more than quadruple in five years, driven by rapid AI infrastructure expansion across the country.
Up to 5M gallons/day per large datacenter
Large AI datacenters — some the size of small towns — can consume as much water per day as 50,000 people, primarily for server cooling arrays.
100-word AI prompt ≈ one 500ml bottle of water
Researchers estimate each modest AI query consumes roughly a standard water bottle's worth of water via the cooling systems of the datacenters processing it.
Datacenters may reach 9% of Texas water use by 2040
Texas researchers calculated this figure as two major new datacenters arrive in Pecos and Carson counties, both recently under drought, while the state already faces rising demand and falling supply.
Utah approves world's largest datacenter complex in drought county
A complex twice the size of Manhattan was approved last month in a Utah county that has been under deep drought since summer 2025, drawing controversy from water resource experts.
Record spring drought: >60% of contiguous US affected
Over 60% of the contiguous US is currently at varying stages of drought — the largest spring drought expanse in modern records — with severe desiccation in the south-east and west raising wildfire and crop failure risks.
Water rights law unprepared for AI-scale demand
Legal frameworks in the eastern US assume water abundance; experts warn there is currently no clear mechanism to prioritize residents and farms over datacenters when supplies run short.
Arid locations chosen for economic incentives
Dry, sparsely populated regions offer lower land costs, generous tax breaks, and reduced equipment corrosion — incentives that drive siting decisions without accounting for long-term water sustainability.
Key data points from Guardian analysis of 809 planned US AI datacenters cross-referenced with federal drought severity records and Cleanview infrastructure data, supplemented by expert commentary on water resource law.
What This Means
The AI infrastructure boom is on a collision course with America's worsening drought crisis: nearly two-thirds of planned datacenters are heading to water-scarce regions, driven by economic incentives that ignore long-term environmental costs. Datacenter water demand is set to quadruple by 2028, placing enormous new pressure on communities already struggling with record drought conditions. As approvals for massive complexes continue in parched counties, the absence of legal frameworks to protect residential water access over industrial use could force hard societal trade-offs in the years ahead. This is a systemic risk that sits at the intersection of AI's physical footprint, climate change, and public resource governance.
