Hyperscale AI infrastructure concentration in water-stressed regions creating cascading resource conflicts

str 8 4/28/2026 · 1 article

structural · economic · AI, Infrastructure · US

Analysis

Data centre developers are clustering facilities in drought-prone areas with spare grid capacity, creating simultaneous peaks in water and electricity demand that collide with agricultural cycles and municipal capacity, forcing communities to choose between AI infrastructure and food/water security.

Key actors

hyperscalersrural communitiesLawrence Berkeley Laboratory

Source article

The great American data centre divide

Financial Times — AI, Data, Robotics and Digital Power · 4/28/2026 · extracted in run pdf-import-2026-04-28-1779224753682-92 · 5/19/2026, 10:29:53 PM

"Around two-fifths of all US data centres are located in areas of high water stress, according to S&P Global." [two-fifths]

This quantifies the structural mismatch: 40% of data centres are in water-stressed regions, and the article shows these same regions (Arizona, Texas, West Virginia) are experiencing severe drought and municipal water shortages, creating direct resource competition.

Reasoning from this article

The article documents that hyperscaler water consumption is forecast to reach 16-33 billion gallons annually by 2028 (excluding indirect electricity-generation water use), while towns like Davis, West Virginia have treatment plants producing only 250,000 gallons daily. The timing overlap is critical: both agriculture and data centres peak water/energy demand in summer. This is not a solvable trade-off but a structural incompatibility that will force regional triage decisions.