The Hidden AI Risk: Water Scarcity and Cooling
How much water do AI datacenters consume?
A medium-sized 100MW AI datacenter using evaporative cooling consumes 1.1 million gallons of water daily—equivalent to 3,000 homes. With AI workloads running at higher power densities (40-100kW/rack), water demand is intensifying just as key markets like Phoenix and Mesa impose stricter water allocations.
Last updated: January 2026
High Water-Stress Data Center Markets
Phoenix / Mesa, AZ
Risk Level: Critical. Recent groundwater restrictions limit new permits. Datacenters must move to air cooling (higher PUE) or recycled water.
Salt Lake City, UT
Risk Level: High. Great Salt Lake depletion has led to legislative scrutiny on industrial water use.
San Antonio, TX
Risk Level: High. Edwards Aquifer levels trigger frequent usage cuts.
The WUE Metric Explained
Water Usage Effectiveness (WUE) measures liters of water used per kWh of IT equipment energy.
- Efficient Evaporative1.2 - 1.8 L/kWh
- Legacy Towers2.0 - 2.5 L/kWh
- Air Cooled (Zero Water)0.0 L/kWh
- Liquid Cooling (Closed)~0.0 L/kWh
*Air cooled systems save water but increase energy use (higher PUE).
Solutions for the Future
Direct-to-Chip Liquid Cooling
Circulates fluid directly to GPUs. Uses a closed loop, consuming negligible water. Essential for >50kW racks.
Immersion Cooling
Submerges servers in dielectric fluid. Extremely efficient and zero water consumption, but operationally complex.
Recycled Water
Using "purple pipe" municipal wastewater for cooling towers instead of potable water. Required in Ashburn and Phoenix.
Frequently Asked Questions
Can we build datacenters without water?
Yes. Air-cooled chillers use zero water but consume ~15-20% more electricity. Closed-loop liquid cooling also uses negligible water and is becoming the standard for high-density AI clusters.
Does liquid cooling cost more?
CapEx is higher (specialized piping, CDUs), but OpEx is lower due to better energy efficiency and zero water bills. For H100/B200 deployments, liquid cooling is often effectively mandatory due to heat density.
How do I check a site's water risk?
Look for the "Water Stress Index" in our site readiness tool. It aggregates data from the World Resources Institute and local municipal allocation plans.
Evaluate Site Water Risk
Don't let water rights derail your project. Check the Hydrological Stress Score for any US zip code.
Check Water Risk Score →Explore More
Related Tools
Water Usage Calculator
Datacenter cooling water consumption analysis
Open Speed-to-Power WatchlistCSS (Curtailment Stress Score)
Real-time grid congestion and curtailment risk by zone
Open Speed-to-Power WatchlistSite Readiness Scores
Project readiness assessments across permitting, interconnection, construction
Buy Texas Time-to-Power PackPhase 1: Water Risk Readiness
Cooling infrastructure constraints and regional water scarcity are becoming primary bottlenecks for AI datacenter deployment. Evaluate your site's water usage effectiveness (WUE) and local municipal water allocations to mitigate regulatory permit risks.
Check Water Stress Scores
Use our interactive readiness map to evaluate water basin stress levels and cooling infrastructure feasibility across all major US markets.
Phase 2: Edge Infrastructure Risk
Edge readiness is critical for latency-sensitive AI inference workloads. Our Edge Risk Index evaluates fiber density, network latency, and colocation availability across 20+ major markets to help you optimize distributed inference deployments.
Explore Edge Readiness Scores
View latency metrics, fiber density, and edge colocation availability across all tracked markets.
Phase 3: Power Infrastructure Risk
Power availability is the primary constraint for AI datacenter deployment. Our Power Risk Index evaluates interconnection queues, curtailment exposure, and behind-the-meter strategies across 20+ major markets to help you de-risk power procurement.
Explore Power Risk Scores
View interconnection timelines, PPA structures, and curtailment risk across all tracked markets.