Datacenter Water Consumption
The freshwater draw from cooling-tower evaporation and direct-evaporative cooling at hyperscale datacenters — up to ~5 million gallons per day per facility for large GPU-heavy builds in hot climates. Active community and regulatory pushback in Bexar/Hood Counties (Texas), central Virginia, central Arizona, and parts of the Pacific Northwest.
Summary
The freshwater draw from cooling-tower evaporation and direct-evaporative cooling at hyperscale datacenters — up to ~5 million gallons per day per facility for large GPU-heavy builds in hot climates. Active community and regulatory pushback in Bexar/Hood Counties (Texas), central Virginia, central Arizona, and parts of the Pacific Northwest.
A siting constraint that compounds **Datacenter Power Shortfall** — a region with available power may still fail water-permit review. Combined with power, the two constraints define which geographies can absorb new S3 capacity at all.
- Closed-loop air cooling solves the water problem at the cost of higher capex and worse PUE — the trade-off is not free.
- Water draws are typically a public-utility-commission disclosure, not a federal one — the regulatory regime varies by state and county, which is what produces the patchwork of moratoria.
- "We just won't disclose" is no longer viable — ESG reporting standards now flag water as a top-tier sustainability disclosure for cloud infrastructure.
- Compounds Datacenter Power Shortfall as a geographic constraint on new S3 region capacity
scoped_toObject Storage, S3
Definition
The freshwater draw from cooling-tower evaporation and direct-evaporative cooling at hyperscale datacenters, currently estimated at up to **5 million gallons per day per facility** for large GPU-heavy builds in hot climates. Active community and regulatory pushback has emerged in **Bexar and Hood Counties (Texas)**, central Virginia, central Arizona, and parts of the Pacific Northwest where datacenter draws compete with municipal supply or agricultural rights. Several Texas counties debated formal moratoria in early 2026.
Recent developments
- US datacenter water demand projected 697M – 1.45B gallons/day by 2030 — comparable to NYC's daily supply. US server farms projected to require 697-1,451 MGD of new water capacity at up to $58B cost. NYC consumes ~1 billion gallons/day — so US datacenter demand at the high end matches an entire major US city. Per The Register — AI Datacenters May Gulp NYC's Daily Water Supply at Peak (March 2026).
- Georgia incident: 29M gallons consumed over 15 months before residents noticed. A 6.2M-sq-ft facility consumed 29M gallons of unauthorized water before residents reported low water pressure; officials refused to fine the builders — the regulatory enforcement gap is as much a 2026 issue as the consumption itself. Per Tom's Hardware — Georgia Data Center Used 29M Gallons of Water.
- 160+ new AI datacenters in past 3 years sited in water-scarce regions. The locational pattern is structurally bad — AI datacenters cluster in low-tax, low-regulation regions that frequently overlap with water-stressed ones. Phoenix, central Virginia, Arizona, parts of Pacific Northwest are all hot spots for new builds AND ongoing water stress. Per Brookings — AI, Data Centers, and Water.
- Phoenix imposed water-use restrictions on new datacenter developments. First major-US-metro formal regulatory response — Singapore + several California water districts have flagged datacenters as competing demand on groundwater. Singapore's national moratorium on new datacenters was partly water-driven. Per Brookings — AI, Data Centers, Water.
- Energy-vs-water efficiency tradeoff: evaporative cooling more energy-efficient but loses water as evaporated waste heat. The structural tradeoff at the cooling-tech level — evaporative cooling minimizes electricity but consumes water; closed-loop air cooling minimizes water but loses 30%+ PUE. Sites must pick which constraint binds. Per Environmental Law Institute — AI's Cooling Problem: How Data Centers Are Transforming Water Use.
- Hyperscaler "water positive by 2030" pledges face physics constraints. Google + Microsoft committed to being "water positive" by 2030 — returning more water to environment than consumed. The pledges are credible only if accompanied by infrastructure investment to actually return water at scale; otherwise they're carbon-offsets-for-water.
Connections 2
Outbound 2
scoped_to2Resources 3
Bexar County (San Antonio) case study documenting a single hyperscale facility's 5-million-gallon-per-day draw — the canonical reference for the per-facility consumption figure.
Lincoln Institute analysis of land and water impacts of datacenter growth — connects siting friction to land-use regulation across multiple jurisdictions.
Public Citizen advocacy piece around the Texas House data-center hearings — primary evidence of the regulatory pushback shaping the moratoria pattern.