Topics / Data centers
Data centers and water
The U.S. data center build pipeline runs faster than host-utility water capacity can follow. A 100 MW hyperscale running at typical PUE consumes between 1.3 and 4.3 billion gallons of cooling water per year. Across the announced pipeline, that is more than any host utility can deliver on its current capital-improvement schedule. A public-private partnership between the host utility and an infrastructure capital partner is the cleanest single solution path.
The numbers
Hyperscale water intensity sits in the 1.5 to 5.0 gallon-per-kWh range, with evaporative-cooled designs at the high end and air-cooled or closed-loop designs at the low end. A 100 MW facility at typical 1.6 PUE draws 1.3 to 4.3 billion gallons per year. A 200 MW facility doubles that. At metro scale, the announced data center pipeline draws billions of gallons annually in direct competition with municipal residential and industrial demand.
The Water Hawk constraint pipeline tracks 11 water-constrained U.S. metros as of mid-2026, three of which were added in Q2 alone (Phoenix MSA, Las Vegas MSA, Reno MSA). The median host-utility capacity gap to the announced data center load is roughly 38% across the constrained cohort.
The structural problem
The host utility carries the entire delivery obligation. The data center developer does not buy water; the utility sells water to the data center customer like it sells to any other industrial customer. When a 200 MW hyperscale lands in a 8-MGD-headroom system, the utility is the entity asked to source the incremental capacity.
The structural mismatch is the procurement timeline. A data center build takes 24 to 36 months from siting to operation. A utility capital expansion takes 5 to 9 years from rate-case filing to commissioning. The math does not work; either the data center stalls, or the utility takes on rate-base risk at a pace its commission has not approved.
The P3 partnership solution
A public-private partnership between the host utility and an infrastructure capital partner closes the timeline gap. The structure: the partner finances and constructs the incremental treatment, conveyance, and reclaim capacity required by the data center load; the utility operates the system; the data center signs a take-or-pay contract for the new capacity at a rate that clears the partner's IRR; existing rate-payers are insulated.
The Water Hawk operator network has worked the math across multiple constrained metros. The median P3 delivery acceleration is 5.2 years versus the utility-only path. The median rate-base avoidance is $1.40 per residential customer per month across a 20-year term. Both numbers are conservative; the high end of the band runs materially better.
How Water Hawk reads this
Water Hawk maintains a per-cluster constraint-pressure read across the U.S. cluster topology, a per-host-utility capacity profile, and a vetted P3 partner shortlist in the operator network. The bespoke Data Center Water Report ($5,000 standard, $12,500 deep, $25,000 with site visit and partner introduction) is the single-engagement deliverable that maps the constraint math to specific candidate sites. The Data Center Intelligence addon ($499 per month) is the recurring subscription for continuous cluster monitoring and supplier-opportunity scoring. The Water Nexus bundle ($899 per month) combines Data Center Intelligence with RNG Intelligence at a small discount.
Read further
- Q2 2026 Data Center Water Impact quarterly , the dated snapshot with the top stressed metros, the implied annual draw, and the P3 partnership pro-forma.
- Bespoke Data Center Water Report service , the engagement that walks the constraint math for your specific candidate metros.
- Sample cluster detail , illustrative cluster page showing what a subscriber sees.