Small Bottle, Big Pipe: Quantifying and Addressing the Impact of Data Centers on Public Water Systems

Overview

This 2026 preprint by Yuelin Han, Pengfei Li, Adam Wierman and Shaolei Ren (arXiv:2603.02705) argues that the binding water constraint for data centres is often the peak-day withdrawal capacity of the local public water system, not annual volume. Using government records and utility data, it estimates the new water capacity US data centres may need through 2030 under several growth and WUE-improvement scenarios.

The results are striking: if 2024 water-use intensity persists, US data centres could need 697-1,451 million gallons per day (MGD) of new water capacity through 2030, comparable to New York City's average supply of about 1,000 MGD. With a 10% compound annual WUE reduction, the need drops to 227-604 MGD. The total valuation of new capacity is on the order of $10 billion, up to $58 billion in the high-growth case.

The paper documents high peaking factors for evaporative facilities. Data-centre daily peaking factors (maximum-day use over average-day use) often reach 6 and sometimes exceed 10. For a Phoenix colocation data centre with cooling towers, peak monthly WUE was about 1.46 times average monthly WUE; applying the regulatory 1.5 adjustment gives an estimated daily peaking factor of at least 2.19. The pressure zone serving a cooling-tower hyperscale data centre in The Dalles, Oregon, shows a measured peaking factor of 2.21.

Recommendations include reporting peak water use, community partnerships, a 'Water Capacity Neutral' ('Pipe Neutral') commitment, and coordinated water-power planning. Peak water capacity limits can push operators to dry cooling, which raises electricity demand during summer grid peaks. As a limitation, the study rests on US data and is a preprint that has not been peer reviewed; results do not transfer directly to other countries.

Key points

1. If 2024 water-use intensity persists, US data centres could need 697-1,451 MGD of new water capacity through 2030, comparable to New York City's average supply of about 1,000 MGD.
2. With a 10% compound annual WUE reduction, the need drops to 227-604 MGD.
3. The total valuation of new capacity is on the order of $10 billion, up to $58 billion in the high-growth case.
4. Data-centre daily peaking factors often reach 6 and sometimes exceed 10.
5. For a Phoenix colocation data centre with cooling towers, peak monthly WUE was about 1.46 times the average; the estimated daily peaking factor is at least 2.19.
6. The pressure zone serving a hyperscale data centre in The Dalles, Oregon, shows a measured peaking factor of 2.21.
7. Peak water capacity limits can push operators to dry cooling, raising electricity demand during summer grid peaks.
8. Recommendations: peak water reporting, community partnerships, a 'Water Capacity Neutral' commitment and coordinated water-power planning.

Using it in design

• Add a peak-day water demand output (annual average times peaking factor, default at least 2.2 for cooling towers, configurable up to about 6-10 for evaporative assist used only on hot days), and compare it with utility supply capacity in site screening.
• Flag designs that rely on evaporation only on design days as having low annual WUE but high peak demand.
• Pair the water peak with the dry-mode electrical peak, since they trade off on the same hot days.

Who should read it

Data centre designers, investors and planners doing cooling-water strategy and site water-capacity screening or negotiating with water utilities.

Related standards

• ISO/IEC 30134-9 (WUE)

Publication details

Authors

Yuelin Han; Pengfei Li; Adam Wierman; Shaolei Ren

Publisher

arXiv (UC Riverside / RIT / Caltech preprint)

Year

2026

Document number

arXiv:2603.02705

Document language

English

Source type

Scientific

Access

Free to read

DOI

10.48550/arXiv.2603.02705

Licence

arXiv distribution permission — reuse conditions remain with the rights holder

Citation