Water is local: generalities do not apply (Jay Dietrich)

Overview

This 2025 article by Jay Dietrich explains where water is actually consumed in a data center: in the heat rejection loop, not in closed IT-side loops. Closed water or water-glycol loops serving CRAHs, rear doors or DLC via CDUs consume little or no water; consumption happens in open towers and evaporative/adiabatic coolers.

Its main message is that each site's water profile depends on climate, heat rejection technology and the local watershed. Newer evaporative/adiabatic systems with higher set points can free-cool 50% to 80% or more of the time. Spray or trickle-media adiabatic precooling can cut water use by half or more versus conventional open systems. In cool climates heat rejection may need water only 4-12 hours a day on 5-20 days a year.

In water-rich regions water-cooled heat rejection is usually most energy efficient, while in water-stressed areas dry free cooling with minimal (peak-shaving) or no water should be prioritised. For context, a typical US household uses about 110,000 US gal (420 m3) per year. The author argues impact must be judged against the safe withdrawal rate of the local supply. The related standard is ISO/IEC 30134-9 (WUE).

Key points

1. Closed water or water-glycol loops (CRAHs, rear doors or DLC via CDUs) consume little or no water; consumption happens in open towers and evaporative/adiabatic coolers.
2. Newer evaporative/adiabatic systems with higher set points can free-cool 50% to 80% or more of the time.
3. Spray or trickle-media adiabatic precooling can cut water use by half or more versus conventional open systems.
4. In cool climates heat rejection may need water only 4-12 hours a day on 5-20 days a year.
5. In water-rich regions water-cooled heat rejection is usually most efficient; in water-stressed areas prioritise dry free cooling with minimal or no water.
6. Context: a typical US household uses about 110,000 US gal (420 m3) per year.
7. Impact must be judged against the safe withdrawal rate of the local supply.

Using it in design

• Model annual water use from heat rejection type, set points and hourly climate rather than a fixed WUE.
• Add a site 'waterStress' field.
• Warn when open cooling towers are chosen in water-stressed locations and suggest dry or hybrid adiabatic options with peak-only water use.

Who should read it

Engineers choosing cooling systems and sites, and managers reporting on water and sustainability.

Related standards

• ISO/IEC 30134-9 (WUE)

Publication details

Authors

Jay Dietrich

Publisher

Uptime Institute Journal

Year

2025

Document language

English

Source type

Industry body

Access

Free to read

Licence

Copyright retained; citation only — the publisher’s terms apply to the original document

Citation