The short answer
Peak water demand is a design-day estimate, not an annual average. Start with either nameplate IT capacity plus an expected utilization, or an expected operating IT load. Then add operating hours, cooling architecture, local design conditions, and a documented Water Usage Effectiveness assumption. Test more than one scenario.
There is no universal data center water consumption per megawatt. The result changes with climate, cooling system, facility efficiency, utilization, and the measurement boundary. LBNL identifies each of these as a material driver.
A practical estimation framework
The U.S. Department of Energy defines WUE as annual site water use divided by annual IT equipment energy:
Site water use in liters ÷ IT equipment energy in kWh
For a defined planning period, the same relationship can support a scenario estimate using either form:
Nameplate IT capacity in kW × expected utilization × operating hours × site WUE in L/kWh
Expected operating IT load in kW × operating hours × site WUE in L/kWh
The result is liters of site water use for the selected period. Do not apply utilization twice. Convert to gallons only after the load, time period, utilization treatment, and WUE boundary are fixed. This is a planning calculation derived from the WUE relationship, not an engineering design standard.
Inputs that change the result
| Input | Why it matters |
|---|---|
| IT load | Sets the energy base behind the water estimate |
| Utilization | Nameplate capacity is not the same as actual operating load |
| Time period | A peak hour, peak day, and annual total answer different questions |
| Cooling architecture | Evaporative, dry, hybrid, and liquid-cooling systems have different site-water profiles |
| Climate | Wet-bulb temperature and economizer hours affect cooling demand and evaporation |
| WUE boundary | Site-only and source-inclusive water accounting are not interchangeable |
| Redundancy and operations | Equipment sequencing, maintenance, and resilience requirements affect actual demand |
What published scenarios can and cannot tell you
EPRI reports that one 150 MW hyperscale scenario at 0.36 L/kWh uses about 340,000 gallons per day. A water-cooled chiller scenario at the same load and 2.8 L/kWh uses about 2.7 million gallons per day.
That range is useful because it shows the sensitivity of the result. It is not a benchmark for every 150 MW facility. The WUE assumption, climate, utilization, and cooling configuration must be stated whenever a scenario is used.
Annual water use is not peak water demand
Annual WUE is useful for consistent performance tracking. It can hide the hottest hours, seasonal peaks, commissioning conditions, and operating transitions that matter for utility and storage planning.
For a design-day range:
- Define the peak operating period.
- Set a realistic IT-load and utilization range.
- Choose WUE assumptions that match the cooling architecture and climate.
- Calculate low, expected, and high cases.
- Review makeup capacity, storage, discharge limits, water cost, and resilience together.
- Confirm the result with the project’s mechanical engineer and local utility.
Move from site demand to recoverable water
Peak site water demand and recoverable blowdown are related, but they are not the same quantity. Evaporation rejects heat and leaves the site as vapor. Blowdown is liquid water discharged to control mineral concentration.
Where evaporative cooling remains the best operating choice, NONA RECOVER continuously demineralizes cooling water and recovers 90% of cooling tower blowdown on site. Use the calculator to model site-specific water and cost inputs, then review the assumptions with Nona’s technical team.