AI Water Consumption Calculator
Estimate the water consumption linked to AI data centres, cooling systems and AI infrastructure based on server load, cooling method and operating assumptions. A UK-focused educational tool for businesses, researchers, journalists and policy analysts.
Important: This tool provides educational estimates only. Real water use varies heavily by facility design, cooling system, water reuse and local climate. Do not treat these results as engineering, planning or environmental advice.
Facility assumptions
What this water use is equivalent to
Insights
- This AI facility could consume approximately 11,589,480 litres (11,589.5 m³) of water per year.
- Changing from evaporative cooling to air cooling could reduce estimated water consumption by roughly 94.4%.
- That is comparable to about 4.6 Olympic swimming pools, or the daily water of roughly 97.2 UK households each day.
- Water use is highly dependent on cooling method, local climate and data centre design — two facilities with identical electricity use can have very different water footprints.
- This calculator provides estimates only and should not be treated as engineering advice.
Why AI data centres use water
AI servers run dense clusters of GPUs and accelerators that draw large amounts of electricity and convert almost all of it into heat. To keep hardware within safe operating temperatures, many facilities use water-based cooling — particularly evaporative cooling, which removes heat by evaporating water. The higher the server density and the more continuously the hardware runs, the more heat must be removed and the more cooling water may be required.
What affects water consumption
Estimated water use depends on several factors working together: the number of AI servers, the power each server draws, the utilisation rate (how busy the hardware is), the PUE (how much overhead energy the facility adds for cooling and infrastructure), and — most importantly — the cooling method. Air and closed-loop liquid cooling use relatively little water, while evaporative cooling can use considerably more. Local climate also matters: warmer regions place more strain on evaporative systems.
Why this matters in the UK
The UK is seeing rapid growth in AI and data centre development at the same time as growing concern about water stress in parts of the country, especially the South East. New AI facilities can add local pressure on water supplies, raise planning and environmental questions, and intersect with infrastructure constraints around both power and water. Understanding the potential scale of water demand helps businesses, communities, journalists and policy analysts ask better questions about new developments.
How this calculator works
IT load (kW) = servers × watts per server × utilisation ÷ 1,000. Facility load = IT load × PUE. We multiply facility load by 24 hours to get daily electricity (kWh), then scale by your operating days for annual electricity. Water use is estimated by multiplying electricity consumption by the water usage intensity in litres per kWh. Defaults assume air cooling 0.1, liquid 0.3, hybrid 0.7, evaporative 1.8 and unknown 1.0 litres/kWh — all editable.
Equivalents assume a UK household uses roughly 327 litres of water per day, a bath holds about 80 litres and an Olympic swimming pool holds about 2,500,000 litres.
Limitations
Please treat these figures as directional estimates, not measurements. In particular:
- Real water use varies heavily by facility design, efficiency and operating practices.
- Some data centres use closed-loop cooling that recirculates water and consumes very little.
- Some facilities use little direct water but still have an indirect water footprint through the electricity they consume (water used in power generation).
- This tool is for education, comparison and early-stage analysis only.
Related Tools & Resources
FAQ
Explore more AI energy tools
Estimate electricity, cost, carbon and infrastructure impact with the full AcrossAI suite of UK AI energy calculators.
Browse AI Energy Tools