Home / Data center power & water estimator
Enter a data-center size and a U.S. market. See how much electricity and water it would really need — and whether the grid can deliver it.
Every figure is an order-of-magnitude estimate from public coefficients — not a site-specific engineering study.
Power. Facility power = IT load × PUE (Power Usage Effectiveness). We use PUE 1.20 for evaporative and 1.40 for closed-loop/air cooling (dry cooling spends more energy to avoid water). Annual energy = facility power × 8,760 h × 0.90 utilization (data centers run near-flat). PUE range: LBNL 2024 / Uptime Institute (global avg ~1.5; hyperscale ~1.2).
Comparisons. Homes = annual kWh ÷ 10,500 kWh (avg U.S. household/yr, EIA). Reactors = average draw ÷ (1,000 MW × 0.925 capacity factor) — a typical U.S. reactor running 24/7 (EIA).
Water. On-site withdrawal = annual kWh × WUE (Water Usage Effectiveness): 1.8 L/kWh evaporative, ~0.1 L/kWh closed-loop/air (Gridlas — "The water problem," hyperscaler ESG reports). Households = gallons ÷ 109,500 gal (~300 gal/day, EPA). Indirect water at the power plant can add ~10× more — not counted here.
Grid wait & workaround reflect each market's reported interconnection reality (national average ~4.5 yr to interconnect; LBNL). Water-stress flags follow WRI Aqueduct.
The Gridlas report ranks every major U.S. metro by capacity, queue and pipeline — with high-res maps and the underlying dataset. We'll email you the free 3-page executive summary now.