What a kilowatt-hour costs in water, state by state
LBNL Water IMPACT balancing-authority data for 2022, mapped to states · as of 2026-08-26
- The water consumed at power plants per kilowatt-hour varies by more than 19x across U.S. states, from 23.01 L/kWh in Montana to 1.23 L/kWh in Texas. Source: LBNL Water IMPACT (Siddik, Shehabi, Rao & Marston 2024), balancing-authority hourly data via HydroShare, mapped to states with EPA eGRID2022
- The same data center consumes about 10 times as much power-plant water in Arizona (12.72 L/kWh) as in Texas (1.23 L/kWh), before it cools anything.
- Load-weighted across all balancing authorities these figures give a national average of 4.28 L/kWh for 2022, against the 4.35 L/kWh LBNL published for U.S. electricity use overall. The two are computed independently, which is the check that this table is on the right basis.
- Excluding hydropower reservoir evaporation the national figure falls to 1.09 L/kWh. Both columns are published below, because in the West almost the entire difference is evaporation off dams.
Every page on this site tracks what a kilowatt-hour costs. This one tracks what it costs in water. The water a data center draws through its cooling is the visible part; the larger part is evaporated at the power plants filling its meter, and that number is a property of the grid it plugs into, not of the building.
Water intensity of electricity, by state
| # | State | Consumed (L/kWh) | Excl. hydro | Withdrawn (L/kWh) | Grid regions |
|---|---|---|---|---|---|
| 1 | Montana | 23.01 | 0.93 | 31.2 | NWMT 65%, BPAT 12%, AVA 6% |
| 2 | South Carolina | 18.10 | 1.20 | 125.6 | DUK 48%, SCEG 28%, SC 17% |
| 3 | Arizona | 12.72 | 1.50 | 15.5 | SRP 60%, TEPC 18%, WALC 11% |
| 4 | Nevada | 10.52 | 0.55 | 23.4 | NEVP 85%, CISO 8%, WALC 5% |
| 5 | North Carolina | 10.49 | 1.24 | 132.0 | DUK 53%, CPLE 41%, PJM 4% |
| 6 | Utah | 8.61 | 1.04 | 14.2 | PACE 84%, LDWP 15%, WAUW 1% |
| 7 | Georgia | 8.29 | 1.52 | 27.3 | SOCO 99%, SEPA 1%, SCEG 0% |
| 8 | California | 7.45 | 0.63 | 29.3 | CISO 85%, LDWP 6%, BANC 4% |
| 9 | Alabama | 6.86 | 1.35 | 42.0 | SOCO 71%, TVA 29% |
| 10 | Idaho | 6.79 | 0.54 | 9.6 | IPCO 50%, AVA 21%, BPAT 20% |
| 11 | Oregon | 6.75 | 0.33 | 7.8 | BPAT 62%, PGE 21%, PACW 11% |
| 12 | Washington | 5.44 | 0.24 | 6.3 | BPAT 62%, CHPD 8%, PSEI 7% |
| 13 | Wyoming | 5.19 | 1.09 | 10.7 | PACE 61%, WACM 37%, WAUW 2% |
| 14 | Mississippi | 4.96 | 1.29 | 57.7 | TVA 43%, SOCO 29%, MISO 28% |
| 15 | Colorado | 3.88 | 0.99 | 15.3 | PSCO 61%, WACM 39%, PACE 0% |
| 16 | South Dakota | 3.87 | 1.19 | 48.2 | SWPP 58%, MISO 39%, WACM 3% |
| 17 | Wisconsin | 3.84 | 1.57 | 57.1 | MISO 100% |
| 18 | Iowa | 3.84 | 1.56 | 57.0 | MISO 99%, SWPP 1%, AECI 0% |
| 19 | Minnesota | 3.84 | 1.57 | 57.0 | MISO 100%, SWPP 0%, PJM 0% |
| 20 | Louisiana | 3.83 | 1.55 | 56.6 | MISO 97%, SWPP 3% |
| 21 | North Dakota | 3.83 | 1.32 | 51.5 | MISO 60%, SWPP 39%, WACM 1% |
| 22 | Nebraska | 3.72 | 0.94 | 43.9 | SWPP 100%, WACM 0% |
| 23 | Kansas | 3.70 | 0.94 | 44.0 | SWPP 100% |
| 24 | Arkansas | 3.66 | 1.40 | 54.8 | MISO 80%, SWPP 12%, AECI 4% |
| 25 | New Mexico | 3.63 | 1.14 | 17.5 | PNM 43%, AZPS 26%, SWPP 24% |
| 26 | Tennessee | 3.46 | 0.96 | 78.5 | TVA 99%, PJM 1% |
| 27 | Michigan | 3.41 | 1.51 | 56.1 | MISO 77%, PJM 23% |
| 28 | Connecticut | 3.33 | 0.63 | 47.4 | ISNE 100% |
| 29 | Massachusetts | 3.33 | 0.63 | 47.4 | ISNE 100% |
| 30 | Maine | 3.33 | 0.63 | 47.4 | ISNE 100% |
| 31 | New Hampshire | 3.33 | 0.63 | 47.4 | ISNE 100% |
| 32 | Rhode Island | 3.33 | 0.63 | 47.4 | ISNE 100% |
| 33 | Vermont | 3.33 | 0.63 | 47.4 | ISNE 100% |
| 34 | Oklahoma | 3.33 | 0.88 | 45.9 | SWPP 84%, AECI 7%, ERCO 6% |
| 35 | Indiana | 3.25 | 1.48 | 55.8 | MISO 68%, PJM 32% |
| 36 | Missouri | 3.10 | 1.11 | 56.0 | MISO 46%, SWPP 27%, AECI 24% |
| 37 | Illinois | 2.46 | 1.38 | 56.7 | PJM 72%, MISO 26%, GLHB 2% |
| 38 | Kentucky | 2.33 | 1.33 | 44.1 | LGEE 50%, TVA 23%, PJM 21% |
| 39 | District of Columbia | 1.96 | 1.30 | 52.9 | PJM 100% |
| 40 | Delaware | 1.96 | 1.30 | 52.9 | PJM 100% |
| 41 | Maryland | 1.96 | 1.30 | 52.9 | PJM 100% |
| 42 | Ohio | 1.96 | 1.30 | 52.9 | PJM 100% |
| 43 | Pennsylvania | 1.96 | 1.30 | 52.9 | PJM 100% |
| 44 | Virginia | 1.96 | 1.30 | 52.9 | PJM 100%, TVA 0% |
| 45 | West Virginia | 1.96 | 1.30 | 52.9 | PJM 100% |
| 46 | New Jersey | 1.90 | 1.23 | 52.6 | PJM 90%, NYIS 10% |
| 47 | New York | 1.33 | 0.63 | 49.5 | NYIS 100%, ISNE 0% |
| 48 | Florida | 1.28 | 1.00 | 52.2 | FPL 53%, FPC 19%, TEC 9% |
| 49 | Texas | 1.23 | 0.80 | 54.1 | ERCO 86%, SWPP 7%, MISO 6% |
Why the Southwest is at the top
The high figures are not cooling towers. They are reservoirs. Water-intensity accounting assigns the evaporation off a hydroelectric reservoir to the electricity that reservoir generates, and an open lake in a desert evaporates enormously. That is why Arizona reads 12.72 L/kWh with hydro allocated and 1.50 without, and why the same gap appears across the Colorado River and Columbia River systems.
There is a circularity worth naming. The reason Arizona's grid carries so much water is that a large share of its power comes from dams on the Colorado River, and the reservoirs behind those dams are Lake Mead and Lake Powell, the same reservoirs whose shrinking drove the 2027 and 2028 cuts. Those cuts, in full.
Allocation is a choice, not a measurement. Reservoirs evaporate whether or not anyone draws power from them, so the hydro-allocated column should be read as the water footprint attributed to a kilowatt-hour under a standard accounting convention, not as water a data center causes to disappear. The excluding-hydro column is the harder floor. This site publishes both and takes no position on which is correct.
How this was built
Hourly water-intensity factors for 58 balancing authorities come from the LBNL Water IMPACT tool and its underlying HydroShare dataset for 2022, published with Siddik et al. (2024). Each balancing authority is load-weighted to an annual mean, then mapped to states using EPA eGRID2022 plant generation shares. States with less than 60% of their generation in covered balancing authorities are left out rather than guessed, which is why 2 are missing. Balancing authorities cross state lines, so a state figure is the mix of grids operating in it, not a reading at any one meter.
Download this table (CSV) · Water hub → · Electricity by state →