Data-center electricity use is substantial and growing, but the totals depend on geography, facility definitions and whether a figure is measured or projected. The International Energy Agency estimates global data centers used about 415 TWh of electricity in 2024. In the United States, Lawrence Berkeley National Laboratory estimated 176 TWh of electricity use and 66 billion liters of direct water consumption in 2023. Operators can reduce demand through more efficient servers and workloads, better airflow and cooling controls, and cooling choices suited to local power and water conditions.
How much electricity do data centers use?
The available estimates below cover different geographies and come from different modeling approaches. They should not be read as a like-for-like comparison.
As an Amazon Associate I earn from qualifying purchases.
| Geography and period | Electricity use | Share of electricity supply | What the figure means |
|---|---|---|---|
| Global, 2024 | About 415 TWh | About 1.5% of global electricity use | Historical estimate from the International Energy Agency (IEA, 2025). |
| United States, 2023 | 176 TWh | 4.4% of U.S. electricity use | Estimate from Lawrence Berkeley National Laboratory (LBNL, 2024). |
| Global, 2030 | About 945 TWh | Just under 3% of global electricity use | IEA Base Case projection (IEA, 2025), not a prediction of a certain outcome. |
| United States, 2030 | 649 TWh | Not stated for the Reference Case | LBNL Reference Case projection (LBNL, 2025). Its sensitivity scenarios put data centers at 9.5%–15.3% of U.S. electricity use by 2030. |
The 2030 figures are scenarios, not measured consumption. The IEA’s alternative global cases differ materially, and LBNL’s U.S. range reflects sensitivity to assumptions. Future demand depends on factors including equipment deployment, utilization, efficiency and cooling technology. The estimates also use different models and boundaries, so differences between the global and U.S. figures do not establish that one region’s facilities are more power-intensive.
Free tools Windows power users keep installed
One-click scans. No signup required.
How much water do data centers use?
Direct water use at U.S. facilities
LBNL estimated U.S. data centers directly consumed 66 billion liters of water in 2023. This is on-site water consumption, not a global total and not a measure of all water associated with supplying data-center electricity. The cited sources do not establish a comparable current global water total.
What water-use metrics include
Water Usage Effectiveness (WUE) is water consumed divided by IT-equipment electricity, commonly reported in liters per kilowatt-hour. Site WUE counts water used at the facility. Source WUE also includes water used to generate the electricity the facility consumes. A value is meaningful only alongside its boundary, period and method.
LBNL’s 2024 report modeled U.S. average site WUE rising to about 0.45–0.48 L/kWh after 2023. This is a modeled aggregate, not a universal benchmark for individual facilities. Site WUE alone cannot show total water impact: a waterless cooling system can consume more electricity, and the water footprint of that electricity depends on the power supply.
What do PUE and WUE tell operators?
Power Usage Effectiveness (PUE) is total data-center facility electricity divided by IT-equipment electricity. It indicates infrastructure overhead; it does not measure how efficiently or usefully the IT equipment performs its workload. WUE relates water consumption to IT electricity, but its interpretation changes depending on whether it is site or source WUE.
Both metrics vary with cooling design, climate and operating practice. To assess an intervention, operators should compare the same facility boundary and time period, and account for workload changes. Track facility and IT electricity separately, report site water, and include source water when the necessary data is available. PUE by itself cannot show whether servers are doing useful work efficiently.
Rank #3
How can operators reduce electricity and water demand?
There is no single best combination of measures for every facility. LBNL’s 2025 review found that modeled workload-level water use varied by more than 10,000-fold across conditions. It identifies server efficiency, grid water intensity, server utilization, cooling type, infrastructure efficiency, climate, inactive-server share and refresh cycle as important factors. Operators should assess actions against their own workloads, equipment, climate, water stress and electricity supply.
1. Improve server and workload efficiency
Use more efficient servers and raise utilization where workload and reliability requirements allow; review inactive-server share and equipment refresh decisions. These choices affect both the electricity required for computing and the amount of cooling infrastructure it needs. Measure workload output alongside power: reducing facility overhead is not the same as improving the efficiency of the computing work itself.
2. Widen temperature and humidity operating ranges where safe
Setpoints can be more restrictive than equipment requires, increasing chiller and cooling-tower demand. DOE’s Federal Energy Management Program says: “Raising the set point for temperature and increasing the range of humidity control set points in the space will result in energy savings and will also result in water savings by reducing the amount of heat that needs to be dissipated by the evaporative process at the cooling tower system.” The statement appears in the DOE’s 2019 guidance, Cooling Water Efficiency Opportunities for Federal Data Centers.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsWidening acceptable ranges can lower energy and water use where the equipment and operating conditions permit. Verify the applicable envelope for the hardware, reliability classification, altitude and site rather than applying a generic setpoint.
3. Improve airflow management
Arrange racks and contain hot and cold aisles so hot exhaust does not mix with cool supply air. Effective separation can support higher chilled-water temperatures and reduced airflow, which can reduce chiller energy. Results depend on the facility’s design and operation.
4. Use economizers when site conditions allow
Air-side economizers use suitable outdoor air for cooling. Water-side economizers use a heat exchanger to bypass or reduce chiller-compressor operation. Feasibility and savings depend on climate, outdoor-air quality, humidity, controls and system configuration; neither approach is a universal fit.
5. Manage cooling towers to limit unnecessary water loss
Cooling towers reject heat through evaporation, while blowdown removes water with concentrated dissolved minerals. DOE guidance explains that increasing cycles of concentration can reduce makeup-water needs. In its example, moving from three to six cycles reduces makeup water by 20% and blowdown by 50%. These are cooling-tower example figures, not guaranteed whole-facility savings; the appropriate operating level depends on water chemistry and system limits.
6. Evaluate the electricity–water trade-off before selecting cooling
In the configurations studied by LBNL, air-cooled chillers used no site water but more energy than water-cooled chillers. That trade-off means optimizing site WUE alone can shift demand to electricity rather than reduce overall resource impacts. Compare facility electricity and cooling overhead, direct site water, source water associated with electricity, local climate and water stress, workload and equipment requirements, reliability and maintenance complexity. Consider capital and operating costs when credible site-specific figures are available; the evidence does not support a universal ranking of cooling strategies.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
