Hyperscale data centers use 2% of US electricity

Hyperscale data centers use 2% of US electricity

As an Amazon Associate I earn from qualifying purchases.

Woodworking Plans Banner

A current Lawrence Berkeley National Laboratory report approximated the overall electrical power usage of all information centers of any kind in the United States (minus crypto mining) at 76 terawatt-hours in 2018. In 2024, that number depended on 192 terawatt-hours. That suggests the hyperscale information centers determined by this research study represent something like one-third to half of all information center energy usage.

Electrons include carbon

The scientists likewise wished to approximate the greenhouse gas emissions related to producing all that electrical power. It’s relatively uncomplicated to get a sensible number here– determine the mix of generation in each information center’s area of the grid, and use that CO2 per watt-hour rate to an information center’s electrical energy usage.

This yields 21 to 31 million metric lots of CO2 for the year. That’s around 0.5 percent of overall United States nonrenewable fuel source CO2 emissions.

These information centers aren’t found in locations of specifically tidy energy– their energy mix was just somewhat cleaner than the United States average. About 54 percent of the electrical power was created by nonrenewable fuel source power plants (dominantly gas), with 25 percent from renewables, and 21 percent from nuclear.

This does not consist of any “behind-the-meter” generation on-site. The xAI information center in Tennessee especially ran a fleet of natural gas turbines throughout this research study’s time duration. Cleanview, which explains itself as a market intelligence business, has actually tracked a boom in on-site generation given that early 2025, consisting of 116 present or scheduled information. While this might not impact this research study’s numbers much, it is ending up being progressively substantial with time.

This matters since the kinds of generators being utilized are much less effective than energy power plants, producing 60 to 80 percent more greenhouse gases to produce the very same quantity of electrical power. (Conversely, if an information center developed solar generation rather, it might lower the carbon footprint of its energy.)

This research study can’t determine the energy usage of AI information centers alone, duplicating this analysis in subsequent years might expose the development associated to the AI buildout– at least as long as that buildout lasts.

Nature Sustainability, 2026. DOI: 10.1038/ s41893-026-01985-7 (About DOIs).

Learn more

As an Amazon Associate I earn from qualifying purchases.

You May Also Like

About the Author: tech