Static

Hyperscale data centers use 2% of US electricity

First reported by Ars Technica ·

The signal ●○○○ Compiled by AI from Ars Technica, the single source so far
Why you might care

Hyperscale data centers now consume 2% of US electricity, a figure that will only grow as AI demand escalates.

What happened

A recent study analyzed the electricity consumption and emissions of hyperscale data centers in the United States between May 2024 and April 2025. Researchers identified 403 such facilities, estimating their total power usage at 68 to 99 terawatt-hours annually, which accounts for 1.5 to 2.2 percent of the total US electricity consumption. This energy usage is comparable to the combined electricity needs of states like Wisconsin or Indiana. The study also estimated the associated greenhouse gas emissions to be 21 to 31 million metric tons of CO2, representing approximately 0.5 percent of total US fossil fuel CO2 emissions. The energy mix powering these centers was found to be only slightly cleaner than the US average, with about 54 percent coming from fossil fuels. The study acknowledges that it could not specifically isolate energy use for AI workloads but suggests future analyses could track the impact of AI-driven buildouts.

What it means

The study highlights a significant increase in data center energy demand, with total data center electricity use (excluding crypto mining) projected to reach 192 terawatt-hours in 2024, up from 76 terawatt-hours in 2018. Hyperscale facilities alone now account for one-third to one-half of this total. The energy mix powering these centers, which relies heavily on fossil fuels (54%), means their carbon footprint is substantial, even before considering less efficient on-site generation that is increasingly being deployed. This presents a growing challenge for grid stability and clean energy transitions as demand for compute power intensifies.

The findings underscore the difficulty in precisely tracking the energy and emissions impact of AI infrastructure, as current studies cannot isolate AI-specific compute. However, the study's methodology, which relies on reported electrical capacity and estimates of facility load factors, provides a baseline for future tracking. As hyperscalers continue their rapid expansion, often without concurrent planning for clean energy supply, the demand for electricity will likely outpace the grid's ability to meet it sustainably. This makes the energy source and efficiency of these facilities a critical factor in the broader climate and energy landscape.

AI-written summary. May contain errors.

Hyperscale