AI Boom Powers a Surge: US Data Centers' Electricity Demand Set to Triple by 2028
The surging demand for artificial intelligence (AI) technologies is driving a significant increase in electricity consumption across the United States, according to recent research supported by the U.S. Department of Energy. This study indicates that electricity demand from data centers in the U.S. is anticipated to nearly triple within the next three years, potentially accounting for up to 12% of the nation's total electricity consumption. This escalation is primarily attributed to the intensifying integration of AI within various industries.
Authored by the Lawrence Berkeley National Laboratory, the report forecasts that by 2028, data centers' annual electricity usage could range from 74 to 132 gigawatts, depending heavily on the supply and demand for AI chips (GPUs). Currently, data centers already consume over 4% of the total electricity demand in the U.S. The deployment of GPU-accelerated servers, a trend gaining traction since 2017, has more than doubled the industry's electricity consumption over the past six years.
Avi Shultz, Director of the Industrial Efficiency and Decarbonization Office at the U.S. Department of Energy, highlighted that AI data centers are among the fastest-growing electricity consumers. The need for advanced chips and efficient cooling systems is crucial to accommodate the anticipated growth in data centers.
The study notes that as recently as 2016, AI servers consumed only about 2% of the energy in data centers. However, with the rapid expansion of AI technologies, particularly those incorporating large language models, there is a growing demand for both training and inference capabilities, driving the surge in electricity consumption.
Concern arises as the current pace of AI development outstrips the capacity of existing energy infrastructure, urging the need for comprehensive strategies to improve energy efficiency and address future electricity demands. The ongoing construction of new AI data centers, each potentially requiring up to 1 gigawatt of power, underscores the urgency of these developments, equivalent to powering all households in a city like Philadelphia.
