The power demand of U.S. data centers could nearly triple by 2028, driven by advancements in artificial intelligence (AI), according to an unpublished Department of Energy-backed report by Lawrence Berkeley National Laboratory.
AI’s reliance on energy-intensive GPUs and cooling systems is expected to push annual energy consumption to between 6.7 percent and 12 percent of the country’s total electricity use, up from the current 4 percent, Reuters news report said.
The report highlights a major shift in the industry since 2016, when AI servers accounted for only 2 percent of server energy use. The rapid deployment of GPU-accelerated servers since 2017 has already more than doubled the sector’s energy consumption.
Researchers call for annual or biannual tracking of data center trends and greater transparency in reporting energy usage and operational details. The report also recommends developing energy-efficiency strategies to mitigate the environmental impact of this growing demand.