The Environmental Costs of the AI Boom
by
Large Seminar Room
Abstract: Artificial intelligence is reshaping the United States' energy landscape. Data centers now consume 4.4% of U.S. electricity and could reach 12% by 2028. Critical questions remain: How much CO2 does a single query generate? What drives this explosive demand growth? How do resource consumption costs compare with AI's benefits?
This talk contextualizes these questions with an overview of the technical developments enabling AI scaling and explores trajectories for future demand growth from this sector. Drawing on recent Union of Concerned Scientists modeling, we'll examine the allocation of AI compute resources, address efficiency paradoxes, and analyze policy solutions for meeting demand with clean energy while limiting climate and health impacts.
David Jaffe