“How can AI data centers in the U.S. meet projected electricity demands by 2030?

Author(s)

Aiden M. Matano, University of San DiegoFollow

Name of Primary Faculty Advisor

Dr. Simon Croom

Publication Date

Fall 11-4-2025

Student Classification

Undergraduate

Disciplines

Energy Systems | Environmental Studies | Operations and Supply Chain Management | Power and Energy

Description / Abstract

The rapid expansion of artificial intelligence is driving a sharp increase in U.S. electricity demand, with AI-driven data centers emerging as a central contributor through 2030. This paper asks how U.S. AI data centers can meet projected electricity needs while simultaneously reducing carbon emissions. Using a supply-chain and systems perspective, it analyzes the full energy chain of AI data centers: upstream electricity supply and grid deliverability, midstream facility design and grid interaction, and downstream operational practices, waste management, and disclosure. Drawing on recent projections from the International Energy Agency, Lawrence Berkeley National Laboratory, and U.S. federal guidance, the paper shows that annual clean-energy procurement and incremental efficiency gains are insufficient to control emissions during the highest-impact hours. Instead, effective decarbonization depends on three coordinated levers: (1) grid-interactive infrastructure that incorporates liquid-first cooling, short-duration storage, and flexible load capabilities; (2) hourly, location-specific clean-energy procurement aligned with grid conditions and supported by market mechanisms such as demand response; and (3) operational and downstream practices that increase compute efficiency, enable heat reuse, manage accelerated hardware refresh cycles, and provide transparent, hourly emissions metrics. The analysis concludes that AI’s growing electricity demand is best understood as a design constraint rather than a fixed limit. When addressed through coordinated infrastructure, procurement, and operational strategies, U.S. AI data centers can realistically meet 2030 electricity demand while aligning growth in computing capacity with measurable reductions in carbon emissions.