Accelerating Large-Scale Grid Infrastructure Projects to Win the AI Race

The U.S. Department of Energy (DOE) aims to quickly scale power grid infrastructure and support growing electric loads to sustain U.S. artificial intelligence (AI) leadership. Constructing new grid infrastructure before 2030 is constrained by permitting, regulatory, and supply chain challenges. Therefore, leveraging existing assets is the most practical near-term strategy to sustain or add generating capacity. This paper provides comments in response to DOE's request for information to identify projects to (1) enable load growth, (2) high-priority geographic areas for targeted DOE investment, and (3) grid infrastructure constraints.

Drawing on insights developed through the study of AI and frontier AI data center power needs, as well as the U.S. energy infrastructure ecosystem supporting them, the authors suggest that the most-feasible single-site example, the Rockport Plant in Indiana, could represent an estimated 4.2 GW of firm capacity for expanding the availability of power to the U.S. grid by 2030. They also propose that two medium-potential AI data center sites — the Portsmouth Gaseous Diffusion Plant within the Ohio River Basin and the Wansley Power Plant in the southeastern United States — could be elevated to high potential with DOE investment. They conclude with a cross section of 66 barriers that the authors have developed to showcase the key issues that constrain substantially increasing electric load.