Data-center investment has become one of the largest capital-expenditure cycles in financial markets, with U.S. hyperscalers expected to deploy roughly $700 billion in 2026. This investment boom has raised concerns that large computing loads impose external costs on households through higher electricity prices. Using a 50-state panel for 2021-2024, we find no statistically significant evidence that data-center presence, installed capacity, or capacity expansion predicts residential electricity-price inflation across extensive-margin, intensive-margin, fixed-effects, and timing specifications. We propose an energyinternalization mechanism: hyperscalers can partially internalize incremental electricity demand through contracted or dedicated generation, including solar and wind energy. Consequently, gross datacenter electricity consumption need not translate one-for-one into net pressure on residential electricity supply. The findings suggest that the extraordinary AI capital-investment cycle has not, thus far, produced a detectable residential electricity-price externality.

Here is the article by Yosef Bonaparte, via the excellent Kevin Lewis.