Conference Agenda

Affordable electricity is essential for sustainable economic growth, but demand from artificial intelligence data centers threatens to undermine this objective. The net effect of data centers on electricity prices is theoretically uncertain. We provide early empirical evidence, studying the United States over the period 2015-2024. We document that data center market entry accelerated sharply in the post-AI period following the year 2021. We map data centers to geographically defined utility territories and implement a shift-share identification strategy to estimate causal effects on retail electricity prices. Our approach exploits the timing of the AI-associated data center construction boom with utility territories' pre-existing fiber infrastructure. We find that a standard deviation of ten data centers raises household electricity prices by 13% and that a single hyperscaler facility increases prices by 4%. Effects are significantly smaller in areas served by a nonprofit utility, with excess energy, and with greater retail competition. Demand-driven channels seem to be the dominant mechanism, although grid infrastructure spending may become more important in future years. Our results highlight the current challenges electricity markets face as we progress further into the AI era.