Click here to download a PDF of the full memo.

Executive Summary

The future of artificial intelligence is bottlenecked by the enormous power demands of AI data centers. Current projections suggest that these facilities could require up to 5 GW of electricity by 2030—more than the output of America's largest nuclear plant. This demand has prompted hard questions about the optimal energy mix to power AI infrastructure, with major implications for energy costs, reliability, and national security.

Each potential power source faces distinct challenges. Nuclear power, while carbon-free and reliable, faces scaling constraints; for example, the recent expansion of the Vogtle plant, which added just 2.2 GW of capacity, required 16 years to complete. The viability of solar-plus-storage depends on continued steep declines in battery costs. Enhanced geothermal systems have shown promise, but remain largely unproven at scale. Meanwhile, tech companies’ varying commitment to climate goals will determine whether they invest in clean energy infrastructure or opt for more readily available natural gas.

The challenge extends beyond generation capacity. Building the gas pipelines and electrical transmission lines needed to deliver power to data centers involves complex regulatory processes that could constrain AI’s growth.

This memo examines the permitting, siting, and regulatory barriers facing these energy sources, and proposes solutions to overcome them. Understanding and reforming these barriers will be critical to meeting the AI industry's ambitious timeline for expansion.