October 19, 2025
NOTUS Perspectives: AI Requires Massive Allocations of Energy. Will Other Sectors Suffer?
This article was originally published on NOTUS.
AI will force hard choices about how energy is allocated in America. Energy demand for developing advanced AI models is skyrocketing. One leading AI company expects to need 5 gigawatts of firm, reliable power (roughly equivalent to the output of five nuclear reactors) to train a single advanced model in 2028, with up to 25 gigawatts required nationwide for advanced AI training.
On the current trajectory, the government may need to prioritize energy for AI over other uses to ensure the United States remains the global leader in advanced AI.
While the Trump administration has taken aggressive steps to speed development — streamlining permitting and allowing use of federal lands — those measures alone will not meet the scale of energy demand. On the current trajectory, the government may need to prioritize energy for AI over other uses to ensure the United States remains the global leader in advanced AI. Tools like the Defense Production Act’s priorities and allocations authorities, or the Department of Energy’s power to override state transmission decisions, make such prioritization possible.
Read the full article on NOTUS.
More from CNAS
-
Commentary
Technology & National Security
The Rise of the Answer MachinesThis article was originally published in Financial Times. Every spring, I take red-eyes from Austin, Texas, to Oxford, England, to teach a graduate seminar on AI and philosoph...
By Brendan McCord
-
Video
Technology & National Security
Selling H200s to China Erodes Main U.S. AdvantageA new report says China could buy twice as much AI computing power as it can produce domestically if Nvidia H200 chips are allowed there. Janet Egan from the Center for a New ...
By Janet Egan
-
Commentary
Technology & National Security
CNAS Insights | Unpacking the H200 Export PolicyAI Chips for China With two new policies, President Donald Trump has implemented his pledge to allow sales of NVIDIA’s H200 AI chips to China in exchange for a quarter of the ...
By Janet Egan & James Sanders
-
Podcast
Indo-Pacific Security / Technology & National Security
AI and Policy, Both Foreign and DomesticIn an episode recorded just before Christmas, Darren interviews Janet Egan, Senior Fellow and Deputy Director of the Technology and National Security Program at CNAS, about AI...
By Janet Egan
