May 20, 2025
Artificial Intelligence Infrastructure on DOE Lands
Summary
Maintaining America’s lead in AI data centers is critical for U.S. AI dominance. However, this dominance is at risk. Access to energy has become a critical bottleneck to building AI data centers domestically—driven in part by lengthy permitting delays. Unaddressed, and absent other controls, these bottlenecks will increasingly incentivize companies to build critical AI infrastructure abroad. Making DOE lands available offers a promising contribution to addressing this bottleneck, particularly if accompanied by streamlined permitting for AI facilities and energy infrastructure.
We recommend DOE continue to make land available and attractive for AI-related infrastructure development as a national security imperative, and offer the following recommendations to strengthen America’s AI infrastructure through effective use of DOE sites.
DOE should prioritize security in its siting decisions
As AI systems grow more powerful, the data centers training and hosting them will become prime targets for theft and sabotage, which could jeopardize America’s national security and technological edge. Siting AI infrastructure on DOE lands provides a unique opportunity to strengthen data center security. DOE should work with other government agencies and private sector partners to develop robust security controls that go beyond baseline compliance with commercial standards. Specifically, DOE should solicit input on:
- Security practices that could advance state-of-the-art data center security against sophisticated adversary threats.
- Opportunities for co-sited AI infrastructure developers and operators to collaborate with DOE on relevant R&D initiatives.
DOE must pursue efficient and predictable environmental review
Achieving DOE’s ambitious timelines will require navigating the environmental review process with efficiency and predictability. DOE should look for opportunities to mitigate potential delays and litigation risks by establishing clear, legally robust NEPA implementation guidance for this initiative upfront. It should consider, for example:
- Guidance on leveraging programmatic review and other opportunities to reduce the need for redundant site-specific analysis.
- A standardized approach for identifying and evaluating the applicability of relevant categorical exclusions.
- Guidance for considering the applicability of national security-related authorities to streamline permitting.
- Establishing transparent and realistic project schedules that include regulatory milestones.
Read the Full Response
More from CNAS
-
Reports
Technology & National Security
Red LinesChinese advanced artificial intelligence (AI) systems pose a serious and growing threat to U.S. national security. At least seven Chinese developers now produce systems with f...
By Daniel Remler
-
Commentary
Technology & National Security
Taiwan Is the Key to AI DominanceA country determined to win the defining technological race of the century can’t allow its chief rival to control the industrial base on which that race depends....
By David Feith
-
Reports
Technology & National Security
American AI Companies Can’t Get Enough ChipsIn 2026, artificial intelligence (AI) chip production has become a binding constraint on the pace of the AI compute buildout. Demand for computing power to train and deploy ad...
By James Sanders, Janet Egan & Rory Madigan
-
Reports
Technology & National Security
Off TargetThe pace of progress in frontier artificial intelligence (AI) capabilities shows no sign of slowing. Frontier models offer transformative potential for national security—from ...
By Caleb Withers, Jay Kim & Ethan Chiu
