On January 14, 2025, with just days remaining before a presidential administration change, President Joseph R. Biden issued an Executive Order on Advancing United States Leadership in Artificial Intelligence Infrastructure (EO) that directs federal government agencies to solicit proposals for the development of “frontier” artificial intelligence (AI) data centers on federal lands.[1] While the EO remains for now, it could soon be rescinded by the new administration. Recognizing the substantial amount of energy infrastructure needed to train AI models, the EO is heavily focused on incentivizing the expeditious development of low or zero carbon energy resources to power these AI data centers. Set out below is a summary of key parts of the EO. Morgan Lewis’s data centers team is available to discuss the EO and assist with data center development and related issues.
Immediate Actions
The EO directs US federal agencies to act promptly to identify eligible federal lands for frontier AI data center development. By February 28, 2025, the Secretaries of the US Department of Defense (DoD) and US Department of Energy (DoE) must identify at least three sites on federal lands that they believe could be leased by the end of 2027 to nonfederal entities for the construction and operation of frontier AI data centers and associated energy facilities.
By March 15, 2025, the US Secretary of Interior must identify land suitable for granting or issuing rights of way to private-sector entities to construct and operate energy facilities that can deliver electricity to frontier AI data centers, and designate at least five regions of lands or subsurface areas that are managed by the Department of Interior as “Priority Geothermal Zones” that may be eligible for expedited permitting treatment. Once sites are identified, proposals from nonfederal entities to construct one or more frontier AI data centers at the sites will be solicited through a competitive process beginning on March 31, 2025, with any winning proposals announced by June 30, 2025.
Energy Obligations for Selected Developers
Developers selected under the solicitation process will be responsible for paying the full cost of building, operating, and maintaining the AI infrastructure, including the costs of new data centers, associated power facilities, transmission development, computing equipment, computing integrated circuits (i.e., graphics processing units or GPUs) and upgrades.
The EO requires selected developers to bring online sufficient zero or low carbon energy generation resources to meet the power demands of those data centers. Eligible generation types include geothermal, nuclear fission, nuclear fusion, solar, wind, hydroelectric, hydrokinetic (including tidal, wave, and current), and marine energy resources. Fossil fuel generation resources are also eligible if paired with carbon capture, utilization, and storage technologies that achieve carbon dioxide capture rates of 90% and permanently sequester the captured carbon dioxide.
Sites with ready access and proximity to high-voltage transmission infrastructure, for example, with planned, unconstructed generation facilities that have executed interconnection agreements but have not made offtake arrangements, are to be prioritized. The EO also directs DoE to develop model contracts for distributed energy resources that, if deployed, could increase the local grid’s capacity to support AI infrastructure.
The DoD and DoE must also seek to facilitate the deployment of additional nuclear power by, among other things, identifying up to 10 sites that are most conducive to quickly and safely deploying nuclear power capacity that can be readily available to serve AI data center electricity demand by December 31, 2035.
Energy Infrastructure and Expedited Permitting
The EO seeks to expedite the development of energy infrastructure needed to support frontier AI data centers by mitigating some of the bottlenecks that can snarl project development. The EO directs applicable permitting agencies to “exercise all applicable authorities” to expedite the process of permits and approvals needed for the construction and operation of AI infrastructure on federal lands (e.g., those required under the National Environmental Policy Act, Clean Air Act, and Clean Water Act) by the end of 2025.
The EO also aims to circumvent delays in transmission development and generator interconnection. The DoE, along with other federal agencies, is directed to promote the development of transmission infrastructure, including consideration of whether any locations around frontier AI data centers should be designated as national interest electric transmission corridors and, as such, be entitled to permitting advantages.
To expedite grid interconnections, the EO directs the Secretary of Energy to take actions to identify surplus interconnection capacity, available transmission capacity for interconnecting generators, and opportunities for “clean repowering,” i.e., siting new zero or low carbon generation at existing points of interconnection serving existing fossil-fueled generation. The Secretary of Energy is also directed to work with transmission providers, transmission organizations, and developers to identify best practices for load interconnection, an issue that is receiving close scrutiny at federal and state levels due to the growth of data centers and other large grid-connected loads.
Next Steps
The EO contains sweeping and ambitious goals for accelerating large-scale AI infrastructure development on federal lands but may be rescinded, replaced, or modified by the incoming Trump-Vance administration. Regardless of the outcome, the EO signals the federal government’s latest acknowledgement of the importance of AI and data center infrastructure to national security, safety, economic competitiveness, and technological advancement.
[1] The EO defines “frontier AI data center” as an AI data center capable of being used to develop, within a reasonable time frame, an AI model with characteristics related either to performance or to the computational resources used in its development that approximately match or surpass the state of the art at the time of the AI model’s development.