The potential federal data center sites are “uniquely positioned for rapid data center construction, including in-place energy infrastructure,” according to the Department of Energy
The U.S. Department of Energy has identified 16 federally owned sites that could potentially host data centers and AI infrastructure and already have associated power resources available.
DoE has also put out an accompanying request for information on proposals for construction of those data centers, with the aim of supporting operational AI infrastructure by the end of 2027.
The move by the Trump administration expands, changes and seeks to move faster than one of the actions taken in the final days of the Biden administration, when the Departments of Defense and Energy were tasked with identifying and leasing at three least federal sites for “frontier AI data centers” that could be powered by clean energy such as geothermal, plus another list of 10 high-priority sites that could be made available to serve AI data centers with nuclear power—albeit by 2035.
“President Trump is committed to ensuring American leadership in artificial intelligence and Secretary Wright is delivering,” said White House Office of Science and Technology Policy Director Michael Kratsios. “The Trump Administration will unleash Federal resources to build out the data resources needed for an AI-powered future.”
So where are the proposed sites, and what’s already there? They are scattered around the country and in most cases, there are already existing federal science laboratories or decommissioned or active nuclear or weapons facilities at the locations.
Here are site summaries, detailed in the official DoE RFI:
–Idaho National Laboratory, where the DoE owns about 62,000 acres of an 890-square-mile site. INL, the RFI notes, “has a legacy of building and testing advanced technologies, including 52 nuclear reactors with four currently in operation, and is also a leader in integrated energy systems.”
-The Paducah Gaseous Diffusion Plant site in Kentucky. That 3,500-acre nuclear production site is undergoing remediation “for potential development of a data center” and is designed for up to 3GW in the Midcontinent Independent System Operator (MISO) power market, and 30 million gal/day of water. It includes 19 miles of road, nine miles of railroad tracks and is close to major railroads, roads and a river.
-The decommissioned Portsmouth Gaseous Diffusion Plant, which operated through 2001 and produced enriched uranium at a site of nearly 3,500 acres (the plant occupies about 1,200 acres). The site is designed for 2.2GW in the PJM Interconnection power market and 40 million gal/day of water, according to the RFI.
–Argonne National Laboratory’s site that is 23 miles from Chicago “could accommodate a 110-acre developable site for a future 1,000 MW AI data park on U.S. Department of Energy (DOE) land with an early target for operations by 2028.” The site has nearby high-voltage power access and Argonne gets 60% of its electricity from two large nuclear power plants that are within 50 miles of the location.
–Brookhaven National Laboratory, where a 5,322-acre site is located 75 miles east of New York City. Of that, around 90 acres is being considered for a data center site, which could be powered by a new 750 MW gas turbine plant for the data center.
-The Femi National Accelerator Laboratory already does large-scale data science and operates one of the largest data centers serving the U.S. DoE. The lab covers 6,800 acres, and there are about 110 acres of available land for development, plus “excellent access to high-speed networking” and has commercially available extra high-voltage (EHV) transmission infrastructure. With its experience and expertise in large-scale construction projects, and a readily available high-tech workforce, Fermi National Accelerator Laboratory is well-positioned and equipped to support major initiatives.
-The National Energy Technology Laboratory campuses in Morgantown, West Virginia and and Pittsburgh, Pennsylvania. The NETL’s Morgantown site includes 137 acres; its Pittsburgh site encompasses 57 acres.
-The National Renewable Energy Laboratory, or NREL, in Colorodo, where its Flatiron Campus “has enough land, power, water, and broadband capability to host a 100 MW data center that could be initiated as soon as this year (2025),” the RFI said, adding: “The site could support an integrated data center energy system test bed, that could be deployed later at scale at other locations. NREL has an 11-acre site located just west of the Flatirons main campus that would be an ideal location for a data center facility.”
–Oak Ridge National Laboratory “has utilities anticipated to be sufficient to support the rapid development of an AI data center,” with local power resources including hydro, nuclear and fossil fuel generation plants. The potential 550+ acre site has about 100 acres suitable for near-term development, the RFI noted, and is also “located approximately 5 miles from the proposed TVA Clinch River Small Modular Reactor (SMR) site, providing a future opportunity to capitalize on regional nuclear infrastructure up to 800MW in capacity.”
–Pacific Northwest National Laboratory in Richland, WA, which offers “a stable, dry climate with low humidity and minimal natural disaster risk, making it ideal for reliable operations” and access to hydroelectric and conventional nuclear power, with the possible addition of “small modular reactors (SMRs) envisioned by energy providers in the region could provide additional power for the data center.” Major hyperscalers already use eastern Washington as a location for large data centers, the RFI noted.
–Princeton Plasma Physics Laboratory, where a recent AI hub for New Jersey has already been announced and the site has 100 MW of energy capacity with district upgrade potential available.
–Los Alamos National Laboratory, which has “already responded to other recent calls for on/offsite data centers” and has work in progress to expand power capacity for AI mission support up to 70MW by 2027 and 160MW by the early 2030s. The RFI noted power capacity constraints at the site and suggested identify and deploy new on-premises power sources such as gas turbines or nuclear small modular reactors. A 40-acre site next to an existing power substation could be used for siting a high-density High Performance Computing facility suitable for AI.
–Sandia National Laboratories, which has locations in New Mexico, California, Hawaii; and Nevada, has identified two sites of roughly nine acres each.
–Savannah River Site, where the National Nuclear Security Administration (NNSA) operates a 310-square-mile site near Aiken, South Carolina, to supply and process tritium for nuclear weapons.
-The Pantex Plant, an ordnance facility constructed in the early 1940s, and was converted to the nation’s primary site for assembly and disassembly of nuclear weapons and is the Center of Excellence for High Explosives Manufacturing, according to the RFI. There is one potential site of about 380 acres and another 5,700 acres currently owned by Texas Tech University and leased by the federal government.
–Kansas City National Security Campus of the National Nuclear Security Administration, located near Kansas City, Missouri, which produces non-nuclear components for nuclear systems. There are 35 acres cleared of a proposed 50 acres with multiple owners in a primarily agricultural/low-density residential area.
