Energy Department selects 4 sites to fast-track AI infrastructure

The US Department of Energy has selected four federal sites for a groundbreaking initiative to accelerate artificial intelligence infrastructure development across the country. This marks the first concrete step in a broader federal strategy to leverage government-owned land and existing energy resources for AI data center construction.

The initiative addresses a critical bottleneck in AI development: the massive power and infrastructure requirements of modern data centers. Unlike traditional computing facilities, AI data centers consume enormous amounts of electricity—often requiring hundreds of megawatts of continuous power—while demanding ultra-reliable connections and specialized cooling systems. By opening federal sites with existing power infrastructure, the government aims to fast-track deployment while strengthening national security capabilities.

Four sites selected for immediate development

The Department of Energy chose Idaho National Laboratory, Oak Ridge Reservation in Tennessee, the Paducah Gaseous Diffusion Plant in Kentucky, and the Savannah River Site in South Carolina as the first locations for public-private partnerships.

“These sites are uniquely positioned to host data centers as well as power generation to bolster grid reliability, strengthen our national security, and reduce energy costs,” said US Energy Secretary Chris Wright.

The selection process began under the Biden administration but gained momentum under President Trump, who signed an executive order last week eliminating clean energy requirements from the original proposal. The order also instructs federal agencies to expedite environmental reviews and permitting processes for data center projects.

Idaho National Laboratory offers the most expansive opportunity, with the Department of Energy controlling approximately 62,000 acres of an 890-square-mile site. The facility has extensive experience with advanced energy technologies, having built and tested 52 nuclear reactors over its history, with four currently operational. This legacy makes it particularly well-suited for integrated energy systems that could power large-scale AI operations.

The Paducah site in Kentucky represents a strategic repurposing of Cold War infrastructure. This former uranium enrichment facility, which operated as a gaseous diffusion plant—a massive industrial complex that separated uranium isotopes for nuclear weapons—spans 3,500 acres and is currently undergoing environmental remediation. The site’s existing infrastructure can support up to 3 gigawatts of power capacity within the Midcontinent Independent System Operator (MISO) regional grid, along with 30 million gallons of daily water capacity. Its 19 miles of internal roads and nine miles of railroad tracks provide excellent logistics access.

Oak Ridge National Laboratory in Tennessee brings both immediate capacity and future potential. The facility has sufficient utilities to support rapid AI data center development, with access to hydroelectric, nuclear, and fossil fuel power generation. The potential 550-acre development area includes 100 acres ready for near-term construction. Significantly, the site sits just five miles from the proposed Tennessee Valley Authority Clinch River Small Modular Reactor location, which could eventually provide up to 800 megawatts of additional nuclear power capacity.

The Savannah River Site in South Carolina operates as a 310-square-mile facility managed by the National Nuclear Security Administration. Currently used for tritium production—a radioactive isotope essential for nuclear weapons—the site’s existing security infrastructure and power systems make it particularly attractive for sensitive AI applications with national security implications.

Twelve additional sites under consideration

Beyond these four selected locations, the Department of Energy identified 12 additional federal sites that could potentially host AI infrastructure. These facilities span the country and represent diverse approaches to repurposing government assets.

Several sites focus on leveraging decommissioned nuclear facilities. The Portsmouth Gaseous Diffusion Plant in Ohio, which produced enriched uranium until 2001, offers nearly 3,500 acres with infrastructure designed to handle 2.2 gigawatts of power within the PJM Interconnection—the regional transmission organization serving 13 states and Washington DC. The facility can access 40 million gallons of water daily, crucial for data center cooling systems.

National laboratories present another category of potential sites. Argonne National Laboratory, located 23 miles from Chicago, could accommodate a 110-acre development targeting 1,000 megawatts of AI data center capacity by 2028. The facility benefits from proximity to two large nuclear power plants within 50 miles, which currently provide 60 percent of Argonne’s electricity.

Brookhaven National Laboratory on Long Island offers 90 acres of its 5,322-acre site, positioned 75 miles east of New York City. Plans call for a new 750-megawatt gas turbine plant specifically designed to power data center operations, providing dedicated energy infrastructure for AI workloads.

The Fermi National Accelerator Laboratory in Illinois brings unique advantages through its existing expertise in large-scale data operations. The facility already operates one of the largest data centers serving the Department of Energy and covers 6,800 acres with 110 acres available for expansion. Its experience with high-performance computing and access to extra-high-voltage transmission infrastructure position it well for AI applications.

Regional facilities across the country offer additional opportunities. The National Energy Technology Laboratory operates campuses in Morgantown, West Virginia (137 acres) and Pittsburgh, Pennsylvania (57 acres), providing strategic locations in energy-rich regions.

The National Renewable Energy Laboratory in Colorado presents an immediate opportunity, with its Flatiron Campus capable of supporting a 100-megawatt data center as early as 2025. The facility could serve as a testing ground for integrated data center energy systems before scaling to other locations.

Pacific Northwest National Laboratory in Washington state capitalizes on the region’s advantages for data center operations. The facility offers a stable, dry climate with minimal natural disaster risk, plus access to abundant hydroelectric power and potential small modular reactor development. Major technology companies already operate large data centers in eastern Washington, demonstrating the region’s suitability for such facilities.

Several sites focus on nuclear weapons infrastructure. Los Alamos National Laboratory in New Mexico has ongoing work to expand power capacity for AI applications, targeting 70 megawatts by 2027 and 160 megawatts by the early 2030s. The facility has identified a 40-acre site adjacent to existing power infrastructure for high-density computing operations.

The Pantex Plant in Texas, originally constructed in the 1940s as an ordnance facility and later converted for nuclear weapons assembly and disassembly, offers substantial space with one potential site covering 380 acres and access to an additional 5,700 acres currently leased from Texas Tech University.

Strategic implications and next steps

This federal initiative reflects growing recognition that AI infrastructure development requires unprecedented coordination between government and private sector resources. Traditional data center development often faces lengthy permitting processes and power grid constraints that can delay projects for years. By offering pre-approved federal sites with existing power infrastructure, the government aims to accelerate deployment timelines significantly.

The Department of Energy plans to release formal solicitations detailing partnership requirements and application processes in the coming months, with the goal of selecting private development partners by year-end. Other agencies, including the Department of Defense, are conducting similar reviews of their land holdings for potential AI, semiconductor, and power infrastructure projects.

The initiative also addresses national security concerns about AI infrastructure development. By maintaining some level of government involvement in critical AI facilities, the program ensures that essential computing resources remain under US control while leveraging private sector expertise and investment.

For private companies, these partnerships offer access to sites that would otherwise require years of development and regulatory approval. The existing power infrastructure, security clearances, and environmental reviews at many federal facilities provide significant advantages over greenfield development projects.

The success of this program could establish a new model for public-private collaboration in critical technology infrastructure, potentially extending beyond AI to other strategic sectors requiring substantial power and security resources.