The U.S. Department of Energy is exploring a new model for AI infrastructure: data centers built inside national laboratories to accelerate scientific discovery rather than serve commercial cloud customers. Pacific Northwest National Laboratory (PNNL) in Richland, Washington, has requested information for a possible AI data center as soon as 2028, signaling a broader pivot in how federal research institutions are approaching artificial intelligence and computing power.

What Makes a Government AI Data Center Different from Commercial Ones?

PNNL's proposed data center would be fundamentally different from the hyperscale facilities that Amazon Web Services and Microsoft are building across the country. Rather than serving paying customers, this facility would focus on training and operating large language models (LLMs), which are AI systems trained on vast amounts of text data, for scientific and national security missions with stringent security requirements.

The lab confirmed in a statement that it is "considering the possibility of a small data center due to increased focus on AI at PNNL." The initial power requirement would be modest by hyperscale standards, starting at 2 megawatts in 2028, though it could expand to 40 megawatts in the future. For context, average-sized data centers typically require 5 to 10 megawatts of power, while large hyperscale facilities demand 100 megawatts or more.

How Does This Fit Into the Genesis Mission?

PNNL's data center plans are directly tied to the Genesis Mission, an ambitious federal initiative led by the Department of Energy to dramatically accelerate the pace of scientific discovery. Energy Secretary Chris Wright has compared the project to the Manhattan Project, the World War II effort that developed nuclear weapons. The Genesis Mission brings together all 17 U.S. national laboratories to use artificial intelligence systems combined with supercomputers and emerging quantum technologies to transform how science is conducted.

Under this initiative, PNNL will focus on three key research challenges: strengthening the nation's electric grid, analyzing nuclear materials related to national security, and integrating AI workflows to improve experiments. The lab has already begun demonstrating this approach with the Anaerobic Microbial Phenotyping Platform (AMP2), a first-of-a-kind system that combines robotics and artificial intelligence to conduct research using microbes like bacteria and fungi.

"What that means is we can ask a question, in the language of a mass spectrometer instrument and get an answer in the language of genes. The goal is to have a platform that can speak multiple data languages, integrate those data and tell us which experiments to do next," explained Chris Oehmen, a technical lead for the project.

Chris Oehmen, Technical Lead, Pacific Northwest National Laboratory

Steps to Understanding PNNL's Data Center Strategy

• Power Flexibility: Battelle, which operates PNNL for the Department of Energy, is exploring multiple power options including utility power, on-site solar, nuclear energy, and power purchase agreements to ensure the facility can operate reliably in a region with limited utility distribution capacity.
• Modular Design: The request for information specifically asked whether the AI data center could be housed in containerized or modular structures rather than traditional buildings, allowing for faster deployment and scalability.
• Heat Recovery: PNNL is investigating whether waste heat from the data center could be reused for industrial processes, turning an operational challenge into a resource opportunity.
• Security Integration: The facility would be located on secure Department of Energy land along the Columbia River that is part of the Hanford nuclear site, providing the stringent security requirements needed for national security research.

Why Is the Tri-Cities Region Becoming an AI Data Center Hub?

PNNL's data center plans are just one piece of a larger infrastructure boom in the Tri-Cities area of Washington. At least three other major data center projects are proposed or being considered for the region, transforming it into an unexpected hub for AI computing infrastructure.

Amazon Web Services intends to develop a $5 billion cluster of 16 data centers at Wallula Gap Business Park to support its AI initiatives, having recently closed a $34 million deal to purchase more than 500 acres from the Port of Walla Walla. Atlas Agro, known for its low-carbon fertilizer plant plans, is facilitating a $500 million data center development on 275 acres near the Framatome nuclear fuel manufacturing campus. Additionally, Trammell Crow Company, a global real estate firm with a data center division, is examining a site at the Lewis and Clark Ranch development area in West Richland.

The concentration of these projects reflects the region's unique advantages: proximity to the Hanford nuclear site, existing power infrastructure, secure federal land, and a skilled workforce. PNNL's proposed facility would complement rather than compete with these commercial ventures, focusing on advancing scientific research rather than serving cloud customers.

The shift toward government-operated AI data centers represents a recognition that artificial intelligence infrastructure is becoming as critical to national competitiveness as nuclear weapons or space exploration once were. By embedding AI computing directly into national laboratories, the federal government is positioning itself to lead in scientific discovery while maintaining the security and control necessary for sensitive research.