The U.S. Department of Energy is partnering with Advanced Micro Devices (AMD) in a combined public-private investment worth over $1 billion to develop and deploy two new artificial intelligence supercomputers at Oak Ridge National Laboratory (ORNL) in Tennessee. The initiative aims to accelerate American research in fields ranging from thermonuclear fusion and cancer treatment to materials science and national security. The collaboration will produce two systems, named Lux and Discovery, that will expand the nation’s sovereign AI infrastructure and support the U.S. AI Action Plan.
This partnership signals a strategic shift in how the government procures and accesses high-performance computing (HPC). Instead of a direct government purchase, this model involves private-sector partners, including AMD, Hewlett Packard Enterprise (HPE), and Oracle, who will contribute capital and hardware. In return for providing the land and infrastructure at ORNL, the Department of Energy will share computing time on the machines with its corporate partners. This approach is designed to foster innovation, reduce direct government costs, and serve as a template for future collaborations between national laboratories and private industry.
A New Model for National Computing
The joint venture between the Department of Energy and technology firms like AMD and HPE marks a departure from traditional government procurement. The more than $1 billion arrangement leverages private capital for the construction of the machines, while the government provides the secure facilities and operational infrastructure at Oak Ridge National Laboratory. This shared-access model ensures that the immense computational power of the new systems is available for both federally funded scientific research and private-sector innovation. A DOE official noted that the two AMD-based supercomputers are expected to be the first in a series of similar partnerships with private companies and government labs. The model is intended to accelerate the deployment of cutting-edge technology by sharing the financial burden and aligning public research goals with industry expertise. This allows the government to maintain its leadership in scientific computing while providing corporate partners with access to world-class research tools.
Lux: The Near-Term AI Powerhouse
The first of the two supercomputers, Lux, is on a remarkably fast track, with deployment scheduled for early 2026. Its rapid rollout was highlighted by AMD Chair and CEO Dr. Lisa Su as the fastest she has seen for a computer of its size, underscoring the agility of the new partnership model. Lux is being co-developed by ORNL, AMD, Oracle Cloud Infrastructure, and HPE. When operational, ORNL Director Stephen Streiffer stated that Lux will offer approximately three times the AI processing capacity of the laboratory’s current leading systems.
Hardware and Architecture
Lux’s performance will be driven by a suite of AMD technologies. At its core are the AMD Instinct MI355X accelerators, which are GPUs specifically designed for the demanding mathematical operations inherent in training and running large AI models. These will be paired with AMD EPYC central processing units and AMD Pensando advanced networking technology to ensure efficient data flow and processing throughout the system. The architecture is designed to support a secure, open, and efficient AI software stack, which will help strengthen America’s broader innovation base.
Immediate Research Applications
Upon its launch, Lux will immediately begin work on several critical national priorities. Researchers will use its power to advance studies in fusion and fission energy, discover new quantum materials, modernize the national energy grid, and spur innovation in advanced manufacturing. In medicine, the system will simulate and model potential cancer treatments with molecular-level precision. Energy Secretary Chris Wright expressed hope that such computational power could help turn many cancers that are today considered terminal into manageable conditions within the next five to eight years.
Discovery: Charting the Exascale Frontier
Following Lux will be the even more powerful Discovery supercomputer, which is slated to arrive at ORNL in 2028 and become fully operational in 2029. Discovery is designed to be the DOE’s next flagship system, building on the breakthroughs of its predecessor, Frontier, which was the world’s first computer to officially break the exascale barrier (a quintillion calculations per second). The system will be based on HPE’s next-generation Cray Supercomputing GX5000 architecture. Its mission is to extend U.S. leadership in both HPC and AI by converging the two fields with quantum computing systems.
Next-Generation Technology
Discovery’s architecture will be powered by next-generation AMD components, including EPYC processors codenamed “Venice” and the advanced AMD Instinct MI430 series accelerators. The MI430 series is specifically engineered to merge the capabilities of traditional HPC processors with the features required to run sophisticated AI applications. This hybrid capability is critical for modern scientific research, where massive simulations are increasingly intertwined with data-driven AI models. The machine will be built to deliver significant improvements in performance and energy efficiency over current systems, enabling new scales of scientific inquiry.
Accelerating Scientific Breakthroughs
The primary mission for both Lux and Discovery is to solve some of the world’s most complex scientific challenges. In the field of fusion energy, the computers will be used to model the highly unstable plasmas needed to replicate the process that powers the sun. Secretary Wright noted the immense difficulty of recreating a star’s core on Earth but expressed optimism that the computational speed of these AI systems could lead to “practical pathways to harness fusion energy in the next two or three years,” a timeline significantly more ambitious than many current estimates.
Beyond energy, the systems will be instrumental in national security, where they will help manage the U.S. nuclear arsenal and advance defense technologies. In medicine and biology, the computers will accelerate drug discovery and help scientists understand the genetic underpinnings of diseases. By enabling researchers to model complex systems with unprecedented fidelity, Lux and Discovery are poised to drive fundamental breakthroughs across a wide spectrum of science and technology.