AMD and the U.S. Department of Energy (DOE) are joining forces again — this time to launch two massive next-gen systems at Oak Ridge National Laboratory (ORNL): the Lux AI Supercomputer and Discovery Supercomputer.
Together, they mark a new phase in America’s race to lead in artificial intelligence (AI) and high-performance computing (HPC) — and they come with a serious commitment: a combined $1 billion in public and private investment.
These systems are built not just to crunch numbers faster, but to train and deploy frontier AI models, accelerate energy research, advance medicine, and strengthen national security. They’re also core to the U.S. AI Action Plan, which aims to secure America’s leadership in AI innovation while keeping infrastructure and data sovereign.
“Discovery and Lux will leverage AMD’s high-performance and AI computing technologies to advance the most critical U.S. research priorities,” said Dr. Lisa Su, AMD Chair and CEO.
Lux: America’s First “AI Factory” Supercomputer
The first of the two, Lux, is set to go live in early 2026. Think of it as America’s AI Factory — a purpose-built platform designed to train, fine-tune, and deploy large AI foundation models for science, energy, and national security.
Co-developed by AMD, ORNL, Oracle Cloud Infrastructure (OCI), and HPE, Lux runs on AMD Instinct™ MI355X GPUs, EPYC™ CPUs, and Pensando™ networking — all optimized for massive AI workloads.
Once online, Lux will serve as a national hub for AI-driven research, powering breakthroughs in materials, biosecurity, energy systems, and advanced manufacturing.
“Winning the AI race requires creative partnerships that bring together the best of American tech and science,” said U.S. Energy Secretary Chris Wright. “Lux represents a new model of collaboration that turns innovation into national strength.”
Discovery: The Next Frontier in Supercomputing
Coming in 2028, Discovery will take over as DOE’s next flagship supercomputer at Oak Ridge — and it’s built for both AI and traditional scientific workloads.
Powered by next-gen AMD EPYC™ CPUs (codenamed “Venice”) and Instinct™ MI430X GPUs — a new series tuned specifically for sovereign AI and scientific computing — Discovery builds on the foundation laid by Frontier, the world’s first exascale supercomputer.
Discovery’s “Bandwidth Everywhere” design massively boosts memory and interconnect bandwidth, giving scientists and AI researchers a smoother, faster pipeline for complex workloads. Despite the big performance leap, it’s engineered for similar power efficiency and based on open-source standards for transparency and interoperability.
“Discovery will drive scientific innovation faster and farther than ever before,” said ORNL Director Stephen Streiffer. “It will let researchers go from problem to solution in record time.”
“Our collaboration with AMD is pushing supercomputing into the AI era,” added Antonio Neri, CEO of HPE. “With HPE’s converged AI and HPC architecture, Oak Ridge will achieve new levels of productivity and scale.”
Why This Matters
Lux and Discovery aren’t just about raw compute power — they’re part of a larger national effort to build a sovereign AI infrastructure that keeps cutting-edge science and innovation inside U.S. borders.
- Lux (2026): America’s first AI Factory for science and national security.
- Discovery (2028–2029): DOE’s next flagship HPC system, continuing the legacy of Frontier with next-gen AI acceleration.
- Partners: AMD, HPE, Oracle Cloud Infrastructure, ORNL, and the DOE.
- Mission: Power breakthroughs in energy, biology, materials, and national security through AI and HPC convergence.
When both systems are fully deployed, they’ll form a critical part of what DOE calls the American AI Stack — a secure, federated infrastructure designed to accelerate innovation for decades to come.