Lux Opens Scientific AI Computing at Oak Ridge
Oak Ridge’s new Lux supercomputer opened to scientific users on November 16, adding a dedicated system for work that combines artificial intelligence with conventional scientific simulation.
Published November 17, 2026Intelligence / Frontier Tech / Experimental Systems

Oak Ridge’s new Lux supercomputer opened to scientific users on November 16, adding a dedicated system for work that combines artificial intelligence with conventional scientific simulation. Its first research allocations reach into fusion power-plant design, materials development and forecasts of water conditions that affect electricity generation. ORNL announcement
The practical opportunity is to bring learning from scientific data closer to the calculations researchers already use to study the physical world. A project allocated to Lux, for example, aims to forecast water availability, streamflow and river temperature in the Tennessee Valley Authority region. Those conditions matter to both hydropower and plants that depend on water for cooling. AMD announcement
Lux also introduces a shared-investment model for access. Half of its annual 3.5 million node-hours is dedicated to the Department of Energy’s Genesis Mission; the other half is reserved for paid public- and private-sector partners. AMD will use some of those cycles for development and testing and intends to make a portion available to new partners. Access framework
For participating organizations, this creates a way to run large scientific AI workloads on national-laboratory infrastructure. The paid-access route is aimed at substantial computing projects: OLCF specifies a minimum commitment of 175,000 node-hours over six months. Access framework
Two computing traditions in one system
HPE built Lux using AMD EPYC processors, Instinct MI355X accelerators, Pensando networking and ROCm software. HPE describes a direct-liquid-cooled system with Ethernet networking and support for multiple research users. HPE system announcement
Lux supports Slurm and Kubernetes, software systems used to schedule work in supercomputing and cloud environments. That combination is intended to accommodate simulations alongside model training, model use and automated research workflows. Researchers with allocations on both Frontier and Lux can also use OLCF’s Orion storage to make simulation data available for training. ORNL announcement · System resources
“Lux will provide researchers with additional capacity and flexibility,” Georgia Tourassi, ORNL’s associate laboratory director for computing and computational sciences, said in HPE’s announcement. She identified Earth-system prediction, materials, biology and fusion energy among the intended beneficiaries. HPE announcement
What the performance number measures
Lux entered the November TOP500 ranking at No. 14, with a reported Rmax of 220.79 petaflops. This is the measured result used for the ranking, distinct from the much larger low-precision arithmetic figures advertised for AI workloads. TOP500 results
A simulation and a neural network may use different numerical precision, software and patterns of communication among processors. TOP500 provides a common benchmark; useful performance for a particular scientific project still depends on its workload.
Access at research scale
One node-hour means using one compute node for one hour. Partners buy time and apply it to jobs of different sizes. OLCF specifies the DOE User Facility rate for proprietary access but publishes no price on the resource page; applicants must discuss arrangements with OLCF. Partner access
Genesis projects will evaluate whether combining AI with simulation and experimental data improves prediction, experimentation or the pace of discovery. Lux supplies the computing capacity for that work; the projects’ scientific results will establish where the combined approach is most useful. ORNL research plan
Primary sources
Sources · 5
- https://www.olcf.ornl.gov/2026/11/16/lux-genesis-mission/
- https://newsroom.amd.com/news/amd-powered-lux-supercomputer-opens-scientific-users/
- https://www.olcf.ornl.gov/olcf-resources/compute-systems/lux/
- https://www.hpe.com/us/en/newsroom/press-release/2026/11/hpe-built-lux-ai-cluster-debuts-at-no-14-on-the-top500.html
- https://top500.org/lists/hpcg/hpcg-november-2026/