AMD expands to 202 systems on the November 2026 TOP500
AMD now powers 202 of the TOP500's 500 systems, up from 177 a year earlier, as Oak Ridge's Lux debuts at No. 14 with 220.79 petaflops of measured HPL performance.
- By
- Rebecca Stone
- Filed
- Channel
- AI & Compute
- Read
- 3 min read
By Rebecca Stone
AMD chips power 202 of the 500 systems on the November 2026 TOP500, about 40% of the list released Monday, Nov. 16 at the SC26 conference. That tally rose from 191 on the June list and 177 in November 2025, a year-over-year gain of 25 systems, according to TOP500 data reported by Tom's Hardware.
Only 27 systems joined. The top 10 did not move.
What does Lux add to Oak Ridge's AI work?
Hewlett Packard Enterprise's Lux AI cluster at Oak Ridge National Laboratory debuted at No. 14 with 220.79 petaflops of measured HPL performance, a result confirmed by the November ranking rather than a vendor projection.
The Next Platform calculated a theoretical peak of 323.9 petaflops from AMD EPYC 9575F processors and Instinct MI355X GPUs. One figure is measured. The other is theoretical.
Tom's Hardware reported that Lux contains more than 4,000 MI355X GPUs across more than 500 nodes, with 288 GB of HBM3E memory on every GPU.
Lux is the first of two systems AMD is building for the Department of Energy under a $1 billion partnership. The second, Discovery, is due in 2028 as Frontier's successor at Oak Ridge, with AMD EPYC Venice CPUs and Instinct MI430X GPUs planned for the system.
HPE described Lux as “the first dedicated AI system to come online” for the Energy Department's Genesis Mission. Genesis Mission receives half of Lux's 3.5 million node-hours a year; the other half is available for proprietary use.
What does the unchanged top 10 show?
China's LineShine, based at the National Supercomputing Centre in Shenzhen, kept the No. 1 position it won in June at 2,198.4 petaflops. TOP500 said there were “no changes in the top 10” from June.
The published Rmax figures for the first five are measured HPL performance, in petaflops:
- LineShine, China: No accelerator listed, CPU system; 2,198.4 petaflops.
- El Capitan, US: AMD MI300A; 1,809.0 petaflops.
- Frontier, US: AMD MI250X; 1,353.0 petaflops.
- Aurora, US: Intel GPU Max; 1,012.0 petaflops.
- JUPITER Booster, Germany: Nvidia GH200; 1,000.0 petaflops.
AMD chips run in four of the top 10: El Capitan and Frontier in the US, plus Eni's HPC7 and HPC6 in Italy. Lux followed at No. 14, rounded to 220.8 petaflops in the table.
How much fresh capacity reached the list?
Only 27 systems joined the November ranking, about half the usual number and the lowest turnover in the 68 lists published since 1993, according to The Next Platform.
The largest newcomer was Hangang at South Korea's KISTI, with 588.3 petaflops of peak performance on Nvidia Grace Hopper chips; Lux ranked second among the additions and was AMD's biggest newcomer. Because the 588.3 figure is peak performance, it is not directly comparable with Lux's measured HPL result.
Five systems using Nvidia Grace Hopper or Grace Blackwell chips supplied 51.8% of the new capacity, The Next Platform said. Across the full list, Nvidia GPUs supplied 52.9% of accelerated FP64 performance, versus AMD's 37.7%, according to the same analysis.
AMD's four top-10 placements show breadth across two US national-laboratory systems and two Italian Eni systems; Lux extends that footprint into a dedicated AI installation. The accelerator split is narrower, with Nvidia ahead in both accelerated FP64 performance and incoming capacity despite AMD's lead in overall system count.
Nvidia used SC26 to announce Rubin HPC, a GPU for scientific computing due in 2028.
The 2028 schedule puts AMD's Discovery and Nvidia's Rubin HPC on the same competitive horizon; the next TOP500 cycle will show whether AMD's 202-system footprint translates into more measured FP64 performance and new capacity.
Source: mlq.ai
More from Rebecca Stone
Show full bio
Correspondent covering technology and product news in Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.
1,037 articles
Related articles
bull-wins-387-8m-contract-for-next-gen-lumi-ai-supercomputer-6d0b12ec
Bull Wins €387.8M Contract for Next-Gen Lumi AI Supercomputer
hpe-s-gx5000-and-k3000-daos-storage-to-power-discovery-at-oak-ridge-1ebd50de
HPE's GX5000 and K3000 DAOS Storage to Power Discovery at Oak Ridge
microsoft-commits-billions-to-amd-helios-racks-for-azure-inference-1f4c04ad
Microsoft Commits Billions To AMD Helios Racks For Azure Inference
ibm-amd-and-oak-ridge-put-quantum-and-hpc-under-one-roof
IBM, AMD and Oak Ridge Put Quantum and HPC Under One Roof



