In the November 2026 ranking, TOP500 kept LineShine at No. 1 with 2.198 exaflop/s on HPL, added 27 systems and placed Lux at No. 14. Announced at SC26 in Chicago on November 16, the 68th edition had the fewest new entries in the list’s history, according to TOP500.
LineShine held the top spot as 27 systems joined
The 27 newcomers made up 5.4% of the 500 entries, and the top ten kept the same order as in June 2026. The five highest-ranked systems were:
| Rank | System | Site | HPL performance |
| 1 | LineShine | National Supercomputing Centre in Shenzhen, China | 2.198 exaflop/s |
| 2 | El Capitan | Lawrence Livermore National Laboratory, United States | 1.809 exaflop/s |
| 3 | Frontier | Oak Ridge National Laboratory, United States | 1.353 exaflop/s |
| 4 | Aurora | Argonne National Laboratory, United States | 1.012 exaflop/s |
| 5 | JUPITER Booster | Jülich Supercomputing Centre, Germany | 1.000 exaflop/s |
Across all 500 entries, aggregate HPL performance reached 19.57 exaflop/s, 4.4% above June’s figure. The November list records the ranked systems and their configurations.
What TOP500 measures
TOP500 ranks systems by their results on High Performance Linpack (HPL), a benchmark that measures how quickly a system solves a dense set of linear equations. TOP500 says the result describes performance on that task, not a system’s overall performance (benchmark details).
A TOP500 first place is an HPL result, not a blanket verdict on every workload.
That distinction matters when comparing systems for a specific job: this ranking measures one defined workload, rather than performance across every scientific, AI or commercial task.
Lux ranked 14th; HANGANG led the newcomers
TOP500’s Lux record lists a measured HPL Rmax of 220.79 petaflop/s and a theoretical peak Rpeak of 323.90 petaflop/s. These are different measures: Rmax is Lux’s measured result for the ranking, while Rpeak is its theoretical peak. The record also lists 1,096,704 cores, AMD EPYC 9575F 64C 3.3GHz processors and AMD Instinct MI355X accelerators (Lux system record).
South Korea’s HANGANG was the highest-ranked newcomer at No. 12, with 342.6 petaflop/s on HPL. It also placed eighth on the separate HPCG benchmark, with 4.29 petaflop/s.
AI-compute estimates
TOP500 also cited Epoch AI estimates that installed AI-compute capacity exceeded the equivalent of 15 million Nvidia H100 GPUs. The estimates put five major U.S. hyperscalers’ share of global AI compute at 71% in the fourth quarter of 2025.