China’s LineShine has kept its position as the world’s fastest supercomputer in the November 2026 TOP500 rankings, as a record low number of new systems suggests a shift in how the tech industry is investing in HPC.

The Shenzhen-based system delivers a remarkable 2.198 exaflops of computing performance, about 21% faster than the second-place El Capitan system at Lawrence Livermore National Laboratory in California. LineShine first claimed the top position in June, ending a nine-year period in which China had not held the No. 1 ranking.

The latest TOP500 results, announced at the SC26 conference in Chicago, show that the world’s leading supercomputers remain largely unchanged. But the rankings also highlight the growing role of AI-focused infrastructure, including a new system developed by HPE and AMD.

HPE’s Lux AI cluster, deployed at Oak Ridge National Laboratory in Tennessee, debuted at No. 14 with 220.79 petaflops of performance. Designed for both AI training and inference, Lux is the first dedicated AI system deployed to support the U.S. Department of Energy’s Genesis Mission. The Genesis initiative aims to combine AI with advanced computing resources and scientific instruments to support research in energy, national security and other fields.

Lux uses AMD Instinct MI355X GPUs and EPYC CPUs, supported by HPE ProLiant Compute XD685 servers and Juniper networking technology. Its Ethernet infrastructure supports 800G connectivity, helping reduce network bottlenecks during demanding AI workloads.

The system will provide 3.5 million node-hours annually, with half allocated to Genesis research. HPE has also committed $37 million in noncash support to the initiative.

Supercomputer Growth Shifts Toward Upgrades

The TOP500 rankings recorded just 27 new entries, or some 5.4% of the list. This is the lowest number of new entrants since the rankings began. Despite the limited turnover, computing capacity continues to grow ever more powerful. Aggregate performance across the 500 systems rose 4.4%, while the performance threshold for the 50th-ranked system climbed 12%.

The slow pace of new installations suggests that organizations are concentrating on upgrading existing systems rather than replacing them with completely new systems.

Among the newcomers, South Korea’s HANGANG ranked 12th. The system uses NVIDIA Grace processors and GH200 Superchips.

The U.S. continues to hold three of the top four positions, including El Capitan at 1.809 exaflops, Frontier at 1.353 exaflops and Aurora at 1.012 exaflops. Germany’s JUPITER Booster ranks fifth with 1 exaflop.

LineShine, meanwhile, demonstrates China’s ability to build advanced supercomputers using domestic technology. Its architecture includes 40,960 Chinese-designed LX2 processors distributed across 92 cabinets, providing nearly 13.8 million processing cores.

AI Infrastructure Changes Computing Investment

The rankings also demonstrate an overall change in computing investment. Research group Epoch AI estimates that five major U.S. cloud companies controlled 71% of global AI computing capacity in late 2025.

Worldwide AI capacity has surpassed the equivalent of 15 million NVIDIA H100 GPUs, indicating that vast computing resources are being deployed outside traditional TOP500 systems.

Accelerator-equipped machines now account for 58.2% of the TOP500, highlighting the growing importance of specialized processors in driving higher performance.

Energy efficiency remains another priority. France’s KAIROS retained first place in the Green500 rankings, whose top ten systems were unchanged.

In the Top500, here are the top ten leaders:

  1. LineShine, National Supercomputer Center. China. 2.198 exaflops
  2. El Capitan, Lawrence Livermore National Laboratory. United States. 1.809 exaflops
  3. Frontier, Oak Ridge National Laboratory, United States. 1.353 exaflops
  4. Aurora, Argonne National Laboratory, United States. 1.012 exaflops
  5. JUPITER Booster, Jülich Supercomputing Centre, Germany. 1.000 exaflops
  6. HPC7, Eni, Italy. 571.5 petaflops
  7. Eagle, Microsoft, United States. 561.2 petaflops
  8. HPC6, Eni, Italy. 477.9 petaflops
  9. Supercomputer Fugaku, RIKEN Center for Computational Science, Japan. 442.01 petaflops
  10. Alps, Swiss National Supercomputing Centre, Switzerland. 434.9 petaflops