China is back at the top of the supercomputer leaderboard.
LineShine, a supercomputer at the National Supercomputing Center in Shenzhen, debuted at No. 1 on the June 2026 TOP500 list, pushing the US system El Capitan into second place.
According to The Verge, the result puts China back on top of the ranking for the first time in years.
LineShine takes No. 1
LineShine reached 2.198 exaflops on the High Performance Linpack (HPL), the benchmark used for the main TOP500 ranking. That puts it ahead of El Capitan, which is listed at 1.809 exaflops.
The system is notable because it reached No. 1 without GPUs. According to the Associated Press, LineShine runs only on traditional CPUs, while many modern supercomputers use GPU accelerators for AI and other high-throughput work. The system also uses about 42.2 megawatts of power.
Tom’s Hardware reported that LineShine uses 304-core LX2 processors, the LingKun platform, LingQi interconnects, and Kylin OS. That gives the ranking geopolitical and technical weight, especially as the US and China continue to compete over advanced computing hardware.
The rest of the top five stayed familiar. El Capitan ranks second, followed by Frontier and Aurora in the US and JUPITER in Germany. Exascale supercomputing is now spread across the US, Europe, and China, even though the US still holds several of the highest-ranked systems.
AI benchmark gap
LineShine’s No. 1 ranking does not mean it is the strongest system for every workload.
The machine also ranked first on HPCG, a benchmark often considered a better signal of memory-heavy scientific computing than HPL alone. That makes LineShine a serious high-performance computing system, not just a headline grabber.
AI is where the split shows up. On HPL-MxP, which tests mixed-precision performance, LineShine ranked fourth. Tom’s Hardware reported that LineShine scored 7.92 exaflops on that benchmark, while GPU-accelerated systems did better on the lower-precision operations common in AI training and inference.
That distinction matters for IT leaders watching supercomputer news through an AI lens. HPL measures sustained double-precision performance. HPL-MxP is closer to the math used in many AI workloads. A system can lead the main list and still trail in the benchmark more closely tied to AI infrastructure.
TechRepublic has covered that divide from the other direction, including Nvidia’s Vera Rubin and OpenAI’s work on a custom AI inference chip. Those systems are not direct comparisons to LineShine, but they help explain why accelerator design matters so much in AI.
LineShine’s debut still matters. It shows that China can return to the top of the public supercomputing rankings despite export controls and intense competition from US systems. It also adds another data point to the broader story of Chinese technology vendors challenging Western assumptions on cost, capability, and enterprise risk, a pattern already visible in Chinese AI models.
For CIOs, infrastructure teams, and AI planners, the lesson is to read the benchmark before buying the bragging rights. LineShine changes the TOP500 leaderboard, but it does not settle who leads in AI compute.
Also read: RIKEN ROQUO supercomputer uses a hybrid quantum-HPC setup to test what practical quantum workloads need.
