China’s LineShine Claims TOP500 Crown, Ending U.S. Lead

On 22-23 June 2026 the TOP500 list announced at the International Supercomputing Conference in Hamburg placed China’s LineShine, at the National Supercomputing Center in Shenzhen, at the top with 2.198 exaflops. The U.S. system El Capitan at Lawrence Livermore fell to second with 1.809 exaflops, ending a decade of U.S. dominance and marking the first Chinese top-rank since 2017.

China reemerges as home to world's fastest supercomputerSouth China Morning PostChina's 'LineShine' declared world's fastest supercomputer at Hamburg conferenceChina Reclaims Leadership in Global Supercomputing

The TOP500, published biannually since 1993, gauges national supercomputing power. After U.S. GPU export controls, China paused submissions in 2023 and returned in 2026 with LineShine.

CNN

LineShine was built by the Shenzhen Cloud Computing Center on the LingKun platform, using ~13.79 million Arm-based LX2 CPUs and 42.2 MW power. It is operated by the National Supercomputing Center, Shenzhen. The displaced U.S. system, El Capitan, runs at Lawrence Livermore.

China Reclaims Leadership in Global Supercomputing

LineShine’s HPL score of 2.198 exaflops is 20 % faster than El Capitan’s 1.809 exaflops. It is the first CPU-only system above two exaflops, consuming 42.2 MW and delivering 52.07 GFLOPS/W.

China Seizes TOP500 Crown With All-Domestic Chips: LineShine Tops List, Trails US on AI

LineShine proves China can achieve exascale performance with wholly domestic processors, reinforcing its self-sufficiency strategy amid U.S. export bans. The machine will support climate, materials and nuclear simulations, but its CPU-only design lags behind GPU-based systems for AI model training.

China's LineShine Overtakes US to Reclaim World's Fastest Supercomputer Title

The National Supercomputing Center called LineShine “the result of breakthroughs across a series of core technological bottlenecks,” highlighting a fully indigenous stack. Premier Li Qiang called the win an “opportunity” for global access to advanced tech. Jack Dongarra noted the CPU-only exascale milestone while stressing the benchmark’s limited AI relevance. The U.S. DOE reaffirmed El Capitan’s nuclear-stockpile role.

South China Morning Post

Experts warn the TOP500 HPL test reflects traditional scientific workloads, not AI. Jimmy Goodrich said, “If the hyperscalers submitted their systems, this ‘world’s fastest’ would not crack the top five.” Addison Snell noted surprise at China’s public submission. Andrew Rohl highlighted missing private AI clusters like xAI’s Colossus.

CNN

LineShine ranks fourth on the AI-focused HPL-MxP benchmark, behind three U.S. GPU systems, showing a mixed-precision gap. Sources differ on its AI relevance, and TOP500 excludes many private AI clusters.

China Seizes TOP500 Crown With All-Domestic Chips: LineShine Tops List, Trails US on AI

• “If the hyperscalers submitted their systems, this 'world's fastest' would not crack the top five,” — Jimmy Goodrich, senior fellow, UC Institute for Global Conflict and Cooperation

Reuters

• “this is the first time a computer with only CPUs has reached exascale” — Jack Dongarra, co-founder, TOP500

China Seizes TOP500 Crown With All-Domestic Chips: LineShine Tops List, Trails US on AI

• “I’m not surprised it’s the number one system. What I’m surprised by is that they submitted it and want recognition for it,” — Addison Snell, CEO, Intersect360 Research

Reuters

• “Speaking at the TOP500 award ceremony in Hamburg, Germany, LineShine chief designer Lu Yutong said the machine broke through from the conventional hybrid architecture of using both CPUs and GPUs for supercomputers.” — Lu Yutong, chief designer, LineShine

CNN