China tops the list of fastest supercomputers with a CPU-only behemoth, ending US champion El Capitan's reign

China's LineShine supercomputer has claimed the top spot on the 67th TOP500 list with a double-precision performance of 2.198 exaflops, overtaking the AMD-powered El Capitan. Built entirely with domes... Read original →

China’s all-CPU ‘LineShine’ supercomputer topping TOP500 signals a strategic pivot toward full-stack HPC sovereignty under sanctions. Its Armv9-based domestic cores achieve 2.198 EFLOPS in FP64 but lag in mixed-precision AI workloads due to the absence of HBM and AI-optimized data types like BF16—revealing a deliberate trade-off: scientific computing over generative AI. This architecture sidesteps U.S. GPU export controls but exposes the limitations of CPU-only scaling for modern AI. For TSMC and AMD, China’s decoupling accelerates supply chain fragmentation, pushing SMIC to fast-track non-EUV 3nm-class HPC chips. Washington may respond by tightening AI accelerator restrictions, inadvertently fueling China’s parallel HPC ecosystem. Over the next 18 months, expect more FP64-optimized domestic systems, yet algorithmic innovation—not hardware—will bridge China’s AI training gap.