Filed under supercomputing.

China's exascale bet against GPUs — and what LineShine's interconnect tells us about post-NVLink fragmentation

LineShine, which debuted atop the June 2026 TOP500, is the first system to exceed two exaflops of sustained double-precision performance without GPU accelerators — built under US export controls from indigenous Arm processors, on-package HBM, and a custom fat-tree fabric called LingQi. The significance is not the ranking: China has closed the interconnect gap that sovereign exascale programmes could not paper over, and the scale-up fabric market is now genuinely fragmented.