China's LineShine supercomputer has achieved 2.198 exaflops of sustained performance using CPUs exclusively, marking a significant milestone in high-performance computing. The system, built with Huawei technology, claims the title of world's fastest FP64 system and is the first CPU-only machine to sustain exascale double-precision performance.

The achievement demonstrates that CPU-based architectures can now reach exascale performance thresholds, a capability previously dominated by GPU-accelerated systems. This development challenges the prevailing industry assumption that sustained exascale computing requires GPU acceleration, the primary driver of recent semiconductor demand and data center capital expenditure in AI infrastructure buildout.

For the AI infrastructure sector, LineShine's performance validates CPU-centric compute strategies for specific workloads, potentially diversifying the hardware pathways for supercomputing deployment. The milestone may influence vendor strategy and investment priorities, particularly for workloads where double-precision floating-point performance and CPU-only architectures offer economic or computational advantages over GPU-heavy systems.