A supercomputer is taking shape, aiming to support science and AI workloads while offering a different approach in the global race for next-generation computing.
China’s National Supercomputing Center in Shenzhen has introduced “LineShine,” a new domestically developed supercomputing platform that is expected to reach 2 exaflops of computing power when fully deployed, marking one of China’s most ambitious efforts to expand its homegrown high-performance computing infrastructure.
The system is being positioned as a fully indigenous supercomputing stack, built on Chinese-designed processors, storage, and networking technologies. According to Lu Yutong, director of the Shenzhen center and LineShine’s chief architect, the goal is to establish a complete domestic ecosystem spanning chip design through system integration.
LineShine is being rolled out in phases. The initial deployment includes 100 Huawei Kunpeng servers, delivering 12,800 CPU cores. A larger second phase is planned, with tens of thousands of additional processors, expanded interconnect capacity, and higher-density storage aimed at compute-intensive workloads.
Technical documents associated with the project describe LineShine as a distributed ARM-based computing system designed for high memory bandwidth and fast internal networking. A research preprint linked to the platform outlines a full-scale exascale configuration of 20,480 compute nodes, each equipped with dual ARMv9-based LX2 processors.
The architecture prioritizes high-speed communication between nodes and strong memory throughput, allowing it to support both traditional scientific simulations and newer AI training workloads. That dual-purpose design reflects a broader shift in supercomputing, where research institutions increasingly want a single platform for modeling, engineering workloads, and machine learning.
One of LineShine’s distinguishing features is its CPU-only design. While many of the world’s fastest supercomputers—especially those deployed in U.S. national laboratories—depend heavily on GPU acceleration, LineShine is being developed as an all-CPU alternative, offering a different architectural path in the race toward exascale computing.
Chinese officials have not disclosed when the full system will be completed or when it is expected to achieve its projected peak performance.
