Presented in Shenzhen, the phased system's first stage uses 100 Huawei Kunpeng servers with 12,800 CPU cores, while technical documents describe a full machine of 20,480 nodes.
NSCC says LineShine is a fully homegrown, all-CPU platform using Chinese chips, storage and networking for scientific computing, engineering simulation and large-scale AI training.
The project forms part of China's push for end-to-end computing self-reliance amid restrictions on advanced chip access, though the full deployment timeline has not been confirmed.
Is China's CPU-only supercomputer a clever bypass of US sanctions or a losing bet against GPU-powered AI?
After a massive military data leak, can China's new homegrown supercomputer prove its security?
In 2026, China unveiled the LineShine exascale supercomputer, built entirely with domestic ARMv9 LX2 CPUs to ensure technological sovereignty amid U.S. export restrictions. This all-CPU design simplifies integration and software control but demands advanced liquid cooling due to higher power needs. LineShine aims to exceed 2 exaflops and supports diverse applications like AI training and climate modeling, with full operation expected by 2029–2030. While the project showcases China's growing self-reliant tech ecosystem, including Huawei's AI infrastructure, it faces challenges in semiconductor manufacturing, software adaptation, and international skepticism due to limited independent performance verification. LineShine symbolizes the deepening U.S.-China tech decoupling and the resulting global ecosystem fragmentation.