China’s National Supercomputing Center (NSCC) in Shenzhen has introduced “LineShine,” a next-generation supercomputing system that it claims will reach 2 exaflops of performance once fully deployed. The system was presented during an institutional meeting focused on domestic computing power and industrial applications last week.

The announcement positions LineShine as a fully homegrown system built using Chinese-designed chips, storage, and networking technologies. According to Lu Yutong, NSCC director in Shenzhen and chief designer of the system, the project represents an effort to build an end-to-end domestic supercomputing stack.

LineShine is being developed in phases. According to an official announcement, the first phase reportedly uses 100 Huawei Kunpeng servers, totaling 12,800 CPU cores. The second phase expands the system significantly, incorporating tens of thousands of CPUs, large-scale interconnect infrastructure, and extensive storage systems designed for high-throughput computing workloads.

The center says the system is intended to support scientific computing, engineering simulation, and artificial intelligence workloads at scale, covering applications from materials research to climate modeling and life sciences.

Full-stack compute push

The system architecture described in technical documents linked to the project outlines a large distributed computing design built around ARM-based processors, high-bandwidth memory, and high-speed interconnects. The preprint referenced by researchers states:

“The LineShine supercomputer, developed by the National Supercomputing Center in Shenzhen (NSCC-SZ), is an exascale system consisting of 20,480 computing nodes, denoted the full machine. Each node is equipped with two ARMv9-based LX2 processors.”

The same description outlines dense interconnectivity and high memory bandwidth designed to support large-scale AI training and simulation workloads.

Officials involved in the program describe the system as part of a broader push toward computing self-reliance. During the meeting last week, Shenzhen Science and Technology Innovation Bureau official Li Xiaoli said the project demonstrates “complete self-reliance and controllability across the entire stack.”

The center also emphasized that the system is designed to integrate scientific computing, engineering modeling, and AI training into a unified platform, enabling multi-domain workloads on a shared infrastructure.

AI and simulation focus

LineShine is being positioned not just as a traditional supercomputer but as a hybrid platform for AI and scientific workloads. Applications highlighted by the center include molecular simulations, fluid dynamics, materials design, and large-scale AI model training.

The project also reflects a shift in global supercomputing strategies, where CPU-centric architectures are being explored alongside GPU-dominated systems used in leading Western installations.

China’s announcement comes at a time when global supercomputing rankings are dominated by GPU-accelerated systems, including those used at US national laboratories. LineShine, however, is being described as an all-CPU approach aimed at large-scale domestic control of compute infrastructure.

The project also reflects China’s broader strategy to reduce dependence on foreign high-performance computing technologies amid ongoing restrictions on advanced chip access. By focusing on domestically designed CPUs and system components, LineShine is positioned as part of a long-term infrastructure push rather than a single-generation system.