China's domestically developed LineShine supercomputer is charting a new path for next-generation computing, combining independent innovation with world-class performance to advance scientific-technological research and artificial intelligence applications.
A breakthrough is the culmination of years of research by Chinese scientists.
At the International Supercomputing Conference (ISC) High Performance 2026 in Germany in June, LineShine topped the latest TOP500 ranking with a sustained performance of 2.198 EFLOPS, becoming the world's first supercomputer to sustain more than 2 EFLOPS on the High Performance Linpack benchmark.
After LineShine received the award, LineShine chief designer Lu Yutong said the recognition marked a milestone for China's supercomputing industry.
"It's an exciting moment because we have fulfilled our mission. China's supercomputing has returned to the world stage. In fact, China's supercomputing has never stopped moving forward," said Lu, who also serves as director of the National Supercomputing Center located in south China's tech hub of Shenzhen City.
For years, most of the world's leading supercomputers have relied on a heterogeneous CPU-GPU architecture.
However, because key technologies for high-end GPUs have long been dominated by overseas suppliers, the LineShine team pursued a different technological route.
Instead of depending on GPUs for AI acceleration, researchers embedded AI matrix acceleration units directly into domestically designed CPUs, creating an all-CPU architecture that reduces data-transfer bottlenecks and makes it easier to develop and migrate scientific applications.
"We've introduced the matrix acceleration units used in graphics processing units (GPUs) into central processing units (CPUs). As a result, researchers can accelerate computing efficiently under the same programming model. It is one of the most user-friendly architectures because it allows more applications to be quickly transplanted into the system and deliver practical benefits," Lu said.
The new architecture also required breakthroughs across the entire technology chain. From processor design and memory packaging to networking and system software, every key component was developed domestically, laying the foundation for broader scientific and industrial applications.
The system has already been deployed in fields including atmospheric and ocean science, engineering simulation, materials science, drug discovery, brain science and AI.
Looking ahead, developers said the next step is to expand the software ecosystem so that the system can better serve both key national projects and a wider range of users.
"We will continue improving the LineShine software ecosystem step by step for both key sectors and broader application fields, and build a more efficient and faster supercomputing service environment," she said.
The TOP500 list, released twice a year, is widely regarded as one of the world's leading rankings of supercomputers.
Chinese systems have topped the list on multiple occasions, including Tianhe-1, Tianhe-2 and Sunway TaihuLight.
China's LineShine supercomputer charts new path for next-generation computing
