China's new supercomputer, LineShine, has dethroned American machines by achieving a record computing power of 2.198 exaFLOPS. Its particularity? It uses only central processing units (CPUs), unlike most supercomputers that combine CPUs and GPUs. This approach allowed China to bypass U.S. restrictions on the export of graphics chips.
Installed at the National Supercomputing Center in Shenzhen, LineShine has been operational since the first half of 2026. Its computing power exceeds that of the previous champion, El Capitan, located in California, by 22%. This is the first time since 2017 that a Chinese supercomputer has taken the top spot on the TOP500 list.
LineShine's design is based on a Chinese technological breakthrough, according to official statements. Unlike other systems that use GPUs to perform parallel calculations, LineShine relies solely on CPUs. This architecture was developed to circumvent U.S. semiconductor restrictions in place since 2018.
Supercomputers like LineShine are used in various fields: weather forecasting, drug discovery, and artificial intelligence. This makes it possible to solve problems that would otherwise be too time-consuming or costly. LineShine has already been employed for projects in atmospheric sciences.
The TOP500 ranking shows there is no single path to performance. The fastest machines use diverse architectures: CPUs, GPUs, custom accelerators. LineShine proves that a pure CPU architecture can compete with the best.
U.S. chip restrictions have pushed China to innovate. Start-ups like DeepSeek have also developed AI models with fewer GPUs. LineShine represents a historic step in Chinese technological sovereignty, according to the national center.
The top 10 of the TOP500 ranking, after this first Chinese place, includes machines from Italy, Switzerland, Japan, Germany, and the United States. Each manufacturer explores different paths, making the supercomputer landscape particularly dynamic. LineShine is a perfect example.