The first fully domestically produced 100,000-card AI supercluster has been put into operation! Experts: Its peak computing power per second is equivalent to all of humanity continuously computing for 200 years.

According to CCTV News reports, recently, China's first fully domestically produced 100,000-card AI supercluster was officially put into operation, marking the country's computing infrastructure construction entering a new stage of 100,000-card-level deployment.

Bu Jingde, Chief Engineer of High-End Computing at the National Research Center for High-Performance Computer Engineering Technology, stated that the peak computing capacity of this new computing cluster per second is equivalent to continuous computing by all of humanity for 200 years.

Currently, the cluster has been applied in 26 fields including new materials, innovative drugs, and artificial intelligence, capable of supporting more than 300 types of computing tasks, providing large-scale computing power support for scientific research innovation and the AI industry.

It is understood that during the Guanghe Organization 2026 Intelligent Computing Application Conference, Sugon announced that China's first fully domestic 100,000-card AI supercluster — the Sugon 8000 (Dengfeng) — was officially completed and simultaneously connected to the National Supercomputing Internet.

The Sugon 8000 adopts a world-first high-density cabinet structure. The computing density of a single computing unit is 20 times higher than that of similar supernodes, achieving higher-scale computing deployment within limited space.

In terms of interconnection, the cluster uses the self-developed scaleFabric IB-native RDMA high-speed interconnect technology, achieving stable high-speed interconnection among 100,000 computing cards, providing support for large-scale AI training and high-performance computing.

Meanwhile, the Sugon 8000 is also equipped with ParaStor distributed storage, immersion phase-change liquid cooling, and other technologies, enabling the computing, networking, storage, and cooling systems to operate collaboratively, reducing energy consumption and cooling pressure on the ultra-large-scale cluster while improving overall performance.

With the 100,000-card-level fully domestic AI supercluster being put into operation, China's ultra-large-scale intelligent computing infrastructure has been further upgraded, which will also provide stronger support for high-computing-demand scenarios such as large model training, scientific computing, and research and development of new materials and new drugs.