China presented the physical system of its first fully domestically developed artificial intelligence supercluster, the Sugon 8000 AI supercluster (also known as Dawning 8000 or Dengfeng), capable of interconnecting more than 100,000 AI accelerator cards. The hardware debuted at the 2026 National Annual Conference on High Performance Computing in Shenzhen.
What happened?
According to CCTV and Global Times reports, Sugon 8000 integrates supercomputing and intelligent computing in a unified architecture. It can run both high-precision scientific calculations and large language model training and inference. The system occupies about 2,000 square meters and uses liquid cooling and high-speed interconnects.
During the event, the company presented 10 advanced research applications, including digital twin brain simulation and universe evolution. More than 300 key applications have been optimized for the system across more than 30 fields. Source: Global Times.
Why is this important?
The project marks the advance of Chinese AI infrastructure from the 10,000 to 100,000 card level, reducing dependence on foreign chips amid export restrictions. The fusion of HPC and AI enables a complete cycle: scientific simulation generates data, AI filters solutions and supercomputing validates, reducing hallucinations. The system is connected to the National Supercomputing Internet.
What changes in practice?
Chinese research institutions, universities and companies gain access to massive national computing capacity for materials, drugs, meteorology, aerospace and AI. The extreme density in 2,000 m² demonstrates advanced engineering in interconnection and cooling.
Image credit: VCG / Global Times — Source: Global Times
By GeekikiBot