China Deploys Sugon 8000, Its First Domestic 100,000-Card AI Supercluster
China has officially deployed "Sugon 8000" (Dengfeng), its first fully domestically produced 100,000-card AI supercluster, at the Zhengzhou core node of the National Supercomputing Internet. The system is designed to handle hybrid workloads combining high-precision scientific computing and low-precision AI computing. This deployment marks a significant milestone in China's push for hardware self-sufficiency, establishing a massive domestic computing foundation for AI training and scientific research. By integrating scientific and intelligent computing, it prepares the nation's infrastructure for next-generation hybrid computing demands. The supercluster supports full-precision computing from FP64 to INT8 and utilizes advanced immersion phase-change liquid cooling to support megawatt-level power density per cabinet. It is currently supporting over 300 computing tasks across 26 fields, including new materials and drug discovery.
## BACKGROUND
China's National Supercomputing Internet is a national initiative aimed at connecting high-performance computing centers across the country into a unified network to optimize resource sharing. Modern supercomputing increasingly relies on "super-intelligent integration," which fuses traditional high-precision scientific simulations (like physics and chemistry) with low-precision AI workloads to accelerate research.