China has taken a major leap in artificial intelligence infrastructure with the official launch of the Dawning 8000 (Dengfeng) supercluster in Zhengzhou, Henan province, on Friday, the country's first AI supercluster capable of supporting over 100,000 domestically developed computing cards.

The milestone, announced at the 2026 Henan Provincial Artificial Intelligence Conference, marks the transition of China's AI infrastructure from the homegrown 10,000-card era to the 100,000-card deployment stage, according to Chinese tech company Dawning Information Industry, also known as Sugon which builds and operates the AI supercluster.

The cluster has been simultaneously connected to the national supercomputing internet, a nationwide computing network which now aggregates over 3.5 million CPU cores and 250,000 GPU cards, serving more than 1.4 million registered users with over 7,300 applications, 1,500 adapted large language models and over 11.3 million visits per month on average.

As large language models, scientific AI, and intelligent agents rapidly evolve, computing infrastructure must scale up accordingly. The 100,000-card threshold is no longer just a number — it represents the new baseline for next-generation AI infrastructure, capable of handling complex workloads that demand high concurrency, high throughput, and mixed precision across multiple tasks, Sugon said.

Scaling from 10,000 to 100,000 cards presents challenges far beyond simply adding more processors. It tests system architecture, network interconnection, memory access efficiency, energy control, and ecosystem capabilities.

What sets this system apart is its end-to-end homegrown infrastructure, spanning chips, computing, storage, networking, cooling, applications and services.

The core node has already completed optimization for over 300 supercomputing-AI fusion applications across more than 20 fields, including large models, robotics, automotive, pharmaceuticals, new materials, quantum computing, and weather forecasting. More than 70 of these applications have achieved 10,000-card scale expansion, validating the system's stability and reliability under massive, high-intensity scientific workloads.

In addition, Sugon announced a strategic partnership with the Beijing Institute of Scientific Intelligence to develop a second domestically developed 100,000-card supercomputing-AI fusion system.

The node is now open to research institutions, universities, enterprises, and individual users, offering inclusive, efficient, and convenient computing services through the national supercomputing internet.