Our Privacy Statement & Cookie Policy

By continuing to browse our site you agree to our use of cookies, revised Privacy Policy and Terms of Use. You can change your cookie settings through your browser.

I agree

Supercomputing plus AI: How China's first homegrown 100,000-card supercluster could simulate a universe

CGTN

China's first homegrown 100,000-card artificial intelligence (AI) supercluster displayed at the WAIC 2026 in Shanghai, China, July 17, 2026. /VCG
China's first homegrown 100,000-card artificial intelligence (AI) supercluster displayed at the WAIC 2026 in Shanghai, China, July 17, 2026. /VCG

China's first homegrown 100,000-card artificial intelligence (AI) supercluster displayed at the WAIC 2026 in Shanghai, China, July 17, 2026. /VCG

China showcased the physical system of its first homegrown 100,000-card artificial intelligence (AI) supercluster at the ongoing 2026 CCF National Annual Conference on High Performance Computing in Shenzhen, south China's Guangdong Province, demonstrating a new approach to computing that integrates supercomputing with AI.

00:43

A 100,000-card AI supercluster is a computing infrastructure that links 100,000 AI computing cards through high-speed interconnects. Completed earlier this year, the Chinese supercluster, known as Dawning 8000 (Dengfeng), has made a core breakthrough – not simply in its scale but also in its ability to bring together two approaches to computing – supercomputing and AI computing on a unified infrastructure, enabling it to handle both high-precision scientific simulations and large-scale AI model training and inference.

These two forms of computing reflect two fundamental ways humans understand the world and anticipate the future.

Supercomputing is used to understand the underlying laws of the physical world and use them to predict the future. The more complex the simulation is, the more computing power it requires.

It is good at simulating the physical world, delivering accurate results but taking more time to compute.

AI computing works by learning from past data to summarize patterns and rules quickly. However, as AI learns patterns from data without necessarily understanding the underlying causes, it can cause AI hallucinations, meaning false or misleading results.

Visitors observe China's homegrown 100,000-card artificial intelligence (AI) supercluster at the WAIC 2026 in Shanghai, China, July 19, 2026. /VCG
Visitors observe China's homegrown 100,000-card artificial intelligence (AI) supercluster at the WAIC 2026 in Shanghai, China, July 19, 2026. /VCG

Visitors observe China's homegrown 100,000-card artificial intelligence (AI) supercluster at the WAIC 2026 in Shanghai, China, July 19, 2026. /VCG

As China's Dawning 8000 combines supercomputing with AI computing, it creates a complete scientific computing loop.

First, supercomputing is used to conduct virtual physical experiments, generating large volumes of reliable simulation data. Then, AI models learn patterns from the data and rapidly screen vast numbers of candidate solutions to a smaller set. Finally, supercomputing verifies these candidates through high-precision simulations and eliminates hallucinations.

With such extraordinary computing power, the system could support the creation of a digital twin of the human brain inside a computer, and even simulate the evolution of a universe spanning tens of billions of light-years.

CHINA NEWS
Keep up to date with all the top trends, insights and developments from China.
subscribe
Search Trends