Shenzhen's 14,000-Petaflop AI Powerhouse Redefines China's Tech Ambitions

Shenzhen has unveiled a cutting-edge intelligent computing cluster with 14,000 petaflops of processing power. The system, built entirely with domestic chips and technology, marks a major leap in China's AI infrastructure. It is designed to support thousands of businesses while maintaining exceptional reliability. The cluster's computing strength equals that of 5.5 million personal computers. Its training performance now rivals the world's leading systems, despite operating with a lower daily failure rate than most global counterparts. This achievement follows China's broader push to expand its computing capacity, with 42 similar 10,000-card clusters completed by the end of 2025.

The project relies on full-stack autonomous technology, including homegrown chips. Its development aligns with national efforts to create an integrated computing network across the country. Officials also aim to accelerate progress in large-language models and semiconductor innovation. While Shenzhen's 11,000-petaflop cluster ranks among the global TOP500, it remains behind the US's **El Capitan**, which leads with over 11,000 petaflops. By late 2025, the US held 171 of the world's top systems, compared to China's 40, maintaining an edge in peak performance.

The cluster will provide AI computing resources to hundreds of enterprises. Its success reinforces China's position in high-performance computing, though global rankings still highlight a gap with US capabilities. The system's efficiency and scale reflect ongoing advancements in domestic technology.