At the 2026 National High-Performance Computing Academic Conference in Shenzhen, China's first fully-domestic 100,000-GPU AI supercluster made its real-machine debut, with ten major application achievements released, Sina Tech reported. The cluster's computing power can support creating a digital-twin brain and even simulating the evolution of a universe spanning tens of billions of light-years.
The debut marks a new level for China's computing-infrastructure buildout. A fully domestic stack — from chips and interconnects to system software — addresses both performance and supply-chain security, providing a stable foundation for training ever-larger models and running massive inference at scale.
The supercluster highlights the strategic weight of compute as the bedrock of the AI industry. As model training shifts toward the "million-card era," domestic clusters reduce dependence on foreign hardware and shorten iteration cycles for Chinese frontier models, reinforcing the country's position in the global AI competition.
The achievement arrives alongside a wave of domestic model releases and underscores how infrastructure, models and applications are advancing in lockstep to power the next phase of intelligent industrialization.
Source: Sina Tech / 2026 HPC Conference
