The first nationally produced 100,000-card AI super cluster is launched! Dialogue with Li Bin, Senior Vice President of Sugon

Intelligent things Author | Yang Jingli Editor | Li Shuiqing Intelligent things reported on July 10 that today, at the Photosynthetic Organization 2026 Intelligent Computing Application Conference, China's first nationally produced 100,000-card AI super cluster Dawn 8000 (Dengfeng) was completed and was simultaneously connected to the national supercomputing Internet. Domestic AI infrastructure has begun to move from the 10,000-card level to the 100,000-card level deployment stage. At present, Sugon 8000 has completed more than 300 super-intelligent fusion application optimizations, and more than 70 applications have achieved Wanka scale expansion. During the conference, Sugon also reached a strategic cooperation with the Beijing Institute of Scientific Intelligence and launched the development and construction of the second nationally produced 100,000-card super-intelligent integrated computing system. The 100,000-card computing power center began to move toward large-scale replication. In addition, Photosynthetic Organization also launched the "Open Computing Token Genealogy" plan to promote the conversion of Tokens into productivity. After the meeting, Li Bin, senior vice president of Sugon, mentioned in an interview with Zhidongxi and other media that the next set of 100,000 card system will be more focused and enhance its support for integration scenarios such as AI4S (AI for Science). With the accelerated development of large models, AI4S and intelligent agent applications, the test of 100,000-card-level AI super clusters has extended from computing power scale to system efficiency, reliability and application support capabilities. 1. The 100,000-card super cluster was launched, and Sugon 8000 completed the full-link localization verification. Sugon 8000 (Dengfeng) adopts the "super-intelligent fusion" technology route to meet the composite needs of high-precision scientific computing and low-precision intelligent computing. It realizes native integrated integration of all types of computing, supports full precision from FP64 to INT8, and can cover scientific research and industrial scenarios such as scientific computing, large model training, AI inference, and industrial simulation. The development and deployment of Sugon 8000 has gone through multiple stages: system development will be completed in 2024, engineering construction will be completed in 2025, 30,000 cards will be put into trial operation in February 2026, 60,000 cards will be put into use in April 2026, and 100,000 cards will be deployed in June 2026. In terms of system construction, Sugon 8000 has a full link of "chip, computing, storage, network, cooling, application, and service" A