
The system was presented at the Photonics Organization 2025 Artificial Intelligence Innovation Conference in Kunshan, Jiangsu province, marking the first full public display of a domestic AI cluster at this scale and positioning it as a strategic asset in China’s broader push for advanced computing capability.
The scaleX platform integrates more than ten thousand graphics processing units into a unified high-performance cluster capable of delivering in excess of five exaFLOPS (exa floating-point operations per second) of total computing power, a measure indicating the system’s ability to perform at least five quintillion calculations per second.
Sugon’s senior vice-president Li Bin described the infrastructure as suitable for complex scenarios such as training trillion-parameter foundation models, scientific AI applications and other data-intensive workloads, drawing on the company’s deep experience in building large computer systems.
The supercluster’s architecture includes world-first high-density ultra-nodes and a self-developed high-speed native RDMA network that supports ultra-low latency and high bandwidth, enabling efficient scaling and deep optimisation of storage, computing and data transmission.
Sugon also highlighted that portions of scaleX’s performance and technical design already exceed the development roadmaps of comparable overseas systems projected for release in 2027. This suggests China’s accelerating capabilities in high-performance AI infrastructure even as global competition intensifies.
The unveiling of scaleX comes amid a growing race for AI computing dominance, with other major Chinese technology groups such as Huawei advancing their own large-scale clusters and supernodes based on domestic AI processors.
These efforts reflect Beijing’s policy emphasis on technological self-reliance and the reduction of dependence on foreign suppliers such as Nvidia, which has traditionally led the market for GPU-based AI infrastructure.
China’s expanded investment in locally designed computing systems underscores the strategic importance of AI and high-performance computing in national innovation agendas and international technology competition.

















