China’s First Open AI Computing Architecture Released

Shanghai Securities News China Securities Network (by reporter Xu Rui) On September 5, China’s first open AI computing architecture was released at the Chongqing Smart Expo. Dawning Information Industry Co., Ltd. (Sugon), in collaboration with over 20 industry chain partners, launched the AI Super Cluster System and announced the establishment of the "Joint Laboratory for Open AI Computing Architecture" under the National Advanced Computing Industry Innovation Center. This initiative aims to break through computational bottlenecks and promote the comprehensive openness and inclusive development of the domestic intelligent computing ecosystem.

Currently, innovations in large-scale models continue to drive a surge in demand for AI computing power. The newly released open AI computing architecture is a high-efficiency collaborative innovation system centered on GPUs, designed for large-scale intelligent computing scenarios. It connects upstream and downstream enterprises, including chips,整机, large models, and industry applications, to promote inclusive computing power through open cross-layer collaboration and optimization.

Based on this open architecture, Sugon’s AI Super Cluster System was simultaneously unveiled. The system features four highlights: "super high performance, super high efficiency, super high reliability, and comprehensive openness," providing a powerful computational foundation for scenarios such as trillion-parameter large model training and推理, as well as industry-specific fine-tuning.

The "Joint Laboratory for Open AI Computing Architecture" will focus on collaborative optimization between chips and underlying basic software, as well as model-hardware co-design, aiming to break down barriers such as the "hardware wall," "software wall," and "ecosystem wall." Sugon has initially opened key technical capabilities, including AI storage optimization, liquid cooling infrastructure design specifications, and the DeepAI basic software stack. This aims to reduce redundant R&D, lower the threshold for small and medium-sized enterprises, and jointly build an open and inclusive intelligent computing industry ecosystem in China.