Supercomputing and AI Infrastructure for Science in China: 2026 Update

James Ling, co-founder and vice director of the HPC Center at Shanghai Jiao Tong University, gives a ten-minute update on supercomputing and AI infrastructure for science in China. He highlights China's return to the TOP500 with a system exceeding two exaflops of double-precision performance, framing it as beneficial for China's visibility, for the credibility of the TOP500 list, and potentially for US funding responses. He explains China's roughly decade-long absence from the list: after open engagement, including hosting international HPC leaders at Tianhe-2A and the Wuxi TaihuLight system, list appearances were used as a reference for placing national centers on the US entity list, prompting withdrawal. He proposes a renewed relationship he calls compete and collaborate, noting that papers on the new system's applications are publicly available. He then describes his university center's shift toward model-as-a-service, operating both the top university supercomputer in China (about six petaflops double precision) and the top university AI cluster, serving 60,000 users. His team is designing its next-generation supercomputer for AI agents rather than human users, arguing that agents integrated with real-time system status will drive future workflows. He also presents the ASC student supercomputing challenge, now 14 years old, as a talent pipeline whose alumni include a young Gordon Bell winner and a core DeepSeek member, and a new international collaboration organization focused on ARM-based systems.