Huawei has unveiled the Atlas 960E SuperPoD, which it describes as the industry’s first NPO-based SuperPoD, designed to support the training and inference of AI models with up to 10 trillion parameters. The system was launched by Huawei Deputy Chairman and Rotating Chairman David Wang at HUAWEI CONNECT 2026 in Shanghai, as the company outlined its strategy to strengthen AI infrastructure through computing, networking and system-level innovation.
Powered by Huawei’s UnifiedBus and Hi-ONE optical technology, the Atlas 960E SuperPoD can scale to 4,096 NPUs, delivering up to 8 EFLOPS at FP8 precision and 16 EFLOPS at FP4 precision. Huawei said its fully liquid-cooled architecture and integration of 5,500 Hi-ONE optical engines reduce power consumption by more than 550 kilowatts compared with conventional configurations, while achieving 99.8% system availability.
Huawei also introduced the TaiShan 950 SuperPoD for general-purpose computing and the OceanStor M900 context-memory storage cluster. Multiple Atlas 960E SuperPoDs can be connected into larger SuperClusters, with Huawei saying the architecture can interconnect up to 512,000 NPUs, or up to one million NPUs using a multi-rail topology. The company also highlighted its move toward open-source development, including the CANN platform, as part of efforts to expand the Ascend computing ecosystem.
Source : www. techafricanews.com
