On September 17, the 11th Huawei Connect 2026 kicked off in Shanghai. Wang Tao, Vice Chairman of Huawei, announced that Ascend supernodes have seen widespread adoption, with more than 1,000 units deployed across over 370 customers. This marks Huawei’s new generation of AI computing power products officially entering commercial deployment after successfully passing multi-stage engineering validations, encompassing chips, servers, interconnections, and software ecosystems.

At the same time, Huawei unveiled the latest progress on its next-generation Ascend 960 chip and supernodes. Wang Tao revealed that the Ascend 960 chip delivers double the performance, with R&D progress outpacing expectations by three-quarters. It is set to be ready in the first quarter of 2027. The Atlas 960 will become the industry’s first supernode to adopt NPO technology, with liquid-cooled supernodes launching in the third quarter of 2027 and air-cooled versions in the second quarter of the following year. Huawei plans to maintain an annual generational evolution pace, leveraging Tao’s Law to continuously double its computing power scale.