On September 17, the 11th Huawei Connectivity Conference 2026 opened in Shanghai. Wang Tao, vice chairman and rotating chairman of Huawei, announced at the conference that Shengteng Super Nodes have been widely used. Currently, more than 1,000 units have been deployed, including more than 370 customers. This means that after experiencing engineering verification in various aspects such as chips, servers, connectivity, and software ecology, Huawei's next-generation AI computing power products have begun to enter the actual commercial deployment stage. At the same time, Huawei also announced the latest developments in the next generation of Ascend 960 chips and supernodes. Wang Tao said that the performance of the Shengteng 960 chip has doubled, and the R&D progress has exceeded expectations, three quarters ahead of schedule, and will be ready in the first quarter of 2027; the Atlas 960 will be the first supernode in the industry to use an NPO. Among them, the liquid-cooled supernode will be launched in the third quarter of 2027, and the air-cooled supernode will be launched in the second quarter of next year. “We will stick to the pace of evolution from generation to year and rely on Tao's law to continue to double the computing power of the national model.”

On September 17, the 11th Huawei Connectivity Conference 2026 opened in Shanghai. Wang Tao, vice chairman and rotating chairman of Huawei, announced at the conference that Shengteng Super Nodes have been widely used. Currently, more than 1,000 units have been deployed, including more than 370 customers. This means that after experiencing engineering verification in various aspects such as chips, servers, connectivity, and software ecology, Huawei's next-generation AI computing power products have begun to enter the actual commercial deployment stage. At the same time, Huawei also announced the latest developments in the next generation of Ascend 960 chips and supernodes. Wang Tao said that the performance of the Shengteng 960 chip has doubled, and the R&D progress has exceeded expectations, three quarters ahead of schedule, and will be ready in the first quarter of 2027; the Atlas 960 will be the first supernode in the industry to use an NPO. Among them, the liquid-cooled supernode will be launched in the third quarter of 2027, and the air-cooled supernode will be launched in the second quarter of next year. “We will stick to the pace of evolution from generation to year and rely on Tao's law to continue to double the computing power of the national model.”