Chinese telecom equipment maker Huawei said it will launch its next-generation artificial intelligence chip in 2027, ahead of its earlier schedule, as the company accelerates its AI computing push in China after declaring that its goal is to become Nvidia.
Wang Tao, Huawei's rotating chairman, told the 2026 Huawei Connect conference in Shanghai that the Ascend 960 chip "has been readied ahead of schedule and delivers twice the performance" of existing products, according to Chinese outlets including Kechuangban Daily on the 17th. The Ascend 960, originally slated for the fourth quarter of next year, is now expected to arrive earlier. The 960DT for AI model training is due in the first quarter of next year and the 960PR for inference in the third quarter, while the Ascend 970 is scheduled for 2028 and the Ascend 980 for 2029.
Huawei is also developing technology that links large numbers of AI chips so they operate as a single, larger computing system. Wang called the approach UnifiedBus and said the company has developed 11 types of semiconductors based on it. Huawei is also focusing on its AI computing supernode system, the Atlas 950 SuperPoD. "More than 1,000 sets of Atlas SuperPoD have already been deployed, with about 370 customers," Wang said. "The Atlas 950 SuperPoD has already begun large-scale commercial deployment."
Guo Ping, chairman of Huawei's supervisory board, the company's top oversight body, earlier told new employees that "Huawei's goal is to become Nvidia." The idea is to make a range of large language models run on Huawei's platform, much as they do on Nvidia's CUDA.
Huawei has argued that Moore's Law, which drove performance gains by continually shrinking transistors, has hit physical limits, and has pointed instead to what it calls Tau's Law, focused on cutting the time signals take to travel inside a chip. Reuters noted that for the approach to work, the chips must be able to exchange information at sufficiently high speeds.
