Huawei has announced an ambitious plan to roll out four new iterations of its Ascend AI chip over the next three years, signaling its intent to compete directly with U.S. chip leader Nvidia. Breaking years of secrecy, the Chinese technology giant revealed the development roadmap during the annual Huawei Connect conference in Shanghai.
Huawei introduced its latest processor, the Ascend 910C, earlier this year and confirmed that two variants of its successor, the Ascend 950, will debut in 2026. Huawei plans to follow up with the Ascend 960 in 2027 and the Ascend 970 in 2028. Vice Chairman Eric Xu emphasized that computing power is critical for advancing artificial intelligence, especially for China’s rapidly expanding AI sector.
A key breakthrough highlighted by Xu is Huawei’s proprietary high-bandwidth memory technology, which removes reliance on South Korean and U.S. suppliers. This innovation addresses a long-standing bottleneck in China’s semiconductor industry and strengthens the country’s domestic chipmaking capabilities.
Alongside the new chips, Huawei unveiled next-generation AI supernodes — Atlas 950 and Atlas 960 — capable of supporting 8,192 and 15,488 Ascend chips respectively. These systems build on the success of the Atlas 900 (CloudMatrix 384), which integrates 384 Ascend 910C chips.
According to research firm SemiAnalysis, Huawei’s latest architecture can outperform Nvidia’s GB200 NVL72 on some performance metrics.
By expanding its Ascend series and introducing powerful supernode architecture, Huawei is positioning itself as a serious global competitor in AI computing, reinforcing China’s drive for technological self-reliance while challenging Nvidia’s dominance in the high-performance AI chip market, Reuters news report said.