Huawei has once again launched a high-performance smartphone chip after six years, aiming to achieve China’s leading level in consumer and AI chips. However, a research institution has released a report stating that Huawei is unlikely to surpass NVIDIA.
On September 7, Huawei unveiled the Mate XT 2 foldable smartphone in Guangzhou, featuring the latest Kirin 9050 Pro chip, the first high-performance chip to adopt logic folding technology. This marks Huawei’s debut of a brand-new Kirin chip at a flagship event following the global launch of the Mate 40 six years ago.
In addition, Huawei has officially announced the HarmonyOS 7 operating system. HarmonyOS 7 integrates a large-scale model with a data volume of 2 billion into the Ark Engine responsible for system-level performance scheduling.
Huawei’s main revenue comes from consumer electronics such as telecommunications equipment and smartphones, while it is also accelerating the development of its own AI chips and computing infrastructure in an attempt to fill the supply gap in the Chinese market caused by the US restrictions on NVIDIA’s high-end AI chip exports.
It is worth mentioning that 2030 is a critical time point for AI development set by the Chinese government. The State Council of the Communist Party of China issued the “Development Plan for the New Generation of Artificial Intelligence” in 2017, proclaiming that by 2030, China’s overall AI theory, technology, and applications are to reach a “world-leading level”, including achieving international competitiveness in the AI industry.
However, on September 4, the AI research institution Epoch AI released a special report comparing Huawei and NVIDIA chips. The core conclusion regarding whether Huawei can catch up with NVIDIA before 2030 in terms of technology and other aspects is “almost certainly not”.
The report examines Huawei’s chip development roadmap and supply chain, including the Ascend 950, 3D stacking layout, and the capacity of China’s domestically produced high-bandwidth memory (HBM). The research team estimates that by 2026, Huawei’s AI computing power production will be less than 4% of NVIDIA’s. NVIDIA’s upcoming B300 chip is expected to have nearly 7 times the theoretical performance of Huawei’s Ascend 950 chip, making it difficult to bridge the technological gap.
AI chip market data shows that as of the fourth quarter of 2025, in terms of AI computing power sold or shipped, NVIDIA holds 68% of the market share, ranking first; Google accounts for 19%, while Amazon and AMD each hold 5%, and Huawei and Cambricon from China have around 3% and 1% respectively.
The report also points out that market share does not necessarily reflect the decrease in AI computing power sold or shipped by a company. The total AI computing power of all AI chip designers tracked by the institution has grown by approximately 3.3 times annually since 2022, indicating that even with a decrease in market share, the supply volume may still be considerable and possibly continue to increase.
As for whether Huawei can narrow the gap in the future, the report specifically considers the inclusion of China’s domestically produced HBM supply. Due to US export controls severing overseas high-performance HBM supply, if Huawei can only rely on domestically produced HBM, its AI computing power production by 2028 may only be equivalent to around 1% of NVIDIA’s.