Nvidia’s share in the domestic market of the PRC fell to 34%, while Huawei’s share rose to 28%
Translated with AI assistance
China is rapidly reducing its artificial intelligence market’s dependence on foreign processors. In the first half of 2026, Nvidia’s share in the Chinese data center AI accelerator market fell to 34%, while Huawei closed in on its American rival with a share of 28%, according to IDC data reviewed by Reuters.
The gap is narrowing primarily due to the rapid growth of Chinese manufacturers. From January to June, Nvidia shipped 678,000 AI accelerators to China – 44% less than a year earlier. Huawei, by contrast, increased its shipments from over 300,000 to 572,000 devices. Another 7% of the market was captured by T-Head, a company owned by Alibaba, 5% by Cambricon, 4% by Baidu’s Kunlunxin, and 3% by Hygon. Thus, according to IDC, these five Chinese manufacturers alone already account for 47% of the market. Nine out of the 11 tracked Chinese developers increased their shipments over the year.
One of the reasons for this shift is U.S. export restrictions, which make Nvidia’s most powerful processors unavailable for official supply to China. At the same time, Beijing is encouraging the use of domestic solutions and, according to Reuters, requires Chinese companies to coordinate purchases of Nvidia chips approved for export. Nvidia CEO Jensen Huang previously estimated the Chinese AI chip market at approximately $50 billion and criticized U.S. restrictions for clearing the way for Huawei.
The Chinese market itself is also changing. In the first half of the year, just over 2 million AI accelerators were shipped in the country. Sales of graphics processing units (GPUs), which form the backbone of a significant portion of modern AI infrastructure, fell by 26% to 991,000 units. Shipments of other specialized AI processors, on the contrary, grew by nearly 80% to 1.016 million, surpassing GPUs for the first time. Huawei, T-Head, Kunlunxin, and Cambricon control 95% of this segment. IDC expects the overall Chinese AI accelerator market to grow by another 19% in 2026, despite a 9% decline in GPU shipments.
But replacing Nvidia with hardware alone is not enough. One of the American company’s main advantages is the software ecosystem built around its processors, which allows developers to efficiently train and run AI models. Now China is trying to close this gap as well.
For example, at the end of September, DeepSeek open-sourced six software tools at once, adapted for Huawei Ascend processors. They replicate the functionality of previously developed tools for Nvidia and, according to the company itself, are intended to help create an «independent and controllable» software ecosystem for Chinese AI chips.
It is precisely the software layer that could turn out to be the next stage of China’s import substitution strategy. Previously, even the emergence of a competitive Chinese processor did not mean that companies were ready to abandon Nvidia: doing so required porting and rewriting software. Expanding DeepSeek’s compatibility with Ascend is gradually lowering this barrier.
As a result, the Chinese market is no longer producing just one competitor to Nvidia, but an entire domestic technology chain: Huawei manufactures AI accelerators, Alibaba, Baidu, and other companies are developing their own chips, while DeepSeek creates software tools that enable their use with modern AI models. China still remains significantly dependent on foreign technologies – Nvidia continues to be the largest single supplier – but IDC data shows just how rapidly this dependence is shrinking.
The development of the Chinese AI ecosystem makes it possible to reduce other countries’ dependence on the Nvidia ecosystem. If Chinese processors and software tools begin to scale beyond the domestic market, data centers and developers from the bloc’s member countries will have a wider choice of hardware and infrastructure — especially where the cost or availability of American GPUs becomes a limiting factor.
