Huawei seeks to seize a larger share of the Chinese market for artificial intelligence chips, which is dominated by Nvidia, by helping local companies adopt its competing silicon in so-called “inference” tasks.
China's leading AI companies are relying on graphics processing units (GPUs) made by Nvidia to “train” large language models, with the $3.4 trillion US chipmaker's products seen as vital to developing the technology.
Instead of challenging Nvidia in training, Huawei is positioning its latest Ascend AI processors as the machine of choice for Chinese groups that run “heuristics,” the calculations MBAs do to generate a response to a prompt.
The Chinese tech giant is betting that inference will be a greater source of demand in the future if the pace of model training slows and AI applications such as chatbots become more widespread.
“Training is important, but it only happens a few times,” said Georgios Zacharopoulos, a senior AI researcher working on accelerating inference at Huawei’s lab in Zurich. “Huawei is mostly focused on inference, which will eventually serve more customers.”
It's focusing on the less technically challenging but potentially profitable path of modifying AI models trained on Nvidia products to run on Ascend chips, according to company employees and Ascend customers. Because Nvidia and Ascend GPUs work Different programsHuawei is helping companies use another software tool to make the two systems compatible.
Huawei's efforts come with top-down government support. Chinese officials have urged local tech giants to buy more AI chips from Huawei and move away from Nvidia.
Huawei was viewed internally as the country's most serious competitor, one of the people familiar with Nvidia's operations in China said, adding that its chip design ability was “advanced.”
Washington has sought to limit Beijing's development of artificial intelligence through export controls aimed at hindering the development of sensitive technologies in China.
Unlike US rivals such as OpenAI and Google, the companies are unable to access the latest GPUs in China. But although Chinese groups are only able to source Nvidia's lower-end H20 chips designed to meet export controls, the less powerful GPUs remain in high demand because they are considered better than domestic alternatives.
Analysts and Huawei researchers He said Ascend was not yet ready to replace Nvidia for model training because of technical issues, such as a breakdown in the ways the chips interact with each other within a broader “collection” of AI chips when training ever-larger models.
“Although Ascend chips perform well on a per-chip basis, there is a communication bottleneck between the chips,” said Lin Qingyuan, a China semiconductor analyst at Bernstein. “When training a large model, you have to break it down into smaller tasks. If one chip goes down, the software needs to find a way for the other chips to take over without delay.
Another challenge facing Huawei is convincing developers to switch from Nvidia's Cuda software, known as the company's “secret sauce” because it is easy for developers to use and able to greatly speed up data processing.
But Huawei's soon-to-be-released updated version of its AI chip, the Ascend 910C, is also expected to address these concerns. “We expect this new generation of devices to come with improved software that will make it more accessible to developers,” said a Huawei employee, who declined to reveal his name.
Huawei and Nvidia face stiff competition. Chinese internet group Baidu and chip designer Cambricon have made great strides in developing artificial intelligence chips. Meanwhile, in the US, Amazon and Microsoft are also betting they can capture a larger share of the market in chips for inference as AI applications proliferate.
SemiAnalogy, a chip consulting firm, estimates that Nvidia generated $12 billion in sales in China last year by delivering 1 million H20 chips in the country, and sold twice as many AI chips as Huawei does with the Ascend 910B.
“Nvidia's H20 GPUs in China make up the majority of AI chips sold in China. But progress is shrinking rapidly as Huawei increases its manufacturing capacity,” said Dylan Patel, senior analyst at SemiAnalogy.
Industry insiders warned that Huawei's push for AI chips was also being constrained by insufficient supply, with two potential customers telling the Financial Times they were unable to secure chips.
Huawei did not respond to a request for comment. Nvidia declined to comment.
Analysts said Huawei's manufacturing likely faces challenges due to US export controls that have left Chinese factories reliant on outdated chip manufacturing equipment.
The focus on heuristics also points to an evolving dynamic in Chinese AI that differs from that of the United States. Washington's export controls mean Chinese AI players are not in the same race as Silicon Valley rivals Meta, Elon Musk's X.AI and OpenAI to build massive clusters of Nvidia's most advanced GPUs.
Chinese companies are playing a different game. “They are paying much more attention to inference from the US because significant efficiency gains can be achieved even with less powerful chips, which also means they can achieve commercialization faster,” said Lin, the Bernstein analyst.
He said that Chinese companies are betting on their ability to maintain their competitiveness in the field of artificial intelligence by reducing the cost of inference, which in turn makes running artificial intelligence applications cheaper.
Last month, the Hangzhou company and Beijing startup DeepSeek launched its V3 model, which has received attention for its lower training and inference costs compared to comparable models in the United States.
The company proposed a new way for an AI model to selectively focus on specific parts of input data as a way to reduce the costs of running the model. It also used the “expert mixture” technique popular in other Chinese AI systems Start-upswhich also helps speed up inference as only part of the model is used to generate the response.
DeepSeek said Huawei has successfully adapted the V3 to Ascend, providing detailed instructions to developers on how to use the chip. The F.T I mentioned Huawei sent engineers to help customers migrate from Nvidia to Ascend.
Additional reporting from Zijing Wu in Hong Kong
