HomeIndustriesHuawei desires to capture market share in AI chips from Nvidia in...

Huawei desires to capture market share in AI chips from Nvidia in China

Huawei is attempting to capture a bigger share of China's artificial intelligence chip market, dominated by Nvidia, by helping local corporations tackle their rival chips for so-called “inference” tasks.

Leading AI corporations in China are counting on Nvidia graphics processing units (GPUs) to “train” large language models, with the $3.4 trillion US chipmaker's products seen as crucial to the technology's development.

Rather than difficult Nvidia in training, Huawei is positioning its latest Ascend AI processors because the hardware of selection for Chinese groups running “inference,” the computation performed by LLMs to generate a solution to a prompt.

The Chinese tech giant expects inference to be a bigger source of demand in the long run because the pace of model training slows and AI applications comparable to chatbots grow to be more widespread.

“Training is significant, nevertheless it only happens a number of times,” said Georgios Zacharopoulos, a senior AI researcher working on inference acceleration at Huawei’s Zurich lab. “Huawei mainly focuses on inference, which is able to ultimately serve more customers.”

Leading AI corporations in China depend on Nvidia graphics processing units to “train” large language models. © Costfoto/NurPhoto via Getty Images

The focus is on the less technically demanding but potentially lucrative route of retrofitting AI models trained on Nvidia products to run on Ascend chips, in response to company officials and Ascend customers. Since Nvidia GPUs and Ascend operate on different software, Huawei helps corporations use a unique software tool to make the 2 systems compatible.

Huawei's move is supported by the federal government from above. Chinese officials have urged local tech giants to purchase more AI chips from Huawei and move away from Nvidia.

An individual acquainted with Nvidia's operations in China said Huawei is viewed internally because the country's most serious competitor, adding that its chip design capabilities are “advanced.”

Washington has sought to curb Beijing's development of AI through export controls aimed toward hampering the event of sensitive technologies in China.

Unlike their US competitors comparable to OpenAI and Google, corporations in China shouldn’t have access to essentially the most advanced GPUs. But despite the fact that Chinese corporations can only purchase Nvidia's less powerful H20 chips, that are specifically tailored to export controls, the less powerful GPUs remain in high demand because they’re considered higher than local alternatives.

A gaming laptop with a super-fast Nvidia chip
Unlike their US competitors comparable to OpenAI and Google, Chinese corporations shouldn’t have access to essentially the most advanced GPUs © Glenn Chapman/AFP via Getty Images

Analysts and Huawei researchers said Ascend was not yet ready to exchange Nvidia for model training attributable to technical issues, comparable to a disruption in the way in which the chips inside a bigger “cluster” of AI chips grow to be larger and bigger during training Models interact with one another.

“While the Ascend chips perform well on a per-chip basis, there’s a bottleneck in connectivity between chips,” said Lin Qingyuan, Bernstein's semiconductor analyst for China. “When you train a big model, it’s good to break it down into smaller tasks. If one chip fails, the software has to seek out a way for the opposite chips to take over directly.”

The other challenge for Huawei is convincing developers to modify from Nvidia's Cuda software, which is generally known as the corporate's “secret sauce” since it is simple for developers to make use of and may significantly speed up data processing.

But Huawei's soon-to-be-released updated version of its AI chip, the Ascend 910C, also goals to deal with these concerns. “We expect this latest generation of hardware to come back with improved software that may make it more accessible to developers,” said a Huawei official who declined to be named.

Huawei and Nvidia face tough competition. The Chinese web company Baidu and the chip designer Cambricon have made progress in the event of AI chips. Meanwhile, Amazon and Microsoft within the US are also betting that as AI applications grow to be more widespread, they’ll capture more market share in chips for inference purposes.

Estimates from SemiAnalysis, a chip consulting firm, suggest Nvidia made $12 billion from sales in China last yr by shipping 1 million of its H20 chips to the country, selling twice as many AI chips like Huawei with its Ascend 910B.

“Nvidia’s China-specific H20 GPUs account for the vast majority of AI chips sold in China. But the lead is shrinking quickly as Huawei increases production capability,” said Dylan Patel, principal analyst at SemiAnalysis.

Industry insiders warned that Huawei's AI chip push was also limited by insufficient supply. Two potential customers told the Financial Times that they were unable to secure the chips.

Huawei didn’t reply to a request for comment. Nvidia declined to comment.

Analysts say Huawei's manufacturing is more likely to face challenges due to U.S. export controls which have left Chinese factories counting on aging chip-making equipment.

The Huawei flagship store in Shanghai, China
Chinese officials have urged local tech giants to purchase more AI chips from Huawei and move away from Nvidia © CPHOT/Sipa USA via Reuters

The give attention to inference also points to an evolving dynamic in Chinese AI that differs from that within the United States. Because of Washington's export controls, Chinese AI players are usually not in the identical race as Silicon Valley rivals Meta, Elon Musk's x.AI and OpenAI to construct large mega-clusters of Nvidia's most advanced GPUs.

“Chinese corporations play a unique game. “They put lots more emphasis on conclusions than the US since it is feasible to attain large efficiency gains even with less powerful chips, which also means they’ll reach commercialization more quickly,” said Bernstein analyst Lin.

Chinese corporations are betting that they’ll remain competitive in AI by reducing inference costs, which in turn makes running AI applications cheaper, he said.

Last month, Hangzhou and Beijing-based startup DeepSeek released its V3 model, which attracted attention for its low training and inference costs in comparison with comparable models within the United States.

The company proposed a brand new way for an AI model to selectively give attention to specific parts of the input data to scale back the fee of running the model. It also used the “Mixture of Experts” technique popular with other Chinese AI startups, which also helps speed up inference as only a part of the model is used to generate a solution.

DeepSeek said Huawei has successfully adapted V3 to Ascend and provided detailed instructions for developers on how one can use the chip. The FT previously reported that Huawei had sent engineers to assist customers migrate from Nvidia to Ascend.

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Must Read