TheChek News Flow

The young chipmaker has software advantages over much larger rival Huawei as it aids Beijing’s race for tech self-sufficiency

In 2019, when Huawei stopped using Cambricon’s technology for its smartphones, the Chinese AI chip designer lost the bulk of its business, in what could have been a devastating blow.

But six years later, a pivot has made Cambricon Huawei’s primary challenger in China, with investors placing big bets that it will be one of the winners from Beijing’s campaign to achieve more technological self-sufficiency and cut its AI industry’s reliance on chips made by US tech giant Nvidia.

Cambricon’s share price has shot higher on speculation that it could become the leading supplier of chips to power models developed by DeepSeek, the Chinese AI champion whose technological breakthrough stunned Silicon Valley and markets this year.

After its shares doubled since the start of the month to Rmb1,495 ($209), lifting its market capitalisation to Rmb580bn, the chipmaker warned investors on Friday that its “stock price may be separating from the [business] fundamentals”.

Huawei still dominates China’s AI chip industry. But policymakers have thrown their weight behind Cambricon in an effort to create a competitive environment as they work to increase domestic chip output. The start-up, less than a decade old, makes AI chips for data centres and edge computing modules, which place computing resources closer to devices instead of faraway data centres.

Cambricon, which on Tuesday reported a Rmb1bn ($140mn) profit in the first six months, has aligned itself with Beijing’s AI development plans.

The company “must shoulder the mission and responsibility of our era” to advance AI technology, founder Chen Tianshi said during an investor call this year.

The start-up emerged from a team, including Chen and his brother, which spent years developing processors at China’s main research institute, the Chinese Academy of Sciences. With CAS backing, the fledgling group spun off in 2016 and soon released its first chip for AI tasks. The institute remains Cambricon’s second-largest shareholder after Chen, with a 15.7 per cent stake.

Huawei, the company’s first major customer, accounted for 98 per cent of revenues in 2018, licensing AI technology for its smartphones. When the conglomerate replaced Cambricon with its own technology, the chip designer shifted its focus to cloud computing accelerators, which process complex AI tasks. The move put the company directly in competition with its former client.

Cambricon suffered another significant setback in 2022, when Washington added it to the “entity list” for allegedly supporting Chinese military development. It had to sever ties with Taiwan Semiconductor Manufacturing Company, its leading fabrication partner, forcing it to turn to Chinese manufacturers.

Between 2020 and 2024, Cambricon invested Rmb5.6bn in research and development, which it ploughed into software improvements to make it easier to deploy models originally trained on Nvidia’s GPUs to run on its rival Siyuan chips.

Several people, including a ByteDance AI engineer, said Cambricon’s software compatibility made its products easier to use than Huawei’s Ascend.

“Cambricon struggled to gain traction until the end of 2024, when it collaborated with ByteDance to make its chip more compatible with algorithms trained on Nvidia’s ecosystem,” said Lin Qingyuan, semiconductor analyst at Bernstein.

Cambricon’s revenues are forecast to climb from Rmb6.5bn this year to Rmb13.8bn in 2026, according to Goldman Sachs estimates, with the company’s market share of China’s AI chip market expected to increase from 3 to 11 per cent by 2028.

Many of China’s leading purchasers of AI chips, including China Telecom, Alibaba, Tencent, and Baidu, also compete with Huawei in other businesses, such as network equipment, cloud computing, and self-driving cars, creating a strong incentive for them to support alternatives to the sprawling tech conglomerate.

Huawei declined to comment. ByteDance and Cambricon did not respond to requests for comment.

Cambricon’s growth is dependent on how quickly its manufacturing partner, Semiconductor Manufacturing International Corporation (SMIC), can expand production of 7nm chips, the advanced nodes used for AI processors.

Last year, Beijing instructed SMIC to set aside a larger portion of its capacity for Cambricon, rather than selling all the advanced nodes to Huawei, according to two people familiar with the matter.

The Financial Times reported that SMIC plans to double its 7nm capacity next year, which will give Cambricon and other Chinese semiconductor designers even more of its output.

Cambricon has struggled to meet demand after selling most of its chips in the first quarter to internet giant ByteDance, according to two people familiar with the matter, including one executive at the internet giant.

Bernstein’s Lin said it was a “matter of time” before Chinese fabs led by SMIC could increase manufacturing capacity for AI chips.

“There has never been a capacity issue that China has not been able to solve. Building things is what we do if there is money to be made,” said an investor in Chinese semiconductors.

Even so, analysts caution that Cambricon will remain a smaller player relative to Huawei.

Huawei has been working to fix software issues that have made its chips difficult to use. It is adjusting its chip architecture to provide developers with greater flexibility and is also designing a new processor to make training and deploying large language models easier, according to several people familiar with the effort.

Huawei is the only chipmaker that offers a viable alternative to Nvidia’s NV-Link, a technology that connects its chips, which could give it an edge over Cambricon in AI training chips, said analysts.

Cambricon’s immediate concern is securing more capacity from SMIC to manufacture its chips.

“Our customers have a very high rating for its products after testing,” said one salesperson at Inspur, one of China’s biggest data centre assemblers. “There just isn’t enough supply.”