Advertisement
Semiconductors
TechTech War

Cambricon, China’s ‘little Nvidia’, sees ‘unprecedented’ scope for home-grown AI chips

AI chips are experiencing unprecedented opportunities, and the firm will continue to focus on innovation, co-founder Chen Tianshi says

Reading Time:2 minutes
Why you can trust SCMP
Cambricon has become the priciest stock in China’s onshore market after gaining 544 per cent over the past 12 months. Photo: CFOTO/Future Publishing via Getty Images
Ann Caoin Shanghai

Cambricon Technologies, the Chinese semiconductor designer seen as a potential alternative to US rival Nvidia, expressed confidence in its revenue growth prospects after a record first half on the back of strong demand for its artificial intelligence chips.

“AI chips, as the core of computing infrastructure, are experiencing unprecedented opportunities,” Chen Tianshi, chairman and co-founder, said at an online event on Thursday, adding that the company would continue to focus on technological innovation in design to enhance its core competitiveness.

The company, known as “little Nvidia”, said it would generate “sustained revenue” in the future due to robust demand for AI computing power, particularly for training large AI models. Its products have been deployed at scale in “multiple key industries”, including mobile carriers, as well as clients in the internet and finance sectors, it said.

Advertisement

Cambricon has become the priciest stock in China’s onshore market after gaining 544 per cent over the past 12 months. Its Shanghai-listed shares fell 5.1 per cent to 1,349.24 yuan on Friday.

Chen Tianshi, chairman and co-founder of Cambricon Technology, pictured in 2021. Photo: Handout
Chen Tianshi, chairman and co-founder of Cambricon Technology, pictured in 2021. Photo: Handout

Investors are confident in the outlook of the Beijing-based company, betting that its AI processors would be deployed in the country’s data centres to replace Nvidia. Cambricon’s stock surge came after Chinese start-up DeepSeek said its latest V3.1 model was trained using a new data format, which was “suitable for soon-to-be-released home-grown chips”.

Advertisement
Advertisement
Select Voice
Choose your listening speed
Get through articles 2x faster
1.25x
250 WPM
Slow
Average
Fast
1.25x