Key Points
- The chip is specifically for inferencing rather than training but Alibaba will still use semiconductors from other vendors like Nvidia.
- Alibaba said it would invest at least 380 billion Chinese yuan ($53.1 billion) in AI over the next three years.
- Chinese technology firms have focused on developing homegrown chips over the last few years amid rising geopolitical tensions between the U.S. and China.
Alibaba is developing a new chip designed for artificial intelligence applications, two people familiar with the matter told CNBC, as Chinese tech firms look to ramp up development of homegrown semiconductors. Alibaba’s semiconductor is designed for inferencing, which is the actual running of AI applications, rather than training which is when huge amounts of data is used to develop a large model, said one source, who wished to remain anonymous due to the sensitive nature of the matter. The Wall Street Journal first reported the news of the chip, which has not officially been launched. Alibaba declined to comment when contacted by CNBC. Chinese technology firms have focused on developing homegrown chips over the last few years as rising geopolitical tensions between the U.S. and China has threatened to cut off access to American technology for firms in the world’s second-largest economy. The latest example is U.S. firm Nvidia , which earlier this year was blocked from selling its H20 product to China. The H20 is a system that has reduced performance and was designed specifically to comply with export curbs to China. Nvidia has since been given the green light to resume shipments to China as long as the firm pays 15% of that revenue to the U.S. government . Still, Nvidia said on its earnings call this week that it has not shipped any of the H20 chips to China yet. Alibaba will continue to use its own chips as well as those from other vendors like Nvidia, one of the sources told CNBC. Nvidia’s chips are capable of both training and inferencing. Nvidia shares extended their recent losses, falling over 3% in early trade in the U.S. Meanwhile, Alibaba shares were up 12% on the chip news as well as its strong earnings posted on Friday. Unlike Nvidia, Alibaba will not sell its chip to external customers. Instead, customers will rent computing power from Alibaba that is partly based on these new chips. Alibaba did not respond to a request for comment from CNBC. AI boost The Hanghzou-headquartered company has a semiconductor design unit known as T-head. The last inference chip it released was the Hanguang 800 in 2019 . These chips go into servers in data centers to process AI workloads. Alibaba has signaled plans to invest heavily in AI infrastructure. This year, the company said it would invest at least 380 billion Chinese yuan ($53.1 billion) next three years. Already, Alibaba is seeing the benefits. On Friday, the company said its cloud division saw revenue jump 26% year-on-year , as quarterly growth continues to accelerate. AI-related product revenue “maintained triple-digit year-over-year growth for the eighth consecutive quarter,” the company said. Some of Alibaba’s rivals including Baidu are also developing their own semiconductors. While these companies still use Nvidia chips where they can, they are also using alternatives from Chinese firms like Huawei , which has emerged as a key player in China’s AI semiconductor space. Nvidia CEO Jensen Huang has been lobbying the U.S. government to allow American chip firms to sell to China or risk Chinese firms filling the void . Cambricon, one of the country’s AI chip designers, this week reported a 4,000% surge in revenue and posted a record profit in the first half of the year.
