Chinese firms ‘falling behind’ US peers in AI

Chip curbs on Beijing have set it back two years in key sector, Alibaba chairman says

China is trailing the United States by two years in the global race to lead artificial intelligence (AI) development, according to Alibaba Group Holding co-founder and chairman Joe Tsai, as companies continue to struggle with technology curbs imposed by Washington.

“China is somewhat behind, obviously,” Tsai said, citing how ChatGPT creator OpenAI had leapfrogged the rest of the tech industry in AI innovation, in a podcast interview published on Wednesday with Nicolai Tangen, CEO of Norges Bank Investment Management, the branch of Norway’s central bank that manages the world’s largest sovereign wealth fund.

Tsai said China’s tech firms were “possibly two years behind” the top AI companies in the US. Washington’s export restrictions that barred Chinese firms’ access to advanced semiconductors, such as graphics processing units from Nvidia, had “definitely affected” tech players, including Alibaba, he added.

“We’ve actually publicly communicated [that] it did affect our cloud business and our ability to offer high-end computing services to our customers,” he said. “So it is an issue in the short run, and probably the medium run.”

Alibaba last November cancelled the spin-off of the Cloud Intelligence Group, saying the plan “may not achieve the intended effect of shareholder value enhancement” because of the uncertainties over US export restrictions on advanced chips. Alibaba owns the Post.

The US last week updated the sweeping export controls it implemented in October, making it harder for China to access advanced AI processors, chipmaking equipment and even laptop computers built with those chips, according to a Reuters report. The revised rules took effect on Thursday.

Tsai’s assessment reflects the concerns of China’s broader tech industry on how export controls are dampening local AI innovation, making the country less competitive in the field.

Tsai said tech firms were continuing to look for ways to mitigate the impact of these restrictions, including sourcing advanced processors from other suppliers and stocking up on available chips in the market.

“I think in the next year or 18 months, the training on large language models can still go ahead, given the inventory that people have,” Tsai said, referring to the technology used to train ChatGPT and similar AI systems.