Key Points
- China’s top AI researchers believe they can catch up to the U.S., but a lack of advanced chip-making tools is a major hurdle.
- The successful IPOs of AI startups MiniMax and Zhipu AI have boosted confidence in the sector.
- China has advantages in cheap electricity and infrastructure, but lags in computing power due to a massive investment gap.
- To overcome their limitations, Chinese researchers are focusing on the co-design of innovative software and hardware.
China’s leading artificial intelligence researchers are confident they can close the technological gap with the United States. Still, they acknowledge that a lack of advanced chip-making tools is a major problem. Speaking at a conference in Beijing on Saturday, they pointed to a new wave of risk-taking and innovation as a sign that China is on the right track.
The mood at the conference was optimistic, especially after the successful Hong Kong stock market debuts of two major Chinese AI startups, MiniMax and Zhipu AI. These “AI tigers” are part of Beijing’s push to build a self-sufficient tech industry that can compete with the U.S.
Yao Shunyu, the new chief AI scientist at Tencent and a former researcher at OpenAI, said there’s a “high likelihood” that a Chinese company could become the world’s leading AI firm in the next three to five years. He highlighted China’s advantages in cheap electricity and infrastructure. However, he also pointed to the “main bottleneck”: the lack of advanced lithography machines needed to produce cutting-edge chips.
While China has reportedly built a prototype of one of these machines, it’s not yet producing working chips and might not be ready until 2030.
Another challenge is the massive investment gap. Chinese researchers acknowledged that the U.S. has a substantial advantage in computing power because of the billions of dollars that have been invested in data centers. “The U.S. computer infrastructure is likely one to two orders of magnitude larger than ours,” said the technical lead for Alibaba’s AI model.
But this lack of resources has also forced Chinese researchers to be more innovative. They are focusing on approaches such as “algorithm-hardware co-design,” which enables them to run large AI models on smaller, cheaper hardware. They also pointed to a new generation of young entrepreneurs who are more willing to take big risks, a trait that has long been the hallmark of Silicon Valley.
