Google DeepMind CEO Demis Hassabis has now challenged the prevailing views in the US about China’s artificial intelligence capabilities. Hassabis believes that Chinese AI models are now only ‘a matter of months’ behind Western systems. Speaking at CNBC’s new podcast The Tech Download, Hassabis said China’s progress has been underestimated, “Maybe they’re only a matter of months behind at this point,” he noted, adding that just one or two years ago, many believed China was far further back. The comments made by the Google DeepMind CEO run counter to the widespread assumptions in Washington and Silicon Valley that China lags significantly behind in AI development.
Evidence of China’s progress in AI race
Chinese AI lab, DeepSeek shocked the markets last year by introducing a model which performed strongly despite using less-advanced chips. On the other hand tech giants like Alibaba and startups like Moonshot AI and Zhipu have since released competitive models. Also, Nvidia CEO Jensen Huang has also credited China’s progress, saying the US is ‘not far ahead’ of China in the AI race.
Catching up vs innovating
Hassabis has acknowledged China’s ability to catch up but also questioned whether its firms can deliver frontier breakthroughs. “They’ve shown they can catch up … and be very close to the frontier. But can they actually innovate something new, like a new transformer … that gets beyond the frontier? I don’t think that’s been shown yet.”The transformer architecture introduced by Google researchers in 2016, underlings today’s large language models which include OpenAI’s ChatGPT and Google’s Gemini. Hassabis also emphiased on the fact that China may face both cultural and technical challenges. “To invent something is about 100 times harder than it is to copy it. That’s the next frontier really, and I haven’t seen evidence of that yet.” He compared DeepMind’s approach to a “modern-day Bell Labs,” emphasizing exploratory innovation rather than scaling existing methods.
Challenges ahead for China
China’s AI industry faces hurdles:* U.S. export bans restrict access to Nvidia’s most advanced chips.* The White House recently approved sales of Nvidia’s H200 chip to China, but it is not top-of-the-line.* Domestic chipmakers like Huawei are trying to fill the gap but still lag behind.Analysts warn that over time, superior U.S. infrastructure could widen the gap again. Richard Clode of Janus Henderson told CNBC that current Chinese AI strength may represent “peak relative capability” versus the US.
About the Author
