China’s artificial intelligence models may be just “months” behind U.S. and Western capabilities, Google DeepMind CEO Demis Hassabis told CNBC.
The director of one of the world’s leading AI research institutes and a key promoter behind Google’s Gemini assistant’s assessment goes against views that suggest China remains far behind.
On CNBC’s new podcast, “The Tech Download,” launched Friday, Hassabis said China’s AI models are closer to the capabilities of the United States and the West than “we probably thought a year or two ago.”
“They’re probably just a few months behind at this point,” Hassabis told The Tech Download.
About a year ago, Chinese AI research institute DeepSeek introduced a model that shocked the market because it was built on a less advanced chip and cost less than its U.S. alternative.
Since then, Deep Seek has released new models and the shock factor has faded, but the Chinese tech giant alibaba Also, startups such as Moonshot AI and Zhipu are releasing very capable models.
Still, Hassabis said that while China may catch up, the country’s companies have yet to prove their ability to create breakthroughs in AI.
DeepMind CEO Demis Hassabis listens to a discussion at the AI Summit held at Imperial College London on July 9, 2025 in central London.
Ludovic Marin | AFP | Getty Images
“The question is, can they innovate something new beyond the frontier? So I think they’ve shown they can catch up…and get very close to the frontier…but can they actually innovate something new beyond the frontier, like a new transformer? I don’t think that’s been shown yet,” Hassabis said.
This transformer is a scientific advance made by researchers at Google in 2017 and powers large-scale language models developed by AI Labs in recent years, including those that power products such as OpenAI’s ChatGPT and Google’s Gemini.
Other top tech figures have also praised China’s progress. Nvidia CEO Jensen Huang said last year that the U.S. is “not that far along” in the AI race.
“China is far ahead of us in energy. We are far ahead in chips. China is right there in infrastructure. China is right there in AI models,” Huang said.
Challenges of Chinese chips
Chinese technology companies face many challenges, with access to critical technology being the biggest one. The United States has implemented a ban on exports of cutting-edge semiconductors from the United States. Nvidia Required for training more advanced AI models.
The White House has indicated it will approve sales to China of Nvidia’s H200 chip, a more advanced semiconductor than what China has recently been able to obtain. However, this is not Nvidia’s top-of-the-line product.
Domestic chip companies like Huawei are trying to fill the gap, but their performance still lags behind Nvidia’s products.
Some analysts have suggested that in the long term, the unavailability of Nvidia chips in China could widen the gap between U.S. and Chinese AI models.
“My suspicion is that we’re going to start to see a divergence as the great AI infrastructure in the U.S. starts to iterate on these models and start to enhance the capabilities of those models over the next few years,” Richard Claude, a portfolio manager at Janus Henderson, told CNBC’s “The China Connection” last week.
“So we expect that from here we will probably see a peak in China’s relative AI capabilities versus the United States.”
Even Chinese companies are admitting difficulties.
According to a report in the South China Morning Post, Lin Junyang, head of technology for Alibaba’s Qwen team, said at an AI conference in Beijing last week that there is less than a 20% chance that Chinese companies will surpass the US tech giant in AI within the next three to five years. Lin reportedly said that the US’ computing infrastructure is “one to two orders of magnitude larger” than China’s.
However, Hassabis believes that the lack of breakthroughs on the frontier is due to “mentality” rather than technological limitations.
“Modern Bell Laboratories”
DeepMind’s CEO likened the company to a “modern-day Bell Labs” that encourages “exploratory innovation, not just extending what is currently known.” Bell Labs was founded in the early 1900s and was responsible for numerous Nobel Prize-winning discoveries.
“And of course, that’s already very difficult, because you already need world-class engineering to make it possible. And China definitely has that,” Hassabis said.
“The scientific innovation part is more difficult,” Hassabis added. “Inventing something is about 100 times harder than imitating it. . . . It’s really the next frontier, and we haven’t seen any evidence of that yet, but it’s very difficult.”
Hassabis is considered one of the leaders in the world of AI. The company he founded more than a decade ago, DeepMind, was acquired by Google in 2014. alphabet– Owned by Google, it has had recent successes with its AI products, including Gemini.
Google announced its latest model Gemini 3 in November, which has been well-received by users and the market as the tech giant seeks to allay concerns that it is lagging behind rivals such as OpenAI.
