China is “months” behind US AI models, Google DeepMind CEO says

China’s artificial intelligence models may be “just a few months” behind US and Western capabilities, Google DeepMind CEO Demis Hassabis told CNBC.

The assessment by the head of one of the world’s leading AI laboratories and a key driver of Google’s Gemini assistant contradicts the perception that China is still lagging behind.

Speaking on CNBC’s new podcast, The Tech Download, which launched Friday, Hassabis said Chinese AI models are “closer than we thought a year or two ago” to US and Western capabilities.

“They’re probably months behind at this point,” Hassabis told The Tech Download.

About a year ago, the Chinese DeepSeek AI laboratory came out with a sample sent shock waves in the markets thanks to him built strong performance on less advanced chips and at a lower price than American equivalents.

Although DeepSeek has released new models since then, and the shock factor is goneChinese tech giants like it Alibaba startups like Moonshot AI and Zhipu there is also produced very capable models.

However, Hassabis said China could catch up, but the country’s companies have yet to prove their ability to make AI breakthroughs.

“The question is, can they introduce something new beyond the border? So I think they’ve shown that they can catch up … and be very close to the border … But can they introduce something new, like a new transformer … that goes beyond the border? I don’t think that’s been shown yet,” Hassabis said.

Transformer was a scientific breakthrough created by Google researchers in 2017 that underpins large language models developed by AI labs in recent years, including products that power OpenAI ChatGPT and Google Gemini.

Other high-tech figures have also credited China’s progress. Nvidia CEO Jensen Huang said last year that the US was “not ahead” in the AI ​​race.

“China is way ahead of us in energy. We’re ahead in chips. They’re in infrastructure. They’re right there in AI models,” Juan said.

China’s technology firms, among the largest, face a number of challenges in accessing critical technologies. A US export ban on advanced semiconductors is in effect Nvidia needed to train more advanced AI models.

The White House said it would Approve the sale of the Nvidia H200 chip to Chinamore advanced semiconductor than the country has recently achieved. However, this is not Nvidia’s best product.

Domestic chip companies like it Huawei tried to fill the gapbut their performance still lags behind Nvidia’s offering.

Some analysts suggest that in the long term, the lack of access to Nvidia chips in China could mean a gap between the US and China’s AI models.

“I suspect we’ll start to see a gap as the U.S. advanced AI infrastructure begins to replicate these models and enable these models over time in the future,” Richard Claude, portfolio manager at Janus Henderson, told CNBC’s “The China Connection” last week.

“So I would expect here that we are at the peak of AI capability relative to China’s US.”

Even Chinese companies have acknowledged their challenges.

Lin Junyang, CTO of Alibaba’s Qwen team, said during an AI conference in Beijing last week that the Chinese firm has less than a 20% chance of surpassing the US tech giant in the next three to five years when it comes to AI. This was reported by the South China Morning Post. According to Lin, the US computing infrastructure is “one to two times bigger” than China’s.

However, Hassabis attributes the lack of frontier breakthroughs to “mentality” rather than technical limitations.

DeepMind’s CEO compared the company to a “modern-day Bell Labs” that encourages “exploratory innovation” rather than “extending what is known today.” Founded in the early 1900s, Bell Labs was responsible for discoveries made by Nobel laureates.

“Obviously, it’s very difficult because you need world-class engineering to do it. And in China it is inevitable,” Hassabis said.

“The scientific innovation part is much more difficult,” Hassabis added. “Inventing something is about 100 times harder than copying it. … It’s really the next frontier, and I haven’t seen the evidence for it yet, but it’s very difficult.”

Hassabis is considered one of the leading figures in the world of AI. A major driving force was DeepMind, the company he founded more than 10 years ago, which was acquired by Google in 2014. Alphabet– Google has had recent success with its AI products, including Gemini.

Google in November introduced Gemini 3, its latest model, was well received by users and the market as the tech giant sought to allay fears that it was falling behind rivals like OpenAI.