China has narrowed the AI development gap with the United States to just three months in some areas, because firms such as DeepSeek have worked out how to use chips and apply algorithms more efficiently, the CEO of Chinese startup 01.AI Lee Kai-fu said.

Lee, a prominent figure in the global artificial intelligence space and a former head of Google China, told Reuters that startup DeepSeek revealed that China had pulled ahead in areas such as infrastructure software engineering

DeepSeek shook the global AI industry when it launched an AI reasoning model in January that it said was trained with less advanced chips and was cheaper to develop than its Western rivals. The announcement challenged the assumption that US sanctions were holding back China’s AI sector.

“Previously I think it was a six to nine month gap and behind in everything. And now I think that’s probably three months behind in some of the core technologies, but actually ahead in some specific areas,” Lee said in an interview in Hong Kong.

Washington’s semiconductor sanctions were a “double edged sword” that created short term challenges but also forced Chinese firms to innovate under constraints, he added, pointing out how Chinese companies had developed their algorithms.

“The fact that DeepSeek are able to figure out the chain of thought with a new way to do reinforcement learning is either catching up with the US, learning quickly, or maybe even more innovative now,” Lee said, referring to how DeepSeek models show users their reasoning process before delivering answers - a capability first developed by OpenAI but not released to users.

China’s tech sector jumped into the global race to develop generative AI soon after the launch of OpenAI’s ChatGPT in late 2022 but until DeepSeek’s launch, many of the country’s tech leaders said they were far behind Western counterparts.

Lee, who also runs his own venture capital firm, founded 01.AI in March 2023, joining other new AI startups such as ZhipuAI and Moonshot as well as Chinese tech giants Baidu, Alibaba and ByteDance in building foundational models.

Lee said investing in proprietary models had become “courageous” for AI startups in a market environment dominated by well-funded tech giants and rapidly evolving open-source alternatives.

He said 01.AI will instead focus on practical AI applications - software solutions that help clients better deploy foundational models, he said.

Earlier this month, 01.AI launched Wanzhi, a new software platform that helps enterprises deploy AI technology. The company has already begun generating revenue and projects growth for 2025 to several times the $15 million achieved last year, Lee added.