While the US holds the crown for cutting-edge chips, China’s advantages in energy and industrial applications could ultimately erode America’s artificial intelligence “moat”, according to a leading Chinese scholar.

Li Cheng, founding director of the Centre on Contemporary China and the World (CCCW) at the University of Hong Kong, said in a co-authored paper on Wednesday that Beijing could dismantle Washington’s AI dominance within the next 10 to 20 years.

“The US AI moat still has a strategic deterrent and delaying effect on competitors in the short term, but its long-term sustainability faces multiple structural challenges,” said the paper, which featured Li as the lead author along with two co-authors.

Borrowing the “five-layer cake” framework espoused by Nvidia CEO Jensen Huang, the researchers benchmarked the two superpowers across the core pillars of the AI sector: energy, semiconductor chips, AI infrastructure, AI models and AI applications.

They found that while the US maintained a decisive stranglehold in advanced chips and AI infrastructure, China had secured clear leads in energy and AI applications. Meanwhile, the gap in AI models – where the US still boasts the world’s most sophisticated systems – was narrowing to “a manageable range”, they said.

The paper highlighted China’s embrace of cost-efficient, open-source models as a major catalyst.

By driving down adoption costs, China had accelerated AI integration across its economy, the researchers said. This stood in stark contrast to the premium, closed-source models favoured by leading US firms, potentially offering China “a new strategic reference for ultimately winning the long-term technological competition”, they said.