China could erode US AI dominance in 10-20 years, HKU scholar warns
Synopsis
A University of Hong Kong scholar warns that China's leads in energy infrastructure and open-source AI applications could erode US artificial intelligence dominance within 10 to 20 years — a timeline that reframes the chip war as only one front in a much broader strategic contest.
Key Takeaways
Li Cheng , founding director of the Centre on Contemporary China and the World (CCCW) at the University of Hong Kong , co-authored a paper on 18 June 2026 warning that China could dismantle US AI dominance within 10 to 20 years .
Using Nvidia CEO Jensen Huang 's 'five-layer cake' framework, researchers found the US leads in advanced chips and AI infrastructure, while China leads in energy and AI applications.
The gap in AI models between the two powers has narrowed to what the paper calls 'a manageable range,' according to the researchers.
China 's adoption of cost-efficient, open-source AI models is accelerating economy-wide AI integration and is identified as a potential long-term strategic advantage over US closed-source approaches.
The paper warns that 'the long-term sustainability [of the US AI moat] faces multiple structural challenges,' even as it acknowledges America 's near-term deterrent strength.
Li Cheng, founding director of the Centre on Contemporary China and the World (CCCW) at the University of Hong Kong, has warned that China could dismantle America's artificial intelligence leadership within the next 10 to 20 years, citing structural weaknesses in the US AI ecosystem. The assessment, published in a co-authored paper on Wednesday, 18 June 2026, identifies energy capacity and industrial AI applications as China's primary levers for closing the gap.
The 'Five-Layer Cake' Scorecard
Drawing on the 'five-layer cake' framework articulated by Nvidia CEO Jensen Huang, the researchers evaluated both superpowers across five core pillars: energy, semiconductor chips, AI infrastructure, AI models, and AI applications. The analysis found the US holding a decisive lead in advanced chips and AI infrastructure, while China had secured clear advantages in energy supply and the deployment of AI applications at scale.
On AI models — where the US still commands the world's most sophisticated systems — the researchers said the gap had narrowed to 'a manageable range,' signalling accelerating competitive pressure from Beijing.
Why It Matters: The Open-Source Wedge
China's strategic embrace of cost-efficient, open-source AI models emerged as a critical differentiator in the paper. By driving down adoption costs, Beijing has accelerated AI integration across its broader economy, the researchers argued. This approach stands in stark contrast to the premium, closed-source model strategy favoured by leading US firms such as those associated with frontier AI development.
The paper suggested this open-source pivot could offer China 'a new strategic reference for ultimately winning the long-term technological competition,' according to the co-authored analysis led by Li.
The US Moat: Strong Now, Fragile Later
The paper stopped short of declaring US AI supremacy immediately at risk. '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,' the authors wrote. The researchers attributed America's near-term resilience primarily to its unmatched semiconductor ecosystem, anchored by companies such as Nvidia.
Li, who previously served as a senior fellow at the Brookings Institution, co-authored the paper with two additional researchers. The work was published under the auspices of the CCCW, which focuses on strategic dimensions of China's global role.
What's Next: Energy and Applications as Battlegrounds
The findings direct attention toward two under-scrutinised arenas in the US-China tech rivalry: grid-scale energy infrastructure and sector-specific AI application deployment. China's National Energy Administration has aggressively expanded renewable capacity, which the paper frames as a structural enabler for powering next-generation AI data centres at lower cost.
As the AI race shifts from model benchmarks to real-world economic integration, the nations and companies best positioned to scale applications cheaply and reliably — not merely those with the most powerful chips — may ultimately define the competitive outcome over the next two decades.
Point of View
It surfaces China's underappreciated structural assets in energy and application-layer deployment. The open-source angle is particularly sharp — Beijing's tolerance for commoditising AI models mirrors how it captured solar and EV supply chains, using volume and cost compression to outflank premium incumbents. What mainstream analysis tends to miss is that AI economic value increasingly accrues at the application layer, not the model layer, which is precisely where China holds its clearest lead. The 10-to-20-year horizon is deliberately hedged, but it implicitly challenges the assumption that export controls on advanced chips are sufficient to preserve US AI leadership indefinitely.
NationPress
5 Aug 2026
Frequently Asked Questions
What did the University of Hong Kong paper say about US AI dominance?
The paper, led by Li Cheng of the University of Hong Kong 's CCCW , argues that China could erode US AI dominance within 10 to 20 years . While the US retains a decisive near-term lead in advanced chips and AI infrastructure, the researchers identified multiple structural challenges to its long-term position.
Where does China lead the US in artificial intelligence?
China holds clear leads over the US in energy infrastructure and AI applications, according to the paper. The researchers also noted that the gap in AI models has narrowed to what they describe as 'a manageable range.'
Why is China's open-source AI strategy significant?
China 's focus on cost-efficient, open-source AI models has accelerated AI adoption across its economy by lowering deployment costs. The paper argues this could give Beijing 'a new strategic reference for ultimately winning the long-term technological competition' against the US .
Who is Li Cheng and what is the CCCW?
Li Cheng is the founding director of the Centre on Contemporary China and the World (CCCW) at the University of Hong Kong and a former senior fellow at the Brookings Institution . The CCCW researches the strategic dimensions of China 's evolving global role.
How does the Nvidia five-layer cake framework apply to the US-China AI race?
Nvidia CEO Jensen Huang 's 'five-layer cake' framework breaks the AI sector into five pillars: energy, semiconductor chips, AI infrastructure, AI models, and AI applications. The researchers used this structure to score the US and China head-to-head, finding each country leads in different layers rather than one side dominating entirely.
