US-China AI trade friction 'manageable'—but open-weight model curbs a wild card

US-China tech tensions are escalating ahead of President Xi Jinping's planned September state visit to the US, with both nations imposing reciprocal measures. Washington has targeted Chinese AI models, advanced robotics, optical transceivers, and expanded its forced labour blacklist, while Beijing has retaliated against US firms. Despite this tit-for-tat escalation, Citi Research analysts argue that the direct bilateral AI trade impact is likely to remain contained because the US-China AI trade channel is already largely suppressed; Chinese AI-related exports to the US fell 4.1 per cent year on year between January and June, accounting for only a 0.5 percentage point drag on overall export growth.

The analysts' core argument rests on the distinction between direct bilateral trade disruption—which is limited—and indirect global effects. Even under an extreme full-decoupling scenario, up to 9.2 per cent of China's exports could be at risk, a scale Citi characterizes as manageable, citing China's demonstrated resilience to Trump's 2024 tariffs when the US represented 14.7 per cent of total exports at year-end 2024. However, analysts flag a significant caveat: restrictions on open-weight AI models (systems with publicly released code and weights) represent a "genuine wild card" whose impact on the global AI supply chain is harder to quantify. Such restrictions could reshape competition and access patterns in ways that go beyond direct trade volumes, potentially affecting how AI technology diffuses globally and how other countries source AI tools.