Policy

China AI Exports Could Weather Full US Decoupling, Citi Says

Analysts estimate even an extreme scenario would put less than 10% of Chinese exports at risk, though open-weight model restrictions pose uncertainty.

Omega Editorial· August 14, 2026· 3 min read

China's AI export vulnerability less severe than feared

China could manage a complete artificial intelligence decoupling from the United States without catastrophic economic damage, according to new analysis from Citi Research. The Hong Kong-based team estimates that even an extreme-case scenario would place roughly 9.2 percent of China's total exports at risk—a figure they describe as manageable given Beijing's track record navigating previous trade shocks.

The assessment comes as bilateral tech tensions escalate ahead of President Xi Jinping's planned September state visit to the US. Both nations have imposed reciprocal measures since last month, with Washington targeting Chinese AI models, advanced robotics, and optical transceivers while expanding forced labor blacklists. Beijing has responded with its own restrictions on American firms.

Direct AI trade already suppressed

Led by analyst Yu Xiangrong, the Citi team argues that Washington's latest restrictions are "unlikely to materially disrupt China's AI exports" because direct bilateral trade in the sector has already contracted significantly. Between January and June, Chinese AI-related exports to the US fell 4.1 percent year-over-year, subtracting half a percentage point from overall export growth.

The analysts note that direct US-China AI trade has largely decoupled at the aggregate level, which limits the potential first-order impact of additional restrictions. For context, the US accounted for 14.7 percent of China's total exports at the end of 2024, yet Beijing weathered President Donald Trump's "Liberation Day" tariffs during that period.

Most recently, Trump signed a proclamation Thursday imposing tariffs up to 100 percent on certain imported drones and components, citing national security and cybersecurity concerns.

Why it matters

While the headline finding suggests resilience, Citi's analysis highlights a critical uncertainty: restrictions on "open-weight" AI models represent what the analysts call a "genuine wild card" that could fundamentally reshape the global technology landscape. Unlike traditional export controls on chips or hardware, limiting access to open-weight models—AI systems whose parameters are publicly available—could fragment the development ecosystem in ways that extend far beyond bilateral trade flows. This distinction matters for multinational companies planning AI strategies and policymakers weighing the second-order effects of technology restrictions.

Indirect benefits persist

Despite the bilateral contraction, China remains an indirect beneficiary of the broader global AI boom, according to the Citi analysis. This suggests Chinese companies continue to participate in AI supply chains and capture value through third-country trade relationships, even as direct US market access narrows.

The findings were detailed in a Friday research note first reported by the South China Morning Post.

#us-china relations#ai trade#export controls#technology decoupling#open-weight models#citi research

This is an original analysis by the Omega editorial team. Source reporting: AI Watch.

Want systems like this working for your business?

Book a Call

More in Policy

Policy· 3 min read

Colorado AI Rules Detail Disclosure, Review Requirements

Proposed regulations clarify how businesses must handle automated decisions and chatbot interactions under two state laws taking effect in 2027.

Via Automation Watch · Aug 14, 2026
Policy· 3 min read

Federal AI Adoption Fails When Workforce Readiness Lags Technology

Half of federal leaders doubt their staffing can meet mission goals, yet commercial change management models ignore government's unique constraints.

Via AI Watch · Aug 14, 2026
Policy· 2 min read

Open-Source AI Models Face Looming Federal Oversight

Trump administration signals frontier open models may require government review as capabilities advance, despite current exemptions.

Via AI Watch · Aug 14, 2026