On Aug 7, China’s General Administration of Customs reported that exports grew 23.9 percent in July from a year earlier, beating forecasts. The figure that deserves Washington’s attention is narrower: Shipments to the United States rose about 17 percent, increasing from roughly 14 percent in June, while imports from the US climbed 15 percent.
After four years of tariffs, entity listings, and export controls designed to reduce American reliance on Chinese factories and slow China’s progress in artificial intelligence, bilateral trade is not contracting. It is accelerating in both directions.
The instinctive reading in Washington could be that enforcement has been too lax, and that the remedy is higher duties and tighter licensing. That interpretation would be incomplete, representing a categorical error rather than a minor miscalculation.
American policy was built to constrain a China that specialized in cheap assembly and depended on foreign technology at the high end. That country has been receding since China joined the World Trade Organization in 2001. The wall is real, but it guards the wrong terrain.
Average US duties on Chinese goods are generally lower today than a year ago, after a detente of repeated truce extensions. American importers had every incentive to switch suppliers but did not, which is a revealed preference. For a wide range of goods, no alternative matches Chinese cost, quality, scale, and delivery reliability at once.
Tariffs raise the price of a transaction; they do not conjure up a substitute producer. Where alternatives did emerge, such as Vietnam, Mexico, and India, these factories often assemble Chinese components. Trade was rerouted; the dependency was disguised, not dissolved.
Chip restrictions have followed a similar arc. Denied Nvidia’s most advanced chips, Chinese firms learned to do more with less. In July, Moonshot AI released Kimi K3, an open-weight model that reached the top tier of global systems, leading Arena’s front-end coding rankings at a fraction of what comparable American models cost.
Testing continues, and Moonshot briefly stopped taking new subscribers because it ran short of computing power. But the signal is clear: The lead US officials had put at six to 12 months shrank far faster than Washington expected. Restricting inputs raised the cost of Chinese progress. It also concentrated Chinese engineering talent on efficiency, precisely the capability that matters most when compute is scarce. The broader lesson is that the AI race will not be won by containment.
The deeper shift is in what China sells. In the first five months of 2026, Chinese exports of intermediate goods rose 25 percent and capital goods 12 percent from a year earlier, while consumer-goods exports rose just 4 percent. China is moving from supplying the world’s shoppers to supplying the world’s factories. This was years in the making: Local-content rules that pulled component production onshore; the “Little Giants” program funneling subsidies and cheap credit to thousands of specialized small manufacturers; and a domestic market whose brutal competition forced continuous upgrading.
Successful economies don’t protect old technology with walls: They use competition as motivation to build something new
Consider Dongguan ICT Technology Co Ltd, an 80-person firm exporting automated systems that assemble the circuit boards inside virtually all modern electronics, selling to American manufacturers, while sourcing almost entirely from Chinese suppliers. Its founder arrived in Dongguan in 2001 as an 18-year-old assembly-line worker. The transition from factory floors to automated machines shows why viewing China as an endless source of cheap labor is an expensive, outdated mistake.
The sharpest consequences fall not on the US but on its allies. Germany, Japan, and South Korea built their prosperity selling machinery, chemicals, and precision components into Chinese factories. Those customers are now competitors, in third markets and at home. For the first time in decades, Germany imports more advanced capital goods from China than it ships there, and European analysts warn Chinese competition could threaten a majority of European manufacturing output over the medium term. Volkswagen’s agreement with XPeng to codevelop two smart electric vehicles is the frankest admission available: The direction of engineering knowledge has reversed.
For the US, resilience cannot be bought with tariffs. Diversifying toward Vietnam and Mexico is sensible risk management, but it is insurance, not capability-building. Rebuilding capability means patient investment in machine tools, process engineering, automation, and above all, the technical workforce that runs them. Duties buy time; they do not buy competence. The only question that matters is what gets built domestically during the time the tariffs bought. So far, the answer is thin.
China’s position is less commanding than the export figures suggest. Growth leans on external demand while household consumption stays relatively weak and property remains a drag, an imbalance that invites precisely the protectionist wave now forming in Brussels. Deepening trade with the Global South and distributing AI capability through open weights extends China’s reach, but neither fixes demand at home. Strengthening the social safety net, raising the household share of national income, and reforming hukou (the Chinese mainland’s household-registration system) to unlock urban consumption would do more for the nation’s durability than another record surplus.
Both capitals are managing symptoms. Washington defends a moat around an industrial base it has not rebuilt. Beijing exports its way around a demand problem it has not solved. Neither is a strategy for the decade ahead. Successful economies don’t protect old technology with walls: They use competition as motivation to build something new.
The author is a distinguished professor at the UCLA Anderson School of Management.
The views do not necessarily reflect those of China Daily.
