The latest argument over AI supply chains is not coming from a new treaty or a formal Chinese government announcement. It is coming from a Chinese academic who says Washington’s strategy of building alliances around technology, chips, and critical minerals will fail if its purpose is to target China. South China Morning Post reports that Zheng Yongnian made that case at a conference, naming Pax Silica, a chip alliance, and a rare-earth alliance.
Zheng heads the public-policy school at CUHK Shenzhen. His comments are an assessment, not a binding policy statement from Beijing. They nonetheless capture a central dispute in the AI race: whether countries can build secure technology supply chains by organizing around exclusions, or whether that approach will deepen fragmentation and raise costs for everyone involved.
Zheng Yongnian Calls U.S. AI Supply-Chain Strategy Self-Defeating
SCMP says Zheng argued that U.S. frameworks such as Pax Silica would not succeed because they explicitly target a third party. In this case, the third party is China. He also said that attempts to restrict U.S. companies’ access to low-cost Chinese AI models would backfire by harming American commercial interests.
The argument rests on the interdependence of the technologies involved. AI systems depend on chips, servers, networking, power, data centers, software, minerals, and specialized labor. No single alliance can easily control every part of that chain without creating costs or bottlenecks for its own members. Zheng’s position is that cooperation on global economic development would be more sustainable than a system designed to contain a rival.
Washington’s position, as described by SCMP, is different. The United States has created or supported international groupings intended to secure supply chains in areas including AI, semiconductors, and critical minerals while limiting China’s ability to use strategic technology. From that perspective, alliances are a response to the concentration of supply and the security risks that come with it.
The disagreement does not mean that either side has a simple path to independence. EastFrontier’s report on Chinese AI firms using Southeast Asian Nvidia clouds showed that Chinese companies have continued looking for capacity through regional infrastructure even as restrictions make direct access more difficult. Supply chains adapt around controls, but that adaptation can add cost and complexity.
Pax Silica Has Become a Symbol of the AI Bloc Debate
Pax Silica is one of the initiatives Zheng singled out. SCMP says the grouping is part of a wider U.S. effort to organize secure supplies of technologies and materials relevant to AI and semiconductor production. The publication also reported that a State Department draft document warned that Pax Silica members could face exclusion if they joined China’s proposed World Artificial Intelligence Cooperation Organisation. That point should be read as SCMP’s description of earlier Reuters reporting.
The issue is not simply about the movement of physical chips. AI alliances can shape access to cloud infrastructure, model services, advanced manufacturing equipment, standards, data governance, and investment. A country deciding which framework to support may find that an AI policy choice has implications for trade, security relations, and its domestic technology sector.
That pressure was already visible in EastFrontier’s reporting that the United States was pressing countries to choose between AI blocs. Zheng’s comments offer a Chinese critique of that dynamic: he argues that a strategy built around forcing choices will not create a stable technology order.
The result is a contest of narratives as well as capabilities. The United States emphasizes trusted supply chains and security among partners. China emphasizes openness, development, and the potential costs of exclusion. Countries outside the two largest AI powers may welcome greater choice, but they may also be reluctant to make a decision that limits access to either market or technology ecosystem.
The Supply-Chain Contest Will Be Decided by Access and Adoption
Zheng’s prediction that the alliances are doomed to fail cannot be treated as a settled forecast. It is a view expressed at a conference, and the performance of any alliance will depend on its members, investment, technology, and the willingness of companies to use the systems it supports.
Still, the debate has practical consequences for Chinese AI companies. Export restrictions can affect the hardware they can buy. Alliance rules can shape where they build cloud capacity and which customers are comfortable adopting their models. At the same time, China’s expanding domestic AI ecosystem gives companies more reasons to develop alternatives in chips, software, data centers, and open-weight models.
EastFrontier’s August 23 report on Nvidia’s China chip denial illustrated the uncertainty that companies now face around advanced hardware. Even rumors about a China-specific product or changes in availability can affect planning because access to computing capacity is central to training and serving AI systems.
The contest will also be decided by adoption outside China and the United States. If companies and governments in Africa, Southeast Asia, Latin America, or the Middle East find Chinese models, clouds, and hardware commercially attractive, China’s ecosystem can gain influence even without full participation in U.S.-led arrangements. If allied supply chains deliver better security, availability, and performance, Washington’s strategy can gain credibility.
Zheng’s intervention is therefore best understood as a current Chinese argument about the direction of AI globalization. He is not announcing a new policy, and his conclusion is not inevitable. But his remarks underline how far the debate has moved. AI supply chains are no longer only a technical question of where chips are manufactured. They are becoming a test of whether the world’s technology systems will remain connected, split into blocs, or develop a more complicated overlap of both.
For companies building AI products, that uncertainty is now part of the operating environment. They must make choices about suppliers, models, customers, and locations before the geopolitical rules are fully settled. Zheng’s message is that exclusionary alliances will make that environment less stable. The response from Washington and its partners is likely to be that security requires precisely the kind of coordination he criticizes. The argument is far from over, but it is already shaping the infrastructure behind the next generation of AI.