U.S. Closes AI Chip Loopholes, Putting New Tech Squeeze on China
Washington has halted shipments of advanced Nvidia and AMD AI chips to Chinese-controlled subsidiaries outside mainland China, moving to shut a loophole that kept cutting-edge processors flowing. Chinese firms, global cloud providers, and investors are now staring at a tighter race to control the hardware behind the AI boom.
The battle over who controls the chips that power artificial intelligence is no longer confined to customs checkpoints in Shenzhen. It now reaches into corporate structures and overseas subsidiaries, as Washington moves to cut off China’s access to the most advanced AI processors wherever they are shipped.
On 31 May, U.S. authorities halted shipments of advanced AI chips made by Nvidia and AMD to Chinese subsidiaries operating outside mainland China, according to intelligence reporting. The step targets a key workaround: Chinese-controlled entities registered in third countries that were able to order high-performance processors not technically destined for China, but still under Beijing’s influence. The new move effectively extends existing export restrictions by treating these offshore subsidiaries as part of the same risk universe as buyers in the PRC.
For engineers and researchers working at Chinese tech firms abroad, the effect will be felt in server racks that do not arrive and training runs that never start. High-end Nvidia and AMD chips are the backbone of modern deep learning systems; without them, large-scale model training becomes slow, expensive, or in some cases impossible. Cloud service providers that have quietly catered to Chinese corporate customers via overseas data centers now face the prospect of delayed hardware deployments, cancelled contracts, and closer scrutiny from regulators and shareholders.
Strategically, the decision signals Washington’s intent to turn what was already a tight choke on AI hardware into something closer to a global dragnet. Rather than relying solely on geographic boundaries—no high-end chips to China—it is now following corporate and ownership links. That approach treats Chinese subsidiaries in Singapore, Europe, or the Middle East as potential conduits for sensitive technology, and it warns Western suppliers and distributors that formal jurisdiction is no longer a shield.
For Beijing, the move deepens a technology squeeze that officials already portray as an attempt to contain China’s rise. It will add urgency to domestic chip development programs, incentives to acquire foreign technology through third parties, and pressure on allies and partners who host major data center hubs. Countries that have attracted Chinese investment into their digital infrastructure will now be asked, implicitly or explicitly, to help police which chips can be installed—and for whom.
Markets are watching three groups in particular: Nvidia and AMD, whose growth has been propelled by insatiable demand for AI hardware; Chinese cloud providers and internet giants, who have raced to build their own large models; and neutral data center operators that have courted both Western and Chinese tenants. Nvidia and AMD may be able to redirect some sales to other regions, but investors will scrutinize any hit to volumes or pricing power. Chinese tech stocks, already under pressure from regulatory and economic headwinds, must now price in slower access to best-in-class hardware.
Technologically, the decision is likely to accelerate two parallel races: China’s push to design and manufacture competitive AI chips using domestic fabs or friendly foundries, and U.S.-aligned efforts to develop monitoring mechanisms that can track where advanced processors end up once they leave the factory. Both are hard problems. Building high-performance GPUs requires not just designs but access to extreme ultraviolet lithography and highly specialized materials. Tracking them demands better end-use verification than current customs systems provide.
If the restrictions hold, firms in the broader Global South may find themselves caught in the crossfire. Chinese-backed labs and startups based in these markets could see hardware deliveries stalled or cancelled, while Western suppliers face pressure to vet customers more aggressively. For host governments, decisions about who can install what chips in their territory suddenly carry implications for relations with both Washington and Beijing.
Key Takeaways
- The U.S. has halted shipments of advanced Nvidia and AMD AI chips to Chinese subsidiaries outside mainland China.
- The move targets a key workaround that allowed Chinese-controlled entities abroad to access high-end processors.
- Chinese tech firms, cloud providers, and global data center operators face new uncertainty over hardware supply and compliance risk.
- Strategically, Washington is shifting from geography-based to ownership-based controls on sensitive AI hardware.
- The decision will likely accelerate China’s drive for domestic chip alternatives and raise the stakes for countries hosting Chinese tech operations.
Outlook & Way Forward
In the coming months, both industry and governments will test the boundaries of the new restrictions. Chipmakers will seek clarity on exactly which subsidiaries and ownership structures trigger a halt, while multinational customers try to reorganize procurement and data flows to remain on the right side of U.S. rules without abandoning Chinese business altogether. That will generate a wave of legal, corporate, and diplomatic maneuvering well beyond the semiconductor sector.
Longer term, the measure underscores that AI hardware is now treated as a strategic asset akin to advanced weapons systems. States that wish to stay neutral in U.S.-China competition may find that neutrality increasingly expensive, as they are asked to align export controls, monitoring regimes, and investment screening with one bloc or the other. The contest over who gets to train the next generation of powerful models is becoming a contest over who shapes the rules of the digital future itself.
Sources
- OSINT