US Tightens AI Chip Exports to Chinese Firms, Closing Offshore Loophole
U.S. Commerce Department rules now require licenses for AI chip exports to companies headquartered in China, tightening controls on AI chip exports and offshore transfers.
The U.S. Commerce Department on Sunday moved to close a loophole that allowed advanced AI chips to reach Chinese-controlled firms through overseas subsidiaries, announcing new export guidance that extends licensing requirements to companies with headquarters or parent companies in China. The rule change targets shipments of the world’s most advanced processors — including Nvidia’s Rubin and Blackwell series and AMD’s MI350X — and explicitly requires export licenses even when a recipient is based outside the People’s Republic. The guidance signals a continued hard line on AI chip exports to Chinese entities following a 2025 enforcement pause.
New export guidance targets offshore subsidiaries
The updated guidance makes clear that firms with a principal place of business or a parent company in China now fall under U.S. export controls, regardless of where their subsidiaries operate. The Bureau of Industry and Security at the Commerce Department said the change closes a route that companies allegedly used to send high-performance processors to Chinese-controlled data centers in other countries. Companies shipping GPUs and AI accelerators will need to seek licenses before exporting to entities that ultimately report to or are controlled by Chinese headquarters.
The policy does not require the shutdown of existing data centers nor force the immediate removal of already-delivered hardware, according to the department’s announcement. Operators will still be able to run and maintain servers that were lawfully imported prior to the new guidance, but future shipments of specified high-end chips will face stricter review and potential denial.
Loophole traced to 2025 enforcement pause
Officials and industry sources say the loophole emerged after an enforcement pause in 2025, when measures intended to restrict global access to cutting-edge AI chips were temporarily not applied. The pause permitted shipments that, while routed to foreign subsidiaries, effectively supplied Chinese-controlled operations. Former U.S. State Department official Chris McGuire described the scale of the problem, saying, “Chinese companies have purchased these chips, very likely on a large scale.”
Analysts estimate that hundreds of thousands of high-performance processors may have been diverted to data centers affiliated with Chinese firms in countries such as Malaysia and elsewhere in Southeast Asia. That influx, they warn, could have accelerated China’s access to compute capacity critical for large-scale AI model training.
Implications for Nvidia, AMD and global supply chains
Nvidia’s top-tier GPUs and AMD’s MI350X are central to the training and inference capabilities of modern AI systems, making them focal points of the new rules. Nvidia has made significant revenue from Chinese markets in prior years and has been working to reestablish business ties after bilateral tensions curtailed earlier sales. Company leadership has warned publicly about the impact of restricted access to those markets, with Chief Executive Jensen Huang estimating that Nvidia missed roughly $50 billion in potential revenue in 2026 due to trade frictions.
Supply-chain managers at cloud providers and chip distributors are now assessing how the guidance will affect contractual commitments and inventory flows. Some firms face the prospect of applying for licenses that could take weeks or months to process, while others may seek alternative suppliers or redirect shipments to markets not covered by the restrictions.
Industry and security reactions are mixed
Technology firms and trade groups have expressed concern about the administrative burden and potential revenue losses stemming from the expanded licensing regime. Corporate compliance teams will need to update screening and export-control mechanisms to identify entities whose parent companies are based in China. At the same time, national security officials and some lawmakers have welcomed the move, arguing it prevents sophisticated compute resources from being used against U.S. interests.
Researchers who track hardware flows warned that the long-term effectiveness of export controls depends on enforcement and international cooperation, including intelligence-sharing with partner countries to detect re-exports. The Commerce Department emphasized that the policy aims to strike a balance between protecting sensitive technologies and minimizing disruption to legitimate commercial activity.
Practical steps companies must take now
Exporters of AI chips and related servers should immediately review customer ownership structures and implement enhanced due diligence on subsidiaries and affiliates. Legal and compliance advisors recommend conducting audits of sales from 2025 onward to identify any transactions that may now require retroactive licensing or further documentation. Firms with ambiguous ownership chains should plan to file license applications or seek formal determinations from the Bureau of Industry and Security.
Data-center operators that host third-party hardware are advised to confirm the chain of custody for all installed processors and to consult counsel on any contractual obligations that might be affected by the new guidance. Vendors planning future shipments of Rubin, Blackwell or MI350X-class devices should factor potential licensing delays into deployment timelines.
Geopolitical context and likely next steps
The guidance arrives amid broader tensions between Washington and Beijing over technology leadership and national security. While U.S. political leaders have at times signaled interest in stabilizing relations with Chinese counterparts, export-control policies remain a key lever for slowing the transfer of critical capabilities. The Commerce Department’s action demonstrates an appetite to refine controls in response to observed evasion tactics.
Moving forward, enforcement agencies may tighten monitoring of re-exports and collaborate with foreign customs authorities to track shipments routed through third countries. Companies and governments alike will watch closely for further clarifications or appeals processes that could shape how the new rules are implemented in practice.
The expanded export guidance is likely to reshape commercial patterns for high-performance chips, prompting exporters and cloud operators to reassess risk, compliance, and market strategies while regulators sharpen tools intended to prevent the offshore transfer of sensitive AI compute capacity.
