Can Chinese firms still access advanced NVIDIA AI chips via overseas subsidiaries? The U.S. Department of Commerce has moved to restrict that path by issuing new guidance targeted at Chinese-headquartered corporations working outside China. The action goals a potential loophole that may have permitted advanced AI chips, including NVIDIA Blackwell processors, to attain Chinese corporations through subsidiaries in nations such as Malaysia.
The guidance, issued through the Commerce Department’s Bureau of Industry and Security, explains that export license needs apply to advanced chips when the obtaining entity is headquartered in China, despite that entity operates abroad. The move shows Washington’s persisted effort to limit China’s access to the semiconductor hardware require to train and deploy frontier AI systems.
The issue took attention after a paper circulated in Washington warning that “the floodgates have quietly opened.” As per Reuters, one industry source approximated that large number of chips may have reached Chinese subsidiaries during the period when the U.S. Left the opening in place.
The potential loophole appeared after the Commerce Department stated in May 2025 that it would not implement the AI Diffusion rule launched near the end of the Biden administration. That rule had set up licensing necessities for global access to advanced AI chips. By stepping back from enforcement, the U.S. May have created room for Chinese firms to gain restricted chips via foreign-based subsidiaries.
Data Center Ops
For AI infrastructure, the guidance matters because high-end GPUs and accelerators stay central to model training, inference, and large-scale data center operations. NVIDIA’s most advanced chips support the compute-heavy workloads behind foundation models, multimodal systems, reasoning models, and agentic AI applications. Limiting access to those chips can affect how fast companies and governments form competitive AI systems.
A Bureau of Industry and Security spokesperson stated the employer issued guidance explaining export license necessities which have been in place since 2023 and would persist implementing export controls to protect critical American technology.
NVIDIA Guidance
NVIDIA stated the new guidance does not change its position because the Commerce Department had already made clear via a letter that the company require a license to ship the affected chips. AMD, another fundamental supplier of AI chips, did not instantly comment.
Moreover, the guidance does not need data centers to stop the using affected chips or quit servicing for advanced computing items inclusive of servers. That means the action specializes in export and licensing necessities instead of that instant operational shutdowns.
Former State Department official Chris McGuire stated the guidance closes one loophole however leaves some other unresolved. He pointed to the lack of extra due diligence necessities for foundries inclusive of TSMC to verify that advanced chips aren’t being produced for Chinese front companies.
The recent move demonstrates how AI policy increasingly more relies upon on supply-chain enforcement. As governments treat computing as a strategic resource, export controls can also form not only chip markets however also the pace of AI development throughout borders.
