US Commerce Department closes loophole on Nvidia chip exports to China

For roughly a year, Chinese companies had a neat workaround for US chip export controls: just buy the chips through a subsidiary in Malaysia or Singapore. The US Commerce Department just killed that trade.

The Bureau of Industry and Security issued guidance on May 31 mandating that advanced AI processors, including Nvidia’s Blackwell and Rubin chips and AMD’s MI350x, now require export licenses when sold to Chinese-headquartered entities. The key change: it doesn’t matter where the delivery happens. A Chinese company’s subsidiary in Southeast Asia is still a Chinese company in the eyes of BIS.

How the loophole worked

The gap in enforcement traces back to May 2025, when the original export control framework left room for interpretation. Chinese firms quickly figured out that routing purchases through foreign subsidiaries, entities technically located outside mainland China, allowed them to acquire restricted chips without triggering license requirements.

The scale of the workaround was not trivial. Estimates suggest hundreds of thousands of advanced chips made their way to Chinese entities through these channels.

What the new guidance actually does

The updated BIS framework shifts the compliance trigger from geography to ownership. It no longer matters where the chips are delivered. What matters is who ultimately controls the entity placing the order.

If a company is headquartered in China, its subsidiaries anywhere in the world now need proper export licenses to purchase high-end AI processors. This applies to Nvidia’s Blackwell and next-generation Rubin architectures, as well as AMD’s MI350x.

The distinction is important because it targets the corporate structure rather than the shipping address. Previous controls essentially asked exporters to verify a destination country. The new guidance asks them to verify the destination entity’s parent company.

For Nvidia and AMD, this means their sales teams now need to conduct deeper due diligence on customers across Asia, not just in China proper.

The bigger picture for semiconductors

The US-China technology decoupling has been accelerating since 2022, when US export controls on advanced semiconductors first began tightening significantly to impede Beijing’s advancements in AI. The AI Diffusion rule introduced during the latter part of the Biden administration sought to impose global licensing requirements but fell short of enforcement as of May 2025, inadvertently enabling Chinese firms to procure advanced chips via overseas channels.

For Nvidia specifically, China and the broader Asian market have been crucial revenue drivers. Nvidia has already navigated multiple rounds of export restrictions by developing China-specific chip variants with reduced capabilities. AMD faces similar headwinds with its MI350x, which is positioned as a direct competitor to Nvidia’s high-end AI accelerators.