US Commerce Department moves to block Nvidia and AMD chip flows to Chinese overseas units
The US Commerce Department has closed a loophole that may have allowed hundreds of thousands of advanced Nvidia and AMD AI chips to reach Chinese firms via overseas subsidiaries over the past year.
Summary:
- The US Commerce Department issued guidance on Sunday enforcing licence requirements for advanced AI chips sold to entities headquartered in China, even when those entities are located outside China, per the Commerce Department website
- The loophole was created in May 2025 when the Trump administration announced it would not enforce the AI Diffusion rule introduced in the final days of the Biden administration, per Reuters
- One chip industry source with supply-chain knowledge estimated that hundreds of thousands of chips may have been exported during the period the opening existed, according to Reuters
- Chips potentially affected include Nvidia’s Rubin and Blackwell processors and AMD’s MI350x, per Commerce Department guidance
- Former State Department official Chris McGuire described the situation as a major problem, saying Chinese companies had likely been purchasing Nvidia Blackwell chips at scale through overseas subsidiaries, per social media
- The new guidance does not require data centres to cease using or servicing chips already installed, per the Commerce Department
The United States moved on Sunday to close a year-old regulatory gap that may have allowed the world’s most advanced artificial intelligence chips to flow to subsidiaries of Chinese technology firms operating outside China, in an unusual weekend intervention by the Commerce Department that caught the industry off guard.
The new guidance, posted on the Commerce Department’s website, enforces licence requirements for advanced semiconductors sold to entities headquartered in China regardless of where those entities are physically located. The practical effect is to extend existing chip export controls to overseas subsidiaries of Chinese AI firms, including those based in locations such as Malaysia, which had previously sat in an ambiguous regulatory position.
The loophole traces back to May 2025, when the Trump administration announced it would not enforce the AI Diffusion rule that the Biden administration had finalised in its closing days. That rule had governed global access to advanced AI chips. By stepping back from enforcement, Washington inadvertently created an opening that, according to one chip industry source with deep supply-chain knowledge cited by Reuters, may have resulted in hundreds of thousands of chips leaving the United States bound for Chinese-affiliated buyers over the intervening year. The precise number of chips exported through this channel remains unclear.
The chips potentially affected are among the most powerful commercially available. Nvidia’s Rubin and Blackwell processors and AMD’s MI350x sit at the frontier of AI computing hardware, and access to them has been a central front in Washington’s broader effort to limit Chinese progress in artificial intelligence development. Neither Nvidia nor AMD responded to requests for comment on the guidance.
Chris McGuire, a technology policy expert and former State Department official, described the situation in stark terms, saying the loophole had allowed the overseas subsidiaries of Chinese companies to acquire Nvidia Blackwell chips without a licence, and that purchases had very likely occurred at scale.
One notable limit to Sunday’s action is that the guidance does not compel data centres to stop operating or servicing chips already deployed. Installed hardware can continue running, which reduces the immediate disruption to existing infrastructure but leaves open questions about the longer-term status of those deployments under tightened rules.
In a separate development, Nvidia has received US government licences to sell its H200 chips, its second-most powerful AI processor, and around ten Chinese firms have been cleared as buyers. No deliveries have been made, however, with Chinese regulatory approval still outstanding as Beijing continues to direct support toward domestic chip suppliers.
Sunday’s guidance represents the latest escalation in a sustained US effort to constrain Chinese access to advanced semiconductors. The move also raises questions about the adequacy of existing enforcement mechanisms and whether further guidance targeting third-country routing of restricted technology is likely to follow.
—
The guidance lands as an unwelcome surprise for Nvidia and AMD given its weekend timing and the scale of potential exposure, with one industry estimate putting chips exported through the loophole in the hundreds of thousands. Markets will focus on whether the clarification triggers retroactive scrutiny of past shipments, which could weigh on both stocks at the open. The carve-out allowing data centres to continue using and servicing already-installed chips limits the immediate operational disruption, but the signal that Washington is actively tightening the perimeter around Chinese AI hardware access adds a fresh layer of regulatory risk to the sector. For semiconductor supply chains routed through Southeast Asia, particularly Malaysia, the guidance introduces new compliance uncertainty that could slow deal flow while firms assess exposure.
This article was written by Eamonn Sheridan at investinglive.com.