The guidance targets a potential pathway that may have allowed some of the world’s most powerful AI processors—including chips from Nvidia and AMD—to reach subsidiaries of Chinese firms operating outside mainland China.
At the center of concern are high-end accelerators such as Nvidia’s Blackwell and Rubin architectures, alongside AMD’s MI350X series, which are widely regarded as critical infrastructure for training frontier AI systems.
A Policy Gap That May Have Been Exploited
The Commerce Department’s updated guidance clarifies that export licensing requirements will now apply to companies headquartered in China, even if those entities are operating through subsidiaries in third countries such as Malaysia or other regional tech hubs.
This closes a gap created after the United States temporarily paused enforcement of the Biden-era AI Diffusion Rule in May 2025. That earlier policy framework was designed to tightly regulate global access to advanced AI chips but was not actively enforced for a period, creating what officials now appear to view as an unintended opening.
According to reporting and industry estimates, the scale of potential diversion could be significant. One supply-chain source suggested that “hundreds of thousands” of advanced chips may have moved through indirect channels over the past year.
“This Is a HUGE Problem”
The reaction from policy and technology analysts was immediate.
Chris McGuire, a former U.S. State Department official, wrote on social media that the loophole represented a serious enforcement failure, stating:
“This is a HUGE problem.”
He added that overseas subsidiaries of Chinese companies were “very likely at scale” acquiring advanced processors, including Nvidia’s most recent AI hardware, without the necessary export approvals.
The concern is not just about hardware movement, but about how quickly such chips can be deployed into AI training clusters capable of supporting large-scale model development.
New Rules Target Ownership, Not Just Location
A key shift in the updated guidance is its focus on corporate ownership rather than geography alone. Under the revised interpretation, licensing requirements will apply based on where a company is headquartered—not just where it is physically located.
This approach is intended to prevent companies from bypassing restrictions by routing purchases through foreign subsidiaries in countries that are not directly subject to the same export controls.
However, the rule does not currently require companies or data center operators to stop using chips that were already acquired. Nor does it mandate shutdowns or service interruptions for existing infrastructure built using the restricted hardware.
Industry Implications for AI and Semiconductor Supply Chains
The update adds another layer of complexity to an already tightly controlled global semiconductor market.
Companies like Nvidia and AMD sit at the center of the AI boom, supplying the GPUs that power everything from large language models to advanced simulation systems. Export restrictions on their most advanced chips have become a key tool in U.S. technology policy aimed at limiting China’s access to frontier AI capabilities.
At the same time, the global nature of chip distribution networks makes enforcement difficult, particularly when hardware is routed through third countries with active data center expansion.
The latest guidance suggests Washington is now attempting to tighten not just the export rules themselves, but also the interpretation of who ultimately qualifies as the end user.
A Sign of Tighter AI Containment Strategy Ahead
While the Commerce Department has not publicly commented on enforcement details, the shift signals a broader tightening of U.S. controls over AI infrastructure.
Rather than focusing only on physical shipment destinations, the new approach expands scrutiny to corporate structures, ownership chains, and indirect procurement pathways.
For the global AI industry, the message is increasingly clear: access to cutting-edge compute is no longer just a commercial issue—it is becoming a central axis of geopolitical policy.
What remains uncertain is how quickly enforcement will follow policy, and whether the reported flow of advanced chips through third countries can now be meaningfully slowed or reversed.