U.S. Tightens AI Chip Export Rules: New Restrictions Target Chinese Subsidiaries and Close Major Loophole

Published: June 1, 2026 Last Updated: June 1, 2026 By Editorial Team

The new enforcement directive targets a gap in export controls that has persisted for roughly a year, potentially enabling the transfer of Nvidia’s most sophisticated processors, including the Rubin and Blackwell architectures, as well as AMD’s MI350x, to Chinese firms based in locations such as Malaysia. The guidance posted on the Commerce Department’s website states that license requirements now apply to entities headquartered in China even when those entities operate outside the country’s borders.

The opening emerged in May 2025 when the Trump administration announced it would not enforce the AI Diffusion rule, a regulation issued in the final days of the Biden administration that governed global access to advanced computing semiconductors. By suspending enforcement, the Commerce Department created a pathway for overseas subsidiaries of Chinese AI companies to acquire restricted chips without obtaining export licenses.

Chris McGuire, a technology expert and former State Department official, highlighted the severity of the situation in a social media post on Sunday. “This is a HUGE problem,” McGuire wrote, noting that the loophole allowed overseas subsidiaries to purchase Nvidia Blackwell chips without a license. “Chinese companies have been buying these chips, very likely at scale,” he added.

Exactly how many advanced processors flowed through this channel remains unclear. However, one chip industry source with deep supply chain knowledge estimated that the volume could be in the hundreds of thousands over the past year. The processors in question represent the cutting edge of AI acceleration technology, with performance characteristics that place them well above thresholds established under current U.S. export control frameworks.

You may also like:  Samsung Exynos 2700 Chip for Galaxy S27 Set to Drop WLP Tech

According to regulatory data, chips like Nvidia’s Blackwell B200 exceed Total Processing Performance thresholds of 21,000 and memory bandwidth above 8 TB/s, placing them firmly in the category of next-generation architectures subject to strict licensing requirements. By contrast, earlier generation hardware such as the H200 and AMD’s MI325X fall below certain ceilings that allow for case-by-case review under specific conditions, though even those transactions face heightened scrutiny.

Industry Impact and Regulatory Context

The semiconductor industry has navigated a rapidly shifting regulatory landscape since October 2022, when the Bureau of Industry and Security first introduced performance-based thresholds to restrict AI-capable chip exports. Controls tightened through 2023 and 2025, eventually establishing a broad presumption of denial for advanced computing exports to China and Macau. A January 2026 amendment introduced case-by-case review for certain chips below specific performance ceilings, but next-generation architectures remain categorically restricted.

Nvidia and AMD, the two American chipmakers most directly affected by the new guidance, did not immediately respond to requests for comment. The development adds fresh complexity to an already tense market, where demand for AI accelerators continues to outpace supply and geopolitical considerations increasingly shape product roadmaps. Companies like Samsung are also navigating this terrain as they prepare future silicon designs, including the Exynos 2700 chip for the Galaxy S27, which must account for evolving global trade policies.

The restrictions also intersect with broader trends in consumer technology and AI deployment. As Samsung and Google reveal new AI smart glasses at recent developer conferences, the underlying hardware powering these innovations faces increasing regulatory oversight. Even compact flagship phones arriving this autumn reflect a market where top-tier silicon allocation is becoming a geopolitical consideration.

You may also like:  Computex 2026 Kicks Off in Taipei With AI Hardware in the Spotlight

Notably, the Commerce Department’s guidance does not require data centers to cease using chips already deployed or to cut off servicing of advanced computing equipment such as servers. This creates a complex operational picture where existing infrastructure may remain in place even as new procurement channels close.

Analysts expect the clampdown to pressure major chipmakers’ revenue projections in the near term, particularly in markets with significant Chinese corporate presence. The move signals Washington’s continued commitment to restricting Beijing’s access to foundational AI hardware, even as companies seek alternative supply routes and foreign subsidiaries attempt to secure advanced components.

The Commerce Department did not immediately respond to requests for comment on the timing of the guidance or the specific intelligence that prompted the weekend announcement.

Topics: ,

What is your Opinion?