On May 31, 2026, the U.S. Department of Commerce's Bureau of Industry and Security (BIS) issued an unusual weekend guidance, clearly announcing that it will enforce export license requirements for advanced computing chips on entities headquartered in China, even if those entities are registered or operate outside of China. This move aims to prevent the most advanced AI chips from companies like Nvidia and AMD from reaching China through overseas subsidiaries of Chinese capital.

Core of the New Rule

According to the BIS guidance titled "Guidance on Imposing a License Requirement for Advanced Computing Items to Entities Headquartered in Country Group D:5 and Macau," exporters, when determining whether a transaction requires a license, cannot rely solely on the registered or destination address of the trading party. They must further identify the headquarters location of the "ultimate parent company." As long as the ultimate parent company is headquartered in China or Macau, the relevant entity must apply for a license from BIS to procure the controlled advanced AI chips anywhere in the world. The controlled chips include advanced computing products, encompassing Nvidia's latest Rubin and Blackwell architecture processors, as well as chips like AMD's MI350X.

Policy Background and Limited Exemption

This new rule is a clarification and supplement to existing regulations rather than a completely new creation. The loophole originated in May 2025 when BIS announced a pause in the enforcement of the "Artificial Intelligence Proliferation Rules" introduced at the end of the Biden administration. This ambiguous stance allowed some Chinese companies to continue procuring high-end AI chips, which should have been strictly restricted, through overseas subsidiaries in places like Malaysia without needing a license.

It is noteworthy that the new guidance does not require data centers that have already deployed these chips to cease operations or interrupt maintenance services. BIS explicitly stated that compliantly operating data centers do not need to stop the ongoing use of equipment because of this. This indicates that the policy focus is on curbing new procurement and circumvention channels, with limited direct impact on existing computing infrastructure.

CONEVO Integrated Circuit Distributor

CONEVO is a global distributor specializing in FPGA, MCU, DSP, and passive components, providing supply chain solutions and technical support to OEM/EMS clients across aerospace, automotive, and industrial sectors. Daily Recommended IC Models:

● PC28F128J3F75A: A 128Mb Parallel NOR Flash memory from Intel/Micron with 75ns access time for embedded code storage.

● AT24C04C-SSHM-T: A Microchip 4Kb I2C Serial EEPROM supporting 1.7V-5.5V for parameter storage.

● EP1AGX20CF780I6N: An Intel (Altera) Arria GX FPGA with 20K logic elements in a 780-pin package for high-performance digital logic.

Website: www.conevoelec.com

Email: info@conevoelec.com