BIS Eases AI Chip Export Restrictions to China

BIS Opens the Door to Increasing AI Chip Exports to China

On January 15, 2026, the U.S. Department of Commerce, Bureau of Industry and Security (BIS) published an amendment to the Export Administration Regulations (EAR) aimed at facilitating increased exports of specific artificial intelligence chips, including NVIDIA H200 and AMD MI325X, to Infrastructure-as-a-Service (IaaS) providers in China and Macau.

While these AI chips still require a license for export, the new rule allows license applications to be reviewed on a case-by-case basis, moving away from the previous categorical presumption of denial for these destinations. To qualify for this review, both the exporter and recipient must meet stringent criteria.

Background

The NVIDIA H200 chips and similar advanced AI chips classified under Export Control Classification Number (ECCN) 3A090.a have historically faced strict restrictions for export to China and Macau, with license applications typically resulting in denial. The January 15 rule marks a significant shift in policy, providing a pathway for exporters to seek approvals more feasibly.

Qualifying for Case-by-Case Review

To qualify for the case-by-case review, the following criteria must be met:

  • Items: AI chips must have a total processing power (TPP) less than 21,000 and a total DRAM bandwidth under 6,500 GB/s.
  • Destination: Exports must be directed to China and Macau.
  • Parties: Transactions must involve parties headquartered outside of Macau or jurisdictions under Country Group D:5.

What Continues to be Reviewed Under a Presumption of Denial?

Despite the new rules, certain items, such as NVIDIA Blackwell and other ECCN 3A090.a AI chips that exceed the specified performance criteria, will still face the presumption of denial for exports to China and Macau.

Rigorous Requirements for Export Licenses

Exporters aiming for a case-by-case review must provide substantial evidence to BIS, including:

  • U.S. Supply Data: Assurance that sufficient supplies exist to fulfill U.S. customer orders without delay, with shipments to China and Macau not exceeding 50% of total TPP shipped to U.S. customers.
  • IaaS Provider Recipient KYC: Verification that recipients employ strict Know Your Customer procedures to prevent restricted end users from accessing AI chip capabilities.
  • Third-Party Lab Testing: Each shipment must be evaluated by a U.S.-based, independent testing lab to confirm compliance with performance criteria.
  • Country of Concern Remote End Users: A list of intended remote users must be provided, ensuring they are not located in countries of concern.
  • Restricted End Uses: The chips must not be intended for military, nuclear, or other restricted applications.
  • IaaS Provider Physical Security: Necessary physical security measures must be implemented at the location where the AI chips will be housed.

Conclusion

The January 15 rule illustrates the U.S. government’s effort to balance national security interests with commercial opportunities for AI chip exports to China. While the pathway for increased exports has opened, rigorous compliance and oversight remain integral to the process, reflecting ongoing concerns regarding the potential misuse of advanced technologies.

More Insights

Revolutionizing Drone Regulations: The EU AI Act Explained

The EU AI Act represents a significant regulatory framework that aims to address the challenges posed by artificial intelligence technologies in various sectors, including the burgeoning field of...

Revolutionizing Drone Regulations: The EU AI Act Explained

The EU AI Act represents a significant regulatory framework that aims to address the challenges posed by artificial intelligence technologies in various sectors, including the burgeoning field of...

Embracing Responsible AI to Mitigate Legal Risks

Businesses must prioritize responsible AI as a frontline defense against legal, financial, and reputational risks, particularly in understanding data lineage. Ignoring these responsibilities could...

AI Governance: Addressing the Shadow IT Challenge

AI tools are rapidly transforming workplace operations, but much of their adoption is happening without proper oversight, leading to the rise of shadow AI as a security concern. Organizations need to...

EU Delays AI Act Implementation to 2027 Amid Industry Pressure

The EU plans to delay the enforcement of high-risk duties in the AI Act until late 2027, allowing companies more time to comply with the regulations. However, this move has drawn criticism from rights...

White House Challenges GAIN AI Act Amid Nvidia Export Controversy

The White House is pushing back against the bipartisan GAIN AI Act, which aims to prioritize U.S. companies in acquiring advanced AI chips. This resistance reflects a strategic decision to maintain...

Experts Warn of EU AI Act’s Impact on Medtech Innovation

Experts at the 2025 European Digital Technology and Software conference expressed concerns that the EU AI Act could hinder the launch of new medtech products in the European market. They emphasized...

Ethical AI: Transforming Compliance into Innovation

Enterprises are racing to innovate with artificial intelligence, often without the proper compliance measures in place. By embedding privacy and ethics into the development lifecycle, organizations...

AI Hiring Compliance Risks Uncovered

Artificial intelligence is reshaping recruitment, with the percentage of HR leaders using generative AI increasing from 19% to 61% between 2023 and 2025. However, this efficiency comes with legal...