On March 5, news outlets reported that the U.S. government is drafting new export control regulations for AI chips. A copy of the draft regulations is not publicly available. According to reports, the draft regulations would cover most high-end processors sold by U.S. companies, positioning the U.S. as a gatekeeper for the global AI industry.
Reports indicated that the licensing and approval process would depend on the amount of computing power involved in any given export. Shipments of up to 1,000 Nvidia GB300 GPUs or the equivalent would undergo “fairly simple review” with exemptions. Larger exports to companies building extensive computing clusters would require more pre-clearance and be subject to a stricter export licensing regime with conditions attached. Conditions could include disclosure of business models or allowing U.S. government access to facilities for site visits. Large exports of more than 200,000 GB300s or the equivalent would likely require host country government involvement. Exports of this size would only be permitted for projects located in allied countries, and host countries would be required to make “matching” investments in American AI.
The reported draft regulations also come as Nvidia announced that it was halting production of H200 chips intended for the Chinese market, amidst continued U.S. regulatory pressure and Chinese restrictions intended to protect its domestic chip industry. The Commerce Department’s Bureau of Industry and Security (BIS) issued a final rule in January 2026, which changed licensing policy for H200 exports from the U.S. from a presumption of denial to a case-by-case review policy. However, license applicants were required to provide extensive certifications and documentation for case-by-case review eligibility. The administration also issued a presidential Proclamation imposing a narrow 25% tariff on certain semiconductors and related equipment, which included H200s.
If the Commerce Department does issue a rule establishing the reported framework, it would be the Trump administration’s most substantive step to date towards a global chip export control strategy.
In May 2025, the Trump administration rescinded the Biden administration’s AI Diffusion Rule. The AI Diffusion Rule was issued as an interim final rule which created a global license requirement for advanced AI chips classified under Export Control Classification Numbers (ECCNs) 3A090.a, 4A090.a, and related .z items, with license exceptions for certain low-risk destinations and transactions that complied with strict safeguards. It also established a tiered framework for all other destinations, including country-specific allocations and expanded Data Center Validated End User (VEU) authorizations. Upon the Trump administration’s recission of the AI Diffusion Rule, it announced that it would issue a replacement diffusion rule, without providing a timeline for release.
Although aspects of the reported draft regulations seem to mirror the Biden administration’s AI Diffusion Rule, the Commerce Department has denied that it will issue rules that resemble Biden administration policies. It remains unclear exactly how restrictive the reported new framework will be. Questions also remain around how the framework will address model weights and reciprocal AI investment from foreign countries. Reporting indicates that the draft regulations will not significantly change the direction of administration efforts to limit China’s chip production and AI-related technological development.