Trump to Revise Biden’s AI Chip Export Controls

On May 8, 2025, the Trump administration announced plans to rescind and replace the Biden-era rule regulating the export of high-end AI accelerator chips—a policy initially intended to curb global proliferation of advanced semiconductors. Officials described the previous framework as “overly complex, overly bureaucratic, and unenforceable,” pledging to implement a streamlined system that both safeguards U.S. national security and fuels domestic AI innovation.

Background: Biden’s AI Chip Export Framework

In January 2025, the outgoing Biden administration issued the Framework for Artificial Intelligence Diffusion, marking the culmination of a four-year effort to restrict so-called “advanced” AI chips abroad. Targeted at limiting China’s access to chips such as Nvidia’s H100 and H200 GPUs, the policy introduced a three-tiered export control system:

• Tier 1: Unlimited exports to 17 allied nations plus Taiwan.
• Tier 2: Quantitative caps on approximately 120 countries.
• Tier 3: Total prohibition for China, Russia, Iran, and North Korea.

The rule was slated to take effect on May 15, but Commerce Department officials have now deferred implementation pending a full rewrite.

Proposed Shift to Global Licensing Model

Sources familiar with the plan suggest the new approach will abandon rigid tiers in favor of a global licensing regime underpinned by government-to-government agreements. Instead of blanket restrictions, U.S. exporters may obtain individual licenses negotiated directly with partner nations—potentially simplifying compliance for manufacturers like Nvidia, AMD, and Intel while preserving strategic oversight.

Technical and Security Implications

High-performance AI accelerators employ multi-die architectures, high-bandwidth memory (HBM3), and proprietary interconnects such as NVLink or Infinity Fabric. These features can significantly boost the throughput of large language models and computer vision workloads. From a security perspective, limiting access to chips with 80+ teraflops of FP16 compute and on-board encryption engines is critical to preventing adversaries from developing autonomous targeting systems.

Market Reaction and Industry Impact

Following the announcement, Nvidia shares rose 3%, reflecting investor confidence that simplified export rules will accelerate global sales of the H100 and the next-gen H200. AMD’s MI300 accelerators and Intel’s Gaudi2 AI ASICs may also benefit. Meanwhile, Chinese foundries like SMIC continue advancing 7nm-class processes, albeit without access to extreme ultraviolet (EUV) lithography, underscoring the geopolitical stakes of semiconductor policy.

Expert Insights and Future Outlook

Morgan Stanley semiconductor analyst Joseph Moore warns that a global licensing model must include robust end-use verification to prevent diversion to military programs. The Commerce Department is expected to release a Notice of Proposed Rulemaking (NPRM) in summer 2025, initiating a 60-day public comment period. Industry groups anticipate that the final rule could be published by Q1 2026, ensuring U.S. dominance in AI hardware while maintaining necessary security controls.