US Commerce Department Drafts Worldwide Licensing Rule for All Advanced AI Chip Exports

The US Commerce Department has drafted sweeping new regulations that would require federal approval for virtually all exports of advanced artificial intelligence accelerator chips worldwide. According to documents reviewed by multiple outlets, the proposed framework would fundamentally shift the global AI landscape, positioning the United States government as the ultimate gatekeeper for the world's most coveted computing hardware.[1][2]

Rather than an outright ban, the draft rules establish a tiered licensing regime based on the size of the hardware order. This represents a significant expansion from the current country-by-country export control framework, extending federal oversight to every nation outside the United States seeking to acquire leading-edge processors from companies like Nvidia and AMD.[1]

The evidence pack supporting this policy shift centers on the strategic necessity of controlling massive AI data center buildouts. By leveraging its near-monopoly on the design of top-tier AI accelerators, the US aims to dictate not just who receives the hardware, but the security and economic conditions under which foreign AI infrastructure is developed.[2]

The core mechanism of the drafted regulation relies on a strict volume-based tier system. For small installations of fewer than 1,000 advanced GPUs—such as Nvidia's latest GB300 chips—exporters would undergo a streamlined review process.[2]

To qualify for exemptions at this lowest tier, recipients must agree to robust monitoring and install software safeguards that prevent the processors from being networked into massive computing clusters. The evidence suggests this is designed to allow routine enterprise and academic use while preventing the stealth accumulation of supercomputing capabilities.[2]

The requirements scale aggressively for larger deployments. Companies seeking to export up to 100,000 chips would need explicit government-to-government assurances. This diplomatic backstop mirrors the agreements the US previously struck with nations like the UAE and Saudi Arabia in late 2025, ensuring host nations commit to strict security protocols.[1][2]

For the largest "cluster" builds exceeding 200,000 GPUs, the draft rules introduce a novel economic condition: host governments would be required to invest directly in US-based AI infrastructure as a prerequisite for the hardware purchase.[1]

This top-tier requirement effectively links the export of American technological supremacy to domestic economic investment, a move that analysts note transforms export controls from a purely defensive national security tool into an offensive economic lever. Installations of this magnitude could also trigger mandatory on-site verification visits by US export-control officials.[2]

The justification for this sweeping global net stems from documented enforcement gaps in previous, narrower export controls. Recent guidance from the Foundation for Defense of Democracies highlighted how American firms had previously exploited loopholes to sell high-end chips to Chinese subsidiaries located in third-party countries, particularly in Southeast Asia.[4]

By moving to a worldwide licensing requirement, the Commerce Department aims to eliminate these geographic workarounds. The new framework shifts the burden of proof, requiring explicit authorization for any international transfer rather than relying on a patchwork list of restricted entities and nations.[4]

However, the evidence regarding the internal political viability of the draft is highly contested. Reporting from Axios indicates significant friction within the administration, with senior White House officials actively pushing back against the Commerce Department's proposal.[3]

The internal critics argue that the sweeping global licensing regime too closely mirrors the Biden administration's "AI Diffusion Rule," which the current administration explicitly rescinded in May 2025. That previous rule was criticized for stifling American innovation and burdening companies with excessive regulatory requirements.[3][4]

This interagency conflict introduces substantial uncertainty about whether the draft rules will be implemented in their current form. The tension highlights the ongoing struggle to balance national security imperatives against the commercial interests of US tech giants seeking to dominate emerging global markets.[3]

From an industry perspective, the proposed regulations threaten to formalize what some analysts describe as a "de facto" licensing regime that has already begun to take shape. Recent ad-hoc interventions, such as the Commerce Department's June directive restricting foreign access to Anthropic's frontier AI models, have created a highly unpredictable environment for tech firms.

Legal and policy experts warn that relying on case-by-case national security authority, rather than transparent statutory frameworks, undermines trust among global AI users. The sudden imposition of worldwide hardware export controls could further accelerate foreign efforts to achieve technological sovereignty.

The long-term efficacy of such sweeping export controls remains a subject of intense debate. While the US currently holds a commanding lead in AI chip design, critics argue that aggressive gatekeeping will inevitably incentivize the development of alternative, non-US supply chains.

Evidence from the broader technology sector suggests that while export controls can delay adversaries, they rarely prevent eventual parity. Nations facing restricted access to US hardware are already heavily subsidizing domestic semiconductor industries and exploring alternative computing architectures to bypass American bottlenecks.[4]

Ultimately, the Commerce Department's drafted rules represent the most ambitious attempt yet to govern the physical layer of the global artificial intelligence boom. Whether this framework successfully secures American dominance or merely accelerates the fragmentation of the global tech ecosystem remains the central, unresolved question of the AI era.[1][2]