The United States considers “globalizing” the export approval system for AI chips.

```

The United States is assessing upgrading AI chip export controls from “restricted to certain countries” to a “global licensing system”.

According to Bloomberg and other media reports, U.S. government officials are drafting an export control proposal, considering requiring companies like NVIDIA and AMD to obtain approval from the U.S. before exporting AI chips to “almost all global destinations”.

Sources say the new rule is not intended to “ban” NVIDIA exports, but to place the U.S. government at the “gatekeeper” position in the global AI industry chain—who can acquire sufficient computing power, and under what conditions, will be determined more directly by the U.S. Department of Commerce’s licensing method.

The U.S. Department of Commerce also proposed that if companies in relevant countries wish to purchase large volumes of NVIDIA and AMD chips for AI data centers, they must commit to investing in the U.S. The new rule proposes a three-tier approval system based on computing power needs: simplified for small orders, pre-approval and additional disclosure or inspection conditions for large clusters; for super-large orders exceeding 200,000 chips, the buyer’s country may be required to make commitments and link them to U.S. investment.

Currently, the draft is still in the interdepartmental consultation phase. Sources told the media the text may be significantly revised or temporarily shelved.

After the news broke, NVIDIA and AMD share prices dropped during Thursday’s trading.

From “40-country controls” to “global licensing”

Reports say the draft will require companies to apply for licenses for exports of “almost all” NVIDIA and AMD AI accelerators, effectively broadening the current framework—which covers about 40 countries—to global scope.

The U.S. Department of Commerce also emphasized this is not a return to the previous administration’s “AI proliferation rule”.

The Department of Commerce said: “There are reports today that we’re returning to the ‘AI proliferation rule.’ We are not. It is burdensome, overly intrusive, and disastrous in consequence.”

“Trading investment for licenses”: Middle East model may be institutionalized

According to the Financial Times, the highest tier in the draft will require the buyer’s country to commit to investing in U.S. domestic AI infrastructure, as one condition to obtain large quantities of advanced chips.

Binding export approval to investment commitment already has a precedent. Last year, when the U.S. approved chip exports to UAE and Saudi Arabia, it reached a similar arrangement: UAE committed to matching every dollar invested domestically with the same amount invested in the U.S.

The Department of Commerce commented:

“We successfully promoted exports through a historic Middle East agreement, and the government is discussing how to formalize such practices.”

An American official also made it clear: “Any regulations issued by the government will aim to promote the U.S. technology stack.”

Bloomberg also pointed out, the speed of approval and strength of attached clauses will determine whether the impact is simply more paperwork or a substantive blockade: If licenses are issued quickly with few conditions, global AI data center construction may proceed, albeit with higher paperwork costs; if there are delays or extended negotiations, project planning and delivery pace could be disrupted. After the U.S. announced the UAE chip export arrangement last year, licenses were only issued months after, with a “reciprocal U.S. investment” condition attached.

Tiered approval: bigger scale, stricter additional conditions

According to the new rules, the approval process is tiered based on the computing power scale needed by companies:

Tier one: purchases of up to 1,000 NVIDIA GB300 GPUs. These transactions will undergo relatively simple review and have specific opportunities for exemption.

Tier two: medium to large cluster deployments. Buyers must undergo pre-approval before applying for export licenses. Depending on circumstances, buyers may be required to accept additional conditions, such as disclosing their business model fully, or allowing U.S. government onsite inspection of their data center.

Tier three: super-large deployments, namely when a single enterprise in one country owns over 200,000 GB300 chips. At this point, the buyer’s national government must intervene. The U.S. will require strict security commitments from the buyer’s country and "matching" investments in the U.S. AI infrastructure. As a scale reference, 200,000 GB300 chips is equivalent to the total scale of the AI infrastructure contract planned by UK NScale for Microsoft.

Under the new export control framework, chip companies must meet direct financial allocation conditions to obtain specific licenses. According to sources, in negotiations over high-end H200 chip export licenses to certain markets, NVIDIA CEO Jensen Huang has agreed to a condition: NVIDIA will remit 25% of the sales revenue of these H200 chips to the U.S. government in exchange for green-lighted exports.

Risk warning and disclaimerThe market has risks; investment requires caution. This article does not constitute personal investment advice, nor does it take into account special investment goals, financial situations, or needs of individual users. Users should consider whether any opinion, viewpoint, or conclusion in this article suits their particular circumstances. Investment based on this, at your own risk. ```