The United States’ control over the export of AI chips may change from “controlling a few countries” to “controlling the whole world.”
The U.S. Department of Commerce has drafted a new regulation: U.S. companies exporting AI chips to any overseas region must obtain a government license and invest in domestic AI infrastructure in the United States.
Approval will be graded based on transaction size. Small orders are subject to basic review; for large orders, the government of the buyer’s country needs to be involved.
This means that when companies like Nvidia and AMD sell chips to Europe, the Middle East, and Southeast Asia, they must first go through Washington. In the past, export controls were mainly targeted at specific countries and specific models. Now, licensing has become the norm.
The accompanying “invest in America” condition is also interesting. It bundles export licensing and industrial policy: if you want to buy American chips, you have to build something in the United States.
For chip companies, this is real uncertainty. The approval process will slow down delivery, and the purchasing plans of major overseas customers will also have to be rearranged. For other countries, the urgency of self-developed chips and local computing power has been pushed one step further.
It is still only a draft, and it has not yet been decided whether the final version will be loosened or when it will take effect. But the direction is already clear: AI chips are treated as strategic materials.