A new export-control rule is taking shape in Washington, and its target is not the chip itself but the wire that connects to it. According to Malaysia's Oriental Daily News, reporting from Washington on Nov 12, the US Commerce Department is preparing a measure to bar Chinese AI companies from remotely accessing advanced AI chip capacity through data centers in third countries. The control scope will add seven nations, including Malaysia, Singapore and Thailand.
Tech outlet The Information, citing two people familiar with the matter, reported that the rule could be announced as soon as this month. It would expand the Bureau of Industry and Security's existing semiconductor controls by adding those countries to a license list with end-user reporting requirements. The aim is to close off remote-use and smuggling channels. US-operated data centers, by contrast, would be conditionally exempt.
The immediate backdrop is the growth of transit routes. Hundreds of thousands of NVIDIA chips have been sitting in Hong Kong awaiting approval. After the Huang scandal, Reuters reported that Chinese AI companies rushed to lease NVIDIA capacity in overseas data centers, and the new rule targets exactly that path.
For transit hubs such as Malaysia and Singapore, the rule brings heavier compliance duties to data-center businesses. For Chinese AI companies, renting overseas compute gets costlier and less predictable, which sharpens the urgency of domestic alternatives. The market has already priced some of this in. Chinese AI-chip stocks have rallied over the past month, with Cambricon hitting successive stage highs on exactly this expectation.
What the rule really does is extend export control from "where the chip is" to "where the compute is", moving the boundary from hardware to access rights. In the window before formal publication, overseas compute rental prices are likely to climb further. Once the rule takes effect, its compliance costs will spread across the whole cross-border AI computing chain.