The US Commerce Department issued a new rule requiring export licenses for advanced AI chips to five countries: Indonesia, Malaysia, Singapore, Thailand and Brazil. The rule, titled "Addressing Remote Access to Advanced Semiconductors," aims to close a well-known loophole: Chinese companies that cannot buy advanced chips directly, but use their computing power through data centers in third countries.

What exactly changes?

Until now, advanced AI chips could be exported to these countries without a special license. From now on, any export, re-export or transfer within these countries requires a license from the Bureau of Industry and Security (BIS) at the Commerce Department. According to reports, the countries were designated as high-risk for chip diversion, and the Commerce Secretary can add more countries. The rule also addresses remote access through overseas data centers, and according to some reports existing data center operators will get about 90 days to comply.

Who is exempt?

According to reports, shipments to companies whose headquarters, or ultimate parent, is in the US are exempt from the license, provided the exporter meets reporting requirements. In practice, that means American cloud giants such as Microsoft, Google and Amazon can keep building data centers in these countries. Those hit are mainly local and Asian operators, and smaller cloud companies whose ownership isn't always clear.

Why Singapore and Malaysia?

In recent years Singapore and Malaysia have become major data center hubs in Southeast Asia, and suspicions of chip diversion to China through them have come up again and again. The rule also arrives against the backdrop of the Jensen Huang affair: at its center is Megaspeed, a Singapore cloud company that Commerce Department investigators suspected, according to the New York Times, of diverting advanced Nvidia chips to China. Megaspeed denies it, and no one has been charged. But the timing speaks for itself.

How did the countries respond?

Singapore's government said it is studying the details of the rule and engaging US authorities on implementation, and that it is looking for ways to enable "trusted" access for Singaporean companies. In Malaysia and the other countries the expectation is similar: an attempt to reach an arrangement that lets them keep attracting data center investment without running into an American block.

What does it do to Nvidia?

This is exactly the risk investors feared. According to reports, Nvidia stock fell about 4% on Monday and closed around $156, a day before its earnings. Meanwhile, several brokerages downgraded the stock in recent days, including Bernstein, which halved its price target to $200. Nvidia's guidance already excludes data center compute revenue from China, but Southeast Asia is a significant market for it, and the rule could slow sales there.

And what about Israel?

Israel is not on the list, and that matters. Under the Biden administration's previous rule, the "AI Diffusion Rule" of January 2025, Israel was placed in the middle tier, with quantity caps on chip purchases, and the current administration announced in May 2025 that it was rescinding it. The new rule targets specific countries instead of classifying the whole world, so, at least for now, Israeli companies building data centers or buying advanced chips are not directly affected. However, the Commerce Secretary can expand the list, and any Israeli company with cloud infrastructure in one of the new countries should check the implications.

The bottom line

The chip war is moving from factories to the cloud. Washington is no longer content with controlling who buys chips; it wants to know who uses them, and where. For Nvidia, it adds another layer of uncertainty in a week when it is already dealing with a leadership crisis. And for countries that made data centers a growth engine, it is a reminder that in the AI era, computing infrastructure is also a geopolitical question.