For roughly a year, Chinese companies found a workaround to US chip export controls that was almost embarrassingly simple: set up a subsidiary in Malaysia or Singapore, order the chips there, and sidestep the licensing requirements entirely. The US Department of Commerce just slammed that door shut.

On May 31, the Bureau of Industry and Security published new guidance mandating that any shipments of advanced computing chips to entities headquartered in China now require appropriate export licenses, regardless of where the purchasing subsidiary is physically located. The ruling specifically targets Nvidia’s Blackwell and Rubin processors, along with AMD’s MI350x chips.

How the loophole worked

The gap traces back to the Biden administration’s “AI Diffusion Rule,” which went into effect in May 2025. Because the original framework focused on the physical destination of shipments rather than the ultimate corporate parent, Chinese-headquartered companies could route purchases through overseas subsidiaries. A data center operator based in Shenzhen could theoretically place orders through an affiliate registered in Southeast Asia and receive the same chips that a direct shipment to mainland China would have required a license for.

Estimates suggest that during this non-enforcement window, hundreds of thousands of advanced computing chips may have reached China-controlled companies abroad.

What the new guidance actually changes

The BIS guidance redefines the compliance trigger. Instead of asking “where is this chip being shipped?” the question becomes “who ultimately controls the entity receiving it?” If the answer is a Chinese-headquartered company, a license is required.

Nvidia confirmed that the new guidance does not alter its pre-existing licensing obligations. This latest move doesn’t change what Nvidia can or can’t sell under existing licenses. It changes who can buy without one.

The guidance also doesn’t affect chips already deployed in data centers. BIS appears to be drawing a line going forward rather than attempting to claw back hardware that’s already operational.

The bigger picture on US-China tech restrictions

This move fits neatly into a multi-year escalation of US semiconductor export controls targeting China, with significant tightening occurring since 2022. The strategy aims to maintain American technological superiority in AI by restricting access to the hardware that makes advanced models possible. China has poured resources into domestic chip development, and companies like Huawei have made notable progress on homegrown alternatives.

What this means for investors

For Nvidia and AMD shareholders, closing the loophole means fewer chips sold to Chinese-controlled entities without licenses, which could reduce total addressable market in the near term. Both companies have already been pricing in export restrictions for years, and the guidance doesn’t retroactively change existing license agreements.

Companies that had been routing purchases through overseas subsidiaries now face a binary choice: apply for a US license (which may or may not be granted) or find alternative, non-restricted hardware.