This website uses cookies

Read our Privacy policy and Terms of use for more information.

US Closes Loophole That Let Chinese Firms Access Nvidia's Most Advanced AI Chips Through Overseas Operations

The U.S. Department of Commerce posted emergency guidance on Sunday, May 31, 2026, closing a loophole that had allowed Chinese companies to receive America's most advanced AI chips by routing purchases through subsidiaries based outside China. The move directly targets Nvidia's Blackwell and Rubin processors, as well as AMD's MI350x, and takes effect immediately.

The unexpected guidance suggests that the United States' best AI chips may have been making their way to the subsidiaries of Chinese AI firms based in places like Malaysia for almost a year, despite broader U.S. efforts to starve Chinese firms of the semiconductors needed to develop critical AI capabilities. CNBC

How the Loophole Worked

The loophole emerged when the Commerce Department opted not to enforce the AI Diffusion rule in May 2025, a decision made during the final days of the Biden administration. This oversight enabled Chinese companies to acquire Nvidia Blackwell chips without a license through overseas subsidiaries. Benzinga

In practice, a Chinese AI company could set up or use a subsidiary in Malaysia, Singapore, or another third country, purchase Nvidia's most advanced chips through that entity, and face no export license requirement - even though the ultimate beneficiary was a Chinese-headquartered firm. That gap stayed open for nearly a full year.

One chip industry source with deep supply-chain knowledge estimated the number of chips that passed through this loophole was in the hundreds of thousands. The Commerce Department did not confirm a specific number, and it remains unclear how much advanced compute capacity reached Chinese firms during that window. South China Morning Post

What the New Guidance Changes

In its weekend guidance, the Commerce Department said it would enforce license requirements for advanced chips to entities headquartered in China, even when those entities are located outside China. That is the core change - the restriction now follows the company's headquarters, not just the physical location of the purchase or delivery. South China Morning Post

However, the new guidance does not mandate data centers to cease using chips already acquired through the loophole or to stop servicing related computing equipment. Chips already delivered are not being recalled. The restriction closes the door on future shipments, not past ones. Benzinga

The Bigger Picture: Nvidia's China Revenue Problem

This story is part of a much longer-running tension between U.S. export policy and Nvidia's business reality. Despite export controls, Nvidia's operations in China remained substantial, with reports suggesting that over 20% of its fiscal year 2026 compute revenue was still derived from China through intermediaries. Benzinga

That figure illustrates why enforcement gaps matter so much. Export controls that exist on paper but leave workarounds open do not meaningfully restrict Chinese AI development - they just redirect the supply chain through third countries. Malaysia, in particular, has become a significant transit point for advanced chips headed to Chinese buyers, a pattern that U.S. officials have been watching for months.

For business leaders tracking the AI industry, this is a reminder that semiconductor access remains the single most contested resource in the global AI race. Every major Chinese AI model - from DeepSeek to the Zhipu GLM series - depends on compute infrastructure. Restricting that infrastructure is Washington's primary lever for slowing Chinese AI development, which is why closing even a technical loophole carries significant strategic weight.

What This Means for Businesses

For companies with AI supply chains that touch Southeast Asia, this guidance creates immediate compliance questions. Any entity headquartered in China - regardless of where it operates - now requires a license to receive Nvidia's top-tier chips. Companies acting as intermediaries or resellers need to review their customer lists against that standard immediately.

For enterprises evaluating AI for business infrastructure decisions, the broader signal is that the chip supply chain will remain a geopolitically managed resource for the foreseeable future. Pricing, availability, and access to the most advanced AI coding tools and infrastructure will continue to be shaped as much by Washington policy as by market dynamics.

Cut Through the Noise

What loophole did the US just close on Nvidia AI chip exports? The Commerce Department posted guidance on May 31, 2026 requiring export licenses for advanced AI chips - including Nvidia's Blackwell and Rubin processors and AMD's MI350x - shipped to any entity headquartered in China, regardless of where that entity is physically located. The loophole had allowed Chinese firms to receive chips through overseas subsidiaries in countries like Malaysia without a license for nearly a year.

How did Chinese firms access Nvidia chips through the loophole? The gap opened in May 2025 when the Commerce Department chose not to enforce the AI Diffusion rule. Chinese AI companies exploited it by routing chip purchases through subsidiaries based outside China, primarily in Southeast Asia. One chip industry source estimated hundreds of thousands of chips passed through this channel before the loophole was closed.

Does the new guidance require data centers to return chips already received? No. The Commerce Department's guidance does not require data centers to stop using chips already acquired through the loophole or to cease servicing existing equipment. The restriction applies only to future shipments, not hardware already delivered.

Why does Nvidia chip access matter for China's AI development? Advanced chips like Nvidia's Blackwell and Rubin processors are the core compute infrastructure required to train and run frontier AI models. Restricting Chinese firms' access to these chips is the U.S. government's primary mechanism for slowing China's AI development. Nvidia derived an estimated 20% or more of its fiscal year 2026 compute revenue from China through various intermediary channels despite existing export controls.

Keep Reading