Washington has moved to close a one-year-old loophole that allowed Chinese technology companies to acquire banned high-end Nvidia graphics processing units (GPUs) by routing purchases through subsidiaries based in Singapore and Malaysia.
The US Commerce Department’s Bureau of Industry and Security (BIS) issued guidance on May 31 clarifying that export licenses are required for advanced computing chips destined for any entity whose ultimate parent company is headquartered in China or Macau, regardless of where that entity is situated. It stressed that the rule has technically been in force since November 2023.
The loophole emerged after the Trump administration announced in May 2025 that the AI Diffusion Rule, which was drafted by the Biden administration to impose stricter worldwide limits on the export of advanced AI chips based on destination tiers, would not take effect. Many Chinese firms grabbed the opportunity, rushing to build data center racks in Singapore and Malaysia and purchase Nvidia chips in large quantities.
Hundreds of thousands of banned Nvidia chips may have already reached Chinese-owned subsidiaries through this channel, according to industry estimates cited by Reuters.
BIS is now making clear that the parent-company requirement was never suspended. Some observers noted that by framing this as a clarification rather than a new rule, the Trump administration stops short of acknowledging that a loophole ever existed, casting the problem instead as one of compliance and enforcement.
BIS also said it would not require data centers already operating with the banned chips to cease using them, a carve-out that sidesteps potential liability and compensation claims against chip suppliers.
Regardless of how Washington frames the measure, the practical effect is clear. Chinese-owned entities operating overseas will no longer be able to freely purchase Nvidia chips that are banned, ending an arrangement that had gone largely unchallenged for a full year.
The enforcement shift comes as Beijing has been pressing its domestic technology firms to prioritize homegrown artificial intelligence (AI) chips over foreign alternatives, such as Nvidia’s H200 and H20 chips. Huawei Technologies’ Ascend 920 has been at the center of those efforts, with Chinese authorities encouraging companies to adopt domestic AI processors rather than rely on Nvidia’s products.
Some Chinese commentators say the tightened enforcement will have a significant impact on Chinese firms’ data center operations in Southeast Asia in the short run. But they add that Chinese firms will find alternative ways to access the banned processors.
“Chinese firms’ overseas subsidiaries had previously exploited the ambiguity in the rules, procuring Nvidia’s Blackwell-architecture chips in full compliance and without any export licenses. The new guidance will tighten the regulatory grip and raise compliance costs and legal risks for these companies,” says a Liaoning-based columnist using the pen name “Little Chunping.”
“But enforcement remains challenging. Proving that an overseas chip buyer is ultimately Chinese-owned is far harder than it sounds. Corporate structures can span multiple jurisdictions, and foreign governments are under no obligation to help Washington trace them,” he says.
He adds that possible countermeasures include deploying more intermediaries and financial instruments to obscure transaction chains, shifting toward technology licensing or purchasing cloud computing services that bundle access to advanced chips, and establishing more substantive research and production partnerships in less regulated jurisdictions to gain indirect access to the technology.
“Do not be fooled by the word ‘clarification.’ In export control language, a clarification often hits harder than a new rule. BIS has now made explicit that any entity whose ultimate parent is headquartered in China or Macau needs an export license, regardless of whether it is registered in Singapore, Malaysia, the Middle East or anywhere else,” says a Tianjin-based writer using the pseudonym “Ling Gen.”
“This is less about discovering a loophole and more about admitting one was created,” he says. “Washington’s own regulatory zigzagging over the past year opened a grey channel for these chips to flow through. Now the political narrative needs to reframe that policy mismanagement as an urgent national security gap that must be plugged.”
He says the move works like a slow squeeze with two immediate consequences:
• Existing chips in data center racks are untouched, but every future purchase order becomes a license application stuck in permanent review. AI companies may be able to cope with existing stockpiles for now, but will hit a bottleneck when upgrading to more powerful next-generation models.
• Projects already mid-stream face immediate uncertainty. The guidance covers future exports but says nothing about chips already at sea or orders signed but not yet shipped. With no transition period offered, contracts are left in legal limbo and suppliers have already begun to hit the brakes. Affected data centers may pivot to Nvidia’s H200 chips, which remain outside the scope of the current restrictions.
AI Diffusion Rule
When Donald Trump returned to the White House in January 2025, one of his early moves in the technology arena was to shelve the AI Diffusion Rule, a sweeping export control framework that the Biden administration had pushed through in its final days in office.
The rule would have imposed tiered restrictions on the worldwide export of advanced AI chips, dividing countries into three categories based on their perceived security risk to the US and capping the total volume of chips that could be shipped to most destinations. Its suspension left a regulatory vacuum that Chinese firms were quick to exploit.
Last July, media reports indicated that Washington was planning to tighten export controls on Nvidia’s AI chips, specifically targeting Malaysia and Thailand, to prevent Chinese firms from using them to train AI models. But the May 31 guidance could only be launched a full year after the loophole was created.
“The reason they had to issue this statement is that BIS’s non-enforcement of certain export controls has potentially inadvertently allowed Chinese companies to both buy Nvidia Blackwell chips and make AI chips at TSMC, all legally and without a license,” Chris McGuire, a former US State Department official specializing in technology and national security, says in a social media post on June 1. “This is a huge problem.”
He adds: “This statement does not say that BIS will enforce the parts of US regulations requiring TSMC to do enhanced due diligence on AI chip orders. This is a massive loophole that still needs to be closed. If Chinese companies can make chips at TSMC, including by using third-country cutouts to receive them, there is no point to restricting China’s access to AI chips or advanced chip-making tools.”
McGuire says BIS created the problem by announcing it would stop enforcing certain regulations without clearly stating which rules were affected or updating its policies to reflect what it was still enforcing. He says the resulting confusion left companies and regulators uncertain about what was and was not permitted, opening gaps that some firms were quick to exploit.
A Shandong-based writer says Trump’s shifting approach to chip export controls over the past year reveals a broader strategic playbook. He says Washington’s strategy operates on three tracks:
- Containing China at the core while keeping the H200 chip available as a revenue stream and to prevent China from becoming fully self-sufficient;
- Pressuring allies such as Taiwan Semiconductor Manufacturing Company (TSMC) and Samsung to build factories on US soil;
- Selectively expanding chip access to Middle Eastern countries to draw them into Washington’s technology orbit.
He says China is now at a disadvantage as Huawei’s best AI chip, Ascend 920, can only match Nvidia’s H20, which is even slower than the H200 chips approved for export to China.
In fact, Washington had been aware of the grey-market chip flow in Southeast Asia long before the May 31 guidance.
Last October, the New York Times reported that the US Commerce Department was investigating Megaspeed, a Singapore-based company spun off from a Chinese gaming firm in 2023, over suspicions that it had helped Chinese companies circumvent US export restrictions to obtain banned Nvidia chips.
The investigation revealed that Megaspeed had purchased through its Malaysian subsidiary nearly US$2 billion worth of advanced Nvidia chips, which were then shipped to data centers in Malaysia and Indonesia that appeared to be remotely serving Chinese clients. US officials were also examining whether some of those chips had been transhipped directly to China in violation of US law.
In April this year, Bain Capital’s Bridge Data Centers removed Megaspeed from their Malaysian computing hub. Since Megaspeed was probed by both US and Singaporean authorities, its office has gone dark.
Read: China’s H200 hunger drives Nvidia chip smugglers to Japan route
Follow Jeff Pao on X at @jeffpao3
