The U.S. Bureau of Industry and Security issued guidance on June 1, 2026, clarifying that export licensing requirements for advanced AI chips apply to any company headquartered in or majority-owned from China, regardless of where that company is physically located. The move closes a loophole that had allowed Chinese-owned subsidiaries operating outside mainland China to purchase Nvidia’s most powerful processors, including its Blackwell-generation GPUs, without the licenses required for direct sales into China.
What the rule actually changes
Under the clarified guidance, any transfer of controlled chips to a company headquartered in China or Macau now requires an export license, even if the purchasing entity’s servers or offices sit in a third country such as Singapore or Malaysia. Former State Department official Chris McGuire said the clarification effectively confirms that “Blackwell shipments to China-headquartered companies outside of China are now illegal again,” and warned that “Chinese companies have been buying these chips, very likely at scale” before the guidance took effect.
A whiplash policy year
The rule is the latest swing in a chip-export policy that has changed direction repeatedly in 2026. The Trump administration scrapped the Biden administration’s Framework for Artificial Intelligence Diffusion in May 2026, having deemed its global licensing-tier system “burdensome” and diplomatically counterproductive. Just months earlier, in January 2026, Commerce had loosened restrictions to allow sales of Nvidia’s H200 chips and AMD’s MI325X chips to Chinese customers, reversing a prior ban. The June clarification on subsidiaries suggests Washington is trying to keep some enforcement teeth even as it relaxes headline restrictions.
Nvidia’s response
Nvidia said in a statement that “licenses are required to ship controlled products to PRC-headquartered companies,” adding that the government’s guidance “reaffirms that NVIDIA’s sales and vetting process is correct.” The company has walked a careful line throughout 2026, publicly supporting continued access to the Chinese market while insisting its compliance processes already meet whatever the current rules require — a position that lets it avoid taking sides in the underlying policy debate.
The market fallout in China
The practical effect has been a collapse in Nvidia’s Chinese AI chip market share, which fell from roughly 95% in 2023 to effectively zero for new H200 shipments by mid-2026, as reported by CNBC. Chinese customers have faced security scrutiny from both governments, and domestic alternatives from firms like Huawei have gained ground in the vacuum, a dynamic Nvidia CEO Jensen Huang has repeatedly flagged as a long-term competitive risk even as he continues to lobby Washington for greater access.
Two camps on chip policy
National-security hawks in Congress argue the subsidiary clarification is overdue and that any gap in enforcement effectively hands frontier AI compute to a strategic rival. Industry advocates and some economists counter that inconsistent, frequently reversed export rules do more to push Chinese firms toward self-sufficient chip supply chains than any coherent export ban would, ultimately weakening rather than strengthening U.S. leverage over the long run.
What to watch next
Nvidia’s fiscal third-quarter earnings, expected in late August 2026, will be the first full quarter reported since the extraterritorial rule took effect, and analysts will be watching the China data-center revenue line closely as a real-time gauge of how strictly the new guidance is being enforced. Any further reversal — in either direction — would ripple through Nvidia’s stock and the broader AI supply chain almost immediately.
The compliance burden on chipmakers
The clarified rule effectively forces Nvidia and its distributors to build corporate-ownership due diligence into every large sale, tracing not just the buyer’s address but its ultimate parent company across potentially multiple jurisdictions. That is a meaningfully harder task than screening a shipping destination, and industry lawyers say it will likely slow deal cycles for large data-center customers in Southeast Asia and the Middle East, where Chinese-linked capital has increasingly funded AI infrastructure buildouts. AMD, which also received clearance to sell certain chips into China under the January 2026 loosening, faces the same expanded screening obligations, meaning the rule’s impact extends well beyond Nvidia alone.
A pattern of policy improvisation
Trade analysts note that 2026’s chip-export policy has been unusually reactive, with the administration abandoning the more structured, multi-tier Diffusion Framework in favor of case-by-case guidance issued as loopholes are discovered. Supporters of the current approach say it lets Washington respond quickly to specific evasion tactics without the diplomatic friction a broad multilateral framework created with allied chip-buying nations. Critics, including some members of Congress on the House Select Committee on the Chinese Communist Party, argue that ad hoc guidance is easier for well-resourced Chinese buyers to route around than a stable, predictable rule set would be, since it invites companies to search for the next unaddressed gap rather than assume broad restrictions apply.