U.S. rule targets Chinese remote access to computing exceeding 12 million performance points
Key Facts
On November 16, the administration of U.S. President Donald Trump announced a new rule restricting Chinese companies’ remote access to advanced U.S. computing chips and servers through data centers outside China, shifting oversight to how computing capacity is used and where it operates. The rule prohibits making available more than 12,000,000 aggregate performance points when serving companies headquartered in China or Macau, including through interconnected data centers. The U.S. Commerce Department said the aim is to prevent circumvention of existing restrictions through remote-computing arrangements and data centers in third countries. The rule covers data centers wherever they are located if it is known that advanced computing chips or equipment will provide substantial capacity to a company headquartered in China or Macau. The controls therefore extend to the computing services provided by the infrastructure, as well as to the movement of chips and equipment themselves.
The rule requires a license to export, re-export or transfer advanced semiconductors to Indonesia, Malaysia, Singapore, Thailand and Brazil. This covers five destinations that host data centers or related computing investments, according to the scope of destinations specified in the rule. The text includes a conditional exception for some shipments destined for companies headquartered in the United States or companies whose parent is headquartered in the United States. This means the shipment’s destination alone does not determine its treatment; application of the exception is tied to the identity of the recipient company and the rule’s conditions. These provisions are intended to limit Chinese companies’ access to advanced computing capacity through parties or locations outside China.
Oversight covers companies headquartered in China or Macau, as well as companies whose ultimate parent is headquartered there, if data centers provide them with substantial computing capacity directly or indirectly. The text includes the use of interconnected centers, bringing computing-capacity networks within the scope of the restrictions rather than treating each center separately. Commercial recommendation systems used exclusively for commercial purposes, such as those used by retail stores or social media platforms, are exempt from the licensing requirement. This exception defines a range of commercial uses outside the licensing requirements set out in the rule. The Commerce Department said the controls aim to prevent circumvention of existing restrictions through remote-computing arrangements.
Data centers that were providing substantial computing capacity to these companies before the rule was published can apply for a license if they want to continue service. The rule gives these centers 90 days from its effective date to submit a license application. This deadline ties continued service to compliance by existing operators, alongside licensing requirements for semiconductor transfers to the specified destinations. The decision’s scope is not limited to new shipments, as its provisions also cover centers already providing computing capacity to the entities concerned. This could put existing operating arrangements under licensing review within the specified period.
The rule comes amid efforts by Chinese artificial intelligence companies to rent computing capacity powered by Nvidia chips at data centers outside China, Reuters reported before its announcement. The report said the companies had been seeking to secure computing capacity amid concerns that Washington would tighten restrictions. John Moolenaar, chairman of the U.S. House Select Committee on China, said the rule sends a message that the United States is serious about closing export-control loopholes exploited by China. In response, a Chinese Foreign Ministry spokesman accused Washington of interfering in commercial and technological cooperation between China and other countries. He said Beijing would take the necessary measures to protect its interests.
In markets, Nvidia shares closed at $156.14 on November 16, compared with a close of $173.73 on November 11, according to EL7 data. In the same session, the S&P 500 fell 1.9%, the Nasdaq dropped 2.5%, and the Dow Jones lost 739 points. The Associated Press said weakness in technology stocks weighed on the Nasdaq and that Nvidia led a selloff in technology shares. These figures show the session’s performance and the stocks’ movement over the period cited, without establishing that the rule caused the decline. Closing prices for AMD or TSM were not included in this context.
Attention is turning to a meeting between U.S. President Donald Trump and Chinese President Xi Jinping in the Chinese city of Shenzhen on Thursday, according to a Reuters report published on November 16. The anticipated meeting comes after Washington announced new controls on cross-border access to computing capacity. Reuters said further measures against China may be in preparation and also pointed to differing views within the Trump administration on how to deal with Beijing. The meeting is set to put U.S.-China relations in focus during the week the new rule took effect. The report did not say whether controls on access to chips and computing would be raised at the meeting.