US Moves to Block China's Overseas AI Chip Rental Channels: Singapore and Six Other Countries Included, Targeting Nvidia's Top-Tier Blackwell Chips

Beating AI News Flash: The Trump administration is finalizing new AI chip export control rules, preparing to restrict Chinese AI companies from remotely using advanced chips such as Nvidia Blackwell through overseas data centers. The new rules are expected to be announced as soon as this month and are currently undergoing interagency coordination.

The new rules will place Singapore, Malaysia, Thailand, Indonesia, Norway, Brazil, and South Africa under a stricter chip export licensing regime. Data centers operated by local non-U.S. companies that import controlled AI chips such as Nvidia Blackwell and the latest-generation Rubin will need to apply to the United States for licenses, declare the equipment recipients and end uses, and accept ongoing supervision.

Data centers operated by U.S. companies will receive conditional exemptions. As long as they prohibit remote use of the relevant advanced chips by countries subject to U.S. arms embargoes, such as China and Russia, they will not need to apply separately for licenses to import these chips. However, they must still report the identities of customers in the relevant countries and the scale of computing power they use.

Previously, although Chinese AI companies could not directly import some advanced chips, they could still rent GPU computing power through overseas data centers. Recently, Nvidia was reported to have required overseas cloud service providers to allocate at least 51% of their services to non-Chinese customers, and Oracle was also reported to have canceled a computing power leasing contract worth about $7 billion with Tencent under pressure from the U.S. government.

The United States is now preparing to directly restrict overseas data centers from providing advanced computing power to Chinese companies through export licensing and customer declaration systems. Facing the risk of tightening policies, Chinese cloud service providers and AI labs are rushing to book overseas Nvidia computing power.