A Commerce Department proposal to prevent data centers outside of the People’s Republic of China from leasing super-Nvidia-chip computing power to firms inside the PRC—or maybe only lessen such leasing—is still in flux.
A New York Times piece that focuses on Commerce Secretary Howard Lutnick’s Nvidia-related resistance to chip regulations and possible conflicts of interest notes that many aspects of the proposed rule are still being debated. These include whether the rule would even apply to data centers already equipped with U.S. superchips (November 15, 2026).
People familiar with the discussions say the Commerce Department has pushed for a more scaled-back version of the rule. The department has proposed applying regulations to a handful of countries—Singapore, Malaysia, Thailand, Indonesia, Brazil, South Africa and Norway. Others in the administration have wanted global coverage.
The proposal was first reported by The Information.
The Commerce Department proposal would also apply only to future shipments of advanced chips, exempting data centers already filled with American technology. Companies would need to notify the government if Chinese customers use their data centers to train A.I. models. But they would not automatically need to obtain a license.
The Commerce Department had also argued for the rules to kick in when companies are using a larger amount of computing power, meaning fewer activities by Chinese companies would be regulated.
Chris McGuire, a senior fellow at the Council on Foreign Relations and a former Biden administration official, predicted that Chinese companies would find ways to sidestep the more narrow controls the Commerce Department is proposing.
“If you’re only restricting remote access to data centers in certain countries, then China will just rent from the data centers outside those countries,” he said.
McGuire’s objection is unanswerable. But what about the data centers of the “certain countries” which would be subject to the proposed rule?
Those data centers would still be useful to Chinese firms determined to evade Nvidia-chip restrictions if the restrictions kick in only after a specified cap on computing-power usage has been reached. In that case, Chinese firms would have an incentive to allocate or label their remote usage in such a way as to appear to satisfy a data center’s compliance requirements, even if the firm in fact uses more than the maximum allowed amount of super-Nvidia-chip computing power during a time period. Or maybe some data centers themselves would cook the books.
Only closing loopholes, not half-closing them, has any chance of slashing the flow of superchips or superchip computing to China.
Also see:
The New York Times: “Lutnick Pushed to Weaken Chip Rules That Overlap With His Family Business”
Bloomberg: “Alice Huang-Linked Firm Procured Banned Nvidia Chips in China”