
Nvidia has stopped shipping its licensed Hopper-generation data center chips, including the H200, to China, interim Chief Executive Colette Kress said on the company’s earnings call on Tuesday. Guidance for the coming quarter includes no such sales, closing a channel the U.S. government approved last year and that produced about $2.4 billion of revenue in the third quarter.
The decision comes on Nvidia’s first call since co-founder Jensen Huang stepped down as CEO last week. The company said the same day it is creating a board-level committee to oversee national security policy, export-control compliance and global trade risks.
The H200 route was negotiated by Huang and President Donald Trump. Washington cleared sales to about 10 Chinese companies, among them Alibaba, Tencent and ByteDance, with the chips required to pass through the U.S. so the government could collect a 25% tariff. The arrangement drew opposition from China hawks in Washington, who saw it as a loophole in the ban on advanced-chip sales to China.
The halt follows new U.S. rules issued on Monday that bar advanced American chips from foreign data centers serving Chinese customers remotely. Jefferies estimates the rules could remove $44 billion to $62 billion from Nvidia’s data center revenue next year. Buyers would have to certify that they serve no Chinese customers, an export-control attorney said, and those that cannot may turn to other suppliers.
Kress said the backlog of demand for advanced products such as Vera Rubin means the rules would have no impact on Nvidia’s top line.
Beijing had already been the obstacle. China largely refused H200 deliveries even before Nvidia’s decision, with a Georgetown researcher saying the government prefers to build demand for domestic chips. Licensed volumes were therefore a small and uncertain part of the business, though the $2.4 billion third-quarter figure shows sales did occur.
The bigger exposure may lie outside China. Critics put up to 20% of Nvidia’s revenue as linked to Chinese demand, much of it routed through Southeast Asian customers in Singapore, Malaysia, Thailand and Indonesia. Nvidia’s Singapore subsidiary reportedly generated $104 billion of revenue in 2025, before the company stopped disclosing such figures.
Chinese AI developers also appear to have kept access to restricted hardware by other means. All of China’s leading AI models show evidence of training on Grace Blackwell chips, which are barred from export to China.
Sources: Bloomberg, techcrunch.com, usnews.com.









