About 600,000 Nvidia Corp. H200 artificial intelligence chips sit in Hong Kong warehouses, awaiting shipment to mainland China that Beijing has not approved, Reuters reported on November 5, 2026, citing two people familiar with the matter1,2.
The stockpile represents a concrete bottleneck in China's access to advanced AI training hardware. Despite strong demand from Chinese technology firms, Beijing has withheld approval for large-scale imports of the H200, a move Reuters linked to concerns about dependence on U.S. technology.
The hold is not absolute. Three people familiar with the matter told Reuters that China had allowed some H200s into the country in recent months. One source said several major Chinese internet companies had each obtained roughly 10,000 units over the past few months. Three sources also described small shipments through intermediaries and indirect routes, though volumes were not specified.
ANALYSIS The gap between the 600,000-unit backlog and the roughly 10,000-unit parcels reaching individual firms indicates that only a fraction of available supply is clearing into mainland hands.
The constrained supply has inflated mainland pricing. One person familiar with the matter said an eight-card H200 server was offered in mainland China for more than 6 million yuan by the end of October 2026, approximately three times its price a year earlier.
Nvidia has newer AI processors, but the H200 remains useful for training AI models, especially when paired with Nvidia's software, and Reuters described capabilities that Chinese alternatives have yet to match.
China's Ministry of Industry and Information Technology and the National Development and Reform Commission did not respond to Reuters' requests for comment. Nvidia, Alibaba, ByteDance, DeepSeek, and Tencent also did not respond.
ANALYSIS The 600,000-unit stockpile and the tripling of server prices together quantify the cost Beijing's import hold is imposing on Chinese AI firms that still depend on Nvidia silicon for competitive training runs.