NVIDIA Directive Requires 51% Share of Cloud Services for Non-Chinese Customers
Key Facts
NVIDIA has told data center customers to allocate at least 51% of their services to non-Chinese companies, according to media reports, in a directive that could restrict Chinese labs’ access to advanced computing capacity through the cloud. Tom’s Hardware reported the information on November 3, citing a Financial Times report, and said sales teams had conveyed the condition to customers. The company has not published an official document confirming the directive, according to the coverage. NVIDIA said in a statement cited by Tom’s Hardware that “American chips are the foundation American model and application developers need to compete globally.” AFP reported that the company neither confirmed nor denied the report, while the reported directive links the distribution of cloud computing services to customer identity.
TechRadar reported that the condition applies globally to future commercial agreements and does not cover existing cloud computing contracts. That means its potential impact is tied to new contracts, while existing arrangements remain outside its scope, according to the coverage. Tom’s Hardware describes the other side of the allocation as a 49% cap on business with Chinese entities. The two reported shares therefore amount to at least 51% for non-Chinese customers and a maximum of 49% for Chinese customers. These details remain media reports, as no official text of the directive has been published.
As companies look to access computing outside China, Moonshot Lab discussed renting clusters of Blackwell units at European data centers, Tom’s Hardware reported, citing CNBC. The talks covered clusters with 8,000 or more processors, with an emphasis on inference and interest also in securing computing capacity for training. The discussion puts European cloud services among the options the lab has considered for accessing large-scale processing capacity. The coverage did not say the talks resulted in a rental agreement. It also gave no details on when the clusters might come online or the terms for making them available.
The discussions come as NVIDIA disclosed the scale of its revenue from customers headquartered in China and Hong Kong. Its quarterly filing with the U.S. Securities and Exchange Commission showed revenue of 7.88 billion dollars from this group in the quarter ended July 26, 2026. The classification measures the location of the direct customer’s headquarters and does not necessarily identify the end user’s location or where the product was shipped. In its second-quarter results announcement, Chief Executive Jensen Huang said “demand is accelerating,” adding that the company sees new AI labs, startups and leading labs expanding in parallel. This provides context for the broadening demand for computing that coincides with the reported directive on how partners distribute services.
A previous company disclosure shows the cost of restrictions related to the Chinese market on its business. NVIDIA said in its filing with the U.S. Securities and Exchange Commission that it recorded a cost of 4.5 billion dollars in the first quarter of its fiscal 2026, related to excess H20 inventory and purchase obligations after demand for the chip declined. The company directly linked the cost to the drop in H20 demand that quarter. In its forecast for the third quarter of fiscal 2027, it estimated revenue at 108 billion dollars, plus or minus 2%. The company said it assumed no data center computing revenue from China in that forecast.
EL7 data showed NVIDIA shares closed at 178.67 dollars on November 9, 2026. A single-session closing price is not enough to determine the reported directive’s effect on the stock. The company is scheduled to hold a call to discuss its fiscal 2027 third-quarter results on November 17. That quarter ended on October 25, 2026. The call will allow the company to present results for the period before the scheduled date, while its announcement of the call did not detail the directive’s impact on its business.