Nvidia asks cloud providers to direct 51% of services to companies outside China
Key Facts
Nvidia has asked its cloud-computing customers to direct at least 51% of their services to companies outside China, Bloomberg reported on November 3, 2026, citing the Financial Times and unnamed sources. The reported directive links cloud providers’ customer mix to their access to Nvidia’s advanced chips. It narrows a route that Chinese companies have used to benefit from computing capacity rented from data centers outside the country. The measure has not appeared as a formal public announcement by the company, according to reports detailing it. The move places controls on the use of computing capacity within efforts to limit China’s access to advanced AI technology.
Tom’s Hardware reported that the instructions cover data-center operators worldwide but apply to new and future contracts, without changing existing Nvidia hardware-rental agreements. The report said sales and regional teams informed customers, though the instructions did not appear in publicly available official documents. This scope means the reported change concerns how new business is distributed among customers, rather than ending existing arrangements. Its application also depends on operators’ decisions when signing subsequent agreements. Nvidia has not announced the policy in a public document, according to the report.
Reports quoted Nvidia as saying, "American chips are the foundation of what American model and application developers need to compete globally." The company added that U.S. policy encourages cloud-service providers to build infrastructure in the United States and accept business from different markets. The statement places the company’s position in the context of competition to develop models and applications and provide the computing capacity they require. According to the statement, the reported policy is based on customers’ share outside China, while leaving room to accept business from multiple markets. The accounts of the company’s remarks did not include publicly disclosed details on how the share would be measured or applied to contracts.
The reports reflect interest from China’s Moonshot lab, developer of the Kimi model, in accessing Nvidia computing capacity through European data centers. According to Tom’s Hardware, citing CNBC, discussions covered renting clusters with 8,000 or more graphics processing units, with a focus on inference and questions also raised about training clusters. The route highlights AI companies’ reliance on renting computing capacity outside China when they cannot access advanced hardware directly. The report said U.S. export restrictions cover Blackwell chips and other advanced Nvidia GPUs. Renting from data centers abroad has provided Chinese companies a way to access this capacity, the report said.
The reported restrictions come amid shifting U.S. policy on semiconductor exports to China. In January 2026, the U.S. Commerce Department’s Bureau of Industry and Security amended its license-review policy to consider export applications for Nvidia H200, AMD MI325X and similar chips on a case-by-case basis, subject to security and buyer-compliance requirements. Undersecretary Jeffrey Kessler said at the time that allowing H200 sales to China under controls "will strengthen the American technology ecosystem." In its disclosure for the second quarter of fiscal 2027, Nvidia said shipments of Hopper data-center products to China were less than 1% of data-center revenue. The segment generated $89.0 billion in revenue in the same quarter, up 117% year over year.
Nvidia said it expects revenue of $108 billion, plus or minus 2%, for the third quarter of fiscal 2027, excluding any data-center computing revenue from China from its forecast. That leaves the Chinese market outside the company’s published estimates for the quarter, while the reported directive targets new and future cloud-service contracts. In broader Chinese trade, customs data cited by the State Council showed growth accelerating from 15.3% in the first quarter of 2026 to 18.3% in the second and 22.3% in the third. Customs official Wang Jun said trade growth in the first three quarters was the fastest for the same period in the past five years. The figures provide trade context ahead of the expected October data.
The MetaTrader calendar page provisionally lists the next release of China’s trade balance for November 10, 2026, at 02:00 GMT, with a forecast of 670.31 billion yuan and a previous reading of 905.97 billion yuan. Separately, Taiwan’s Ministry of Finance Department of Statistics has scheduled publication of preliminary October 2026 trade statistics before 10 a.m. on the same day. That release concerns Taiwan’s statistics, not China’s. The forthcoming figures will update the regional trade picture after Chinese trade growth accelerated in the first three quarters. The releases follow reports of efforts to restrict Chinese companies’ access to advanced computing through overseas providers.