Nvidia reportedly requires cloud providers to get 51% of revenue outside China

Nvidia has reportedly told foreign cloud providers that at least 51% of their revenue must come from customers outside China to qualify for future commercial agreements. The requirement applies to customers worldwide and does not change existing cloud computing contracts.

The Financial Times, citing people familiar with the discussions, reported that Nvidia communicated the condition verbally through sales representatives and regional managers during the summer. The company said American-built data centers and cloud services support US companies competing worldwide, and that US policy has long backed building American infrastructure around the world.

The reported requirement comes amid US scrutiny of how advanced processors reach Chinese technology companies. Some Chinese firms have used foreign cloud providers to access processors they cannot buy directly, including for demanding AI work such as training large language models.

Nvidia’s B300 processors show the price gap described in the report: an entire rack reportedly sells internationally for about $600,000, while a rack sold in China for about $2.8 million in October 2026, amid limited availability. US export restrictions prohibit direct sales of the B300 to China.

Nvidia has also reduced by more than half this year the number of Asian companies authorized to buy its AI chips, and added compliance checks to prevent restricted processors from reaching China through unauthorized suppliers or intermediaries. A representative of a Singapore-based cloud provider said finding enough international customers outside the United States had become difficult and warned that strict enforcement would hurt business. The policy could make it harder for Chinese technology companies to rent advanced computing capacity abroad.