Nvidia Tells Global Cloud Clients to Cut Reliance on Chinese Customers

Nvidia has begun attaching a new condition to its dealings with overseas cloud providers, asking them to lessen their dependence on Chinese firms that rely on its high-end AI chips, according to TechRadar. The requirement calls for the affected operators to draw no less than 51 percent of their revenue from buyers located outside China, a threshold that could upend arrangements already in place.

The rule is not limited to any single region, and two sources said it covers Nvidia customers across the globe rather than only facilities operating in Asian markets. Data center operators that serve Chinese technology companies stand to feel the impact most sharply.

The Financial Times, citing people familiar with the discussions, reported that Nvidia passed the condition along by word of mouth through its sales staff and regional management. The company brought the measure in during the summer and aimed it at future commercial agreements, leaving cloud computing contracts that already exist untouched.

Under the policy, cloud businesses must show they have enough demand from international clients before they can carry on their commercial ties with Nvidia. The company said in a statement that computing facilities running on American silicon give US model and application developers the base they require to rival others around the world. It added that Washington has long backed the construction of American infrastructure abroad, which it said brings prosperity, influence and jobs back home.

The move arrives as US authorities grow more watchful about how cutting-edge American processors find their way to Chinese technology businesses. Officials have been trying to close off access gained through go-betweens, overseas computing sites and other commercial routes that sidestep existing export controls. Chinese firms have leaned increasingly on foreign cloud operators to reach advanced chips they cannot buy directly, using that capacity for heavy AI work such as training large language models.

The financial gap shows why the market persists. A full rack of Nvidia's B300 processors sells for roughly $600,000 internationally, yet one reportedly changed hands in China for about $2.8 million in October 2026 amid tight supply. Direct sales of the B300 into China are barred by American export rules. Nvidia has also more than halved the number of Asian companies cleared to buy its AI chips this year and added tighter compliance checks to keep restricted parts from reaching China via unauthorised suppliers.

Foreign data center operators that once counted on Chinese clients now have to find replacement customers. A cloud provider based in Singapore said lining up enough international business outside America has proved hard, and a representative of the group said strict enforcement would definitely hurt business.