US believes Nvidia’s Huang had romantic relationship with alleged Chinese spy

Nvidia has reportedly imposed new restrictions on foreign cloud providers, compelling them to reduce their exposure to the Chinese market. This development is forcing data center operators serving Chinese technology firms to reconsider their business models. According to The Financial Times, Nvidia communicated this requirement verbally to its sales representatives and regional management teams during the summer.

The new condition applies to all future commercial agreements and necessitates that affected companies generate at least 51% of their revenue from customers outside China. Two sources confirmed that the restrictions are not limited to facilities operating in Asian markets but extend to Nvidia's customers worldwide. The company's new policy aims to encourage American cloud providers to build on American infrastructure and secure business opportunities globally.

The US government has been scrutinizing how advanced American processors reach Chinese technology businesses, attempting to restrict access through intermediaries and foreign computing facilities. Chinese technology companies have increasingly relied on foreign cloud providers to access powerful American processors unavailable through direct purchases.

The advanced Nvidia B300 processors, for instance, command approximately $600,000 internationally, while the same rack reportedly sold for about $2.8 million in China due to limited availability. However, the B300 is off-limits for direct sale to China due to American export restrictions. The new restrictions have created challenges for foreign data center operators that previously depended on Chinese businesses for revenue.

These operators must now find sufficient international customers to ensure that at least 51% of their services go to non-Chinese companies. Some cloud providers have expressed concerns that strictly enforcing this policy could significantly impact their business. As Chinese technology companies face mounting computing constraints, finding alternative sources for advanced computing capacity may become increasingly difficult, affecting their access to resources necessary for AI development.