Nvidia pushes cloud customers to cut back on China business

Accessing high-end Nvidia chips remotely through foreign cloud providers becomes more difficult.

Nvidia is guiding its cloud customers to reduce business with China to close a loophole in U.S. export controls. The directive, communicated verbally, asks cloud providers and their backed neoclouds to diversify their customer base and disclose the identity of end users accessing their chips. This measure targets the practice of Chinese AI developers remotely accessing advanced Nvidia chips via cloud services hosted in data centers outside China, bypassing import restrictions. U.S. authorities are scrutinizing this practice, particularly after allegations surfaced involving Moonshot AI's alleged use of such methods. The guidance applies to future contracts and is not mandatory.

This move by Nvidia suggests a proactive effort to comply with or anticipate evolving U.S. geopolitical tech policies, particularly concerning advanced AI hardware and China. By discouraging direct cloud rentals to Chinese entities, Nvidia aims to preempt further government restrictions and maintain its market access in other regions. The requirement for end-user disclosure indicates a push toward greater transparency in the supply chain for critical AI components.

The directive signals a growing complexity in the global AI hardware market, where geopolitical considerations increasingly influence business operations. Cloud providers and AI developers will need to navigate these restrictions, potentially leading to shifts in infrastructure choices and development strategies for Chinese AI companies. The effectiveness of this guidance will depend on adoption by cloud providers and the potential for U.S. authorities to implement more direct regulatory measures.