Nvidia restricts Chinese companies' access to cloud computing power

The company Nvidia has begun pressuring its cloud customers, requiring them to reduce the volume of services provided to companies from China. These measures are aimed at complying with U.S. export restrictions, which prohibit the supply of the most powerful GPUs for AI training to the PRC.

According to internal requirements, cloud service providers must review their contracts and restrict access to computing power based on Nvidia chips for clients associated with the Chinese market. This strategy is intended to minimize the risks of violating the sanctions regime, which is constantly being tightened by the U.S. government.

For cloud providers, this means the need to conduct more thorough Know Your Customer (KYC) checks and potentially revise the resource allocation architecture in data centers. In the event of non-compliance with the new directives, the company may restrict providers' access to future hardware shipments, which is critical for maintaining competitiveness amid the shortage of high-performance accelerators.

A practical consequence of these actions will be more difficult access for Chinese tech giants to cloud computing outside the country. Companies using leased capacity to train their models will face the need to find alternative suppliers or revise their strategies for scaling their AI projects.