Nvidia has verbally directed its cloud compute clients to ensure more than 51 percent of their services go to non-Chinese companies, effectively capping Chinese customer revenue below the halfway mark.
The directive targets the indirect supply of AI compute capacity to Chinese entities. Cloud providers that purchase Nvidia's high-performance GPUs are now subject to these usage requirements, closing a loophole in U.S. export controls that previously allowed Chinese firms to access advanced chips through third-party cloud infrastructure.
This approach extends the U.S. government's semiconductor restrictions beyond direct sales. While prior export bans focused on chip shipments to Chinese companies, the compute mandate targets consumption patterns, requiring cloud service providers to actively manage their customer mix.
The verbal nature of the instruction operates outside formal written regulations, allowing Nvidia flexibility while placing compliance responsibility on its cloud partners. Major global providers must now adjust sales strategies and client portfolios to align with the 51 percent threshold—a material constraint on their addressable market in China, one of the fastest-growing AI regions.
For Nvidia, the measure protects its regulatory standing while maintaining deniability through an informal enforcement mechanism. The company avoids the reputational and legal exposure of a formal policy while achieving the same outcome: preventing Chinese AI development from scaling via cloud compute workarounds.
