The U.S. government is exploring a new strategy to tighten the noose on China’s AI development. According to a report from The Information, officials are considering placing a specific cap on the number of AI chips individual customers in China can purchase from Nvidia. This move would limit the volume of silicon any single entity can acquire, marking a shift from restricting hardware capabilities to restricting scale.
Why this is happening
To understand the significance, you have to look at the current dynamic between Washington and Silicon Valley. The U.S. previously banned the export of Nvidia’s top-tier chips (like the H100) based on performance thresholds. Nvidia responded by designing the H20, a chip that sits just below those limits, specifically for the Chinese market.
The concern now is “stacking.” While an individual H20 chip complies with regulations, a Chinese tech giant buying 100,000 of them to link together creates a cluster capable of training frontier models. A per-customer cap directly targets this workaround.
The strategic shift
This represents a significant change in regulatory philosophy. Until now, the Department of Commerce focused on what you can buy. Now, they are looking at how much you can buy.
Here is why this matters for the industry:
- The end of the volume loophole: This closes the door on using volume to overcome performance deficits. It effectively puts a ceiling on the size of training clusters Chinese firms can build using American hardware.
- Compliance pressure: If enacted, this puts immense pressure on Nvidia and its distributors to verify end-users. Expect stricter “Know Your Customer” (KYC) protocols to prevent shell companies from buying chips to aggregate later.
- Accelerating domestic alternatives: Every restriction placed on Nvidia makes Huawei’s Ascend chips look more attractive to Chinese buyers. The U.S. risks accelerating the very independence it hopes to delay.
What comes next
Nvidia has walked a fine line, aiming to comply with U.S. laws while maintaining its foothold in the massive Chinese market. If these caps go into effect, it forces a hard limit on Nvidia’s revenue potential in the region, regardless of how creative their engineering teams are at staying compliant. This signals that the U.S. is willing to micromanage supply chains down to the customer level to maintain its lead in AI development.