Nvidia limits Chinese chip access via new sales cap

The 49 percent sales cap

Nvidia is reportedly implementing a new restriction on how many of its chips can be sold to companies that rent out hardware to Chinese customers. According to the Financial Times, the company is capping these sales at 49 percent. This means that firms providing cloud services must balance their hardware distribution to ensure they are not primarily serving Chinese interests.

The company is reportedly conveying this rule to its global clients through unwritten, verbal agreements. This method of communication allows Nvidia to exert influence over its distribution network without making a formal public policy change. It serves as a mechanism to manage how American technology reaches the Chinese market via third-party intermediaries.

The goal of this cap is to narrow the backdoor that Chinese firms use to access high-end American artificial intelligence chips. By limiting the volume of hardware available to these specific types of rental companies, Nvidia aims to reduce the amount of computing power flowing into China. This move comes as the company faces scrutiny over how it manages its business relationship with the Chinese market.

Protecting Super Intelligence

An Nvidia spokesperson stated that the United States cannot win the Super Intelligence race unless the nation's infrastructure wins worldwide. Super Intelligence, or SI, is a term used to describe a push to rebrand AI, which became official following an executive order on September 29. The spokesperson argued that if global clouds rely on foreign-made chips, then the resulting models and applications will also be built by foreign powers.

The spokesperson emphasized that American policy has long encouraged global data centers to run on U.S. chips. This strategy is designed to ensure that American AI companies can compete on a global scale. By securing the underlying hardware layer, the U.S. hopes to maintain control over the foundational technology that powers modern AI systems.

The move is a direct response to the need to protect the model and application layers of the AI industry. If foreign chips become the standard for cloud computing, the underlying logic and software of the future could be dictated by other nations. Nvidia's actions reflect a broader effort to align its commercial interests with these national security and infrastructure goals.

Impact on global clients

The new verbal requirements began appearing in Nvidia's communications during the summer months. The company reportedly initiated these instructions to get ahead of criticisms regarding its handling of China. By establishing these limits early, Nvidia seeks to proactively manage its reputation and its compliance with broader geopolitical trends.

However, enforcing these rules presents significant logistical challenges for the company. A Singapore-based group told the Financial Times that it is difficult to find enough non-Chinese clients outside of the United States to fill the gap. This scarcity makes it hard to maintain high sales volumes while simultaneously adhering to the new restrictions on Chinese-linked hardware.

There is a risk that these restrictions could harm Nvidia's overall business growth. If the company cannot find enough alternative customers to replace the volume lost from Chinese-linked sales, it may face production surpluses. The company must navigate a delicate balance between meeting its commercial goals and adhering to the new unwritten rules regarding Chinese access.

Demand vs. production limits

The pressure on Nvidia's supply chain is evident in the recent requests from Chinese tech giants. Reuters reported on October 20 that companies including ByteDance, Alibaba, and Tencent requested more than 3 million RTX Pro 5500 chips. This massive order represents a total value of over $35 billion, highlighting the intense demand for high-performance computing power.

Nvidia may not be able to honor these specific requests because they exceed the company's current production capacity. Even without the new 49 percent cap, the sheer volume of the request would be difficult to fulfill. This creates a situation where demand for AI hardware continues to outpace the ability of the manufacturer to provide it to the market.

The situation is further complicated by security concerns regarding how AI models are developed. On September 8, a joint cybersecurity advisory from the NSA, CISA, and the FBI urged American AI companies to identify bad actors who are distilling model data. These security concerns add another layer of complexity to how Nvidia manages its exports and distribution of high-end chips.

The white list strategy

In addition to the 49 percent cap for rental companies, Nvidia has created a separate 'white list' for its Asian clients. This specific list is designed to cut back sales to certain Asian customers by more than 50 percent. This dual-track approach allows the company to apply different levels of restriction depending on the specific nature of the client and their geographic location.

The use of a white list suggests a more targeted approach to export controls and sales management. Rather than a blanket ban, Nvidia is choosing specific partners to limit based on their potential impact on the AI landscape. This allows the company to maintain its presence in the Asian market while specifically curbing the flow of technology to high-priority targets.

Moving forward, the effectiveness of these unwritten rules and white lists will depend on how strictly they are enforced by third-party clients. As the competition for AI dominance intensifies, Nvidia's ability to manage its supply chain while navigating complex international regulations will be a key factor in its market position. The company remains at the center of a struggle over who will provide the foundational hardware for the next generation of technology.