Nvidia reportedly tells data center customers to trim exposure to the Chinese market

Nvidia is reportedly asking foreign cloud providers to prove that most of their business comes from outside China. The new condition requires at least 51% of revenue to come from non-Chinese customers. It matters because providers that depend heavily on Chinese technology companies may struggle to keep buying or deploying Nvidia’s advanced AI processors.

Nvidia reportedly communicated the rule verbally through sales representatives and regional managers. The requirement applies globally and was introduced during the summer for future commercial agreements. Providers must show enough international demand before continuing commercial relationships with Nvidia. Existing cloud contracts are not covered by the reported change.

The policy could reshape cloud businesses serving Chinese customers. A Singapore-based provider said finding enough international customers outside America was difficult. If the rule is strictly enforced, providers may need to diversify their customer base, reduce Chinese business, or risk losing access to future Nvidia arrangements.

The rule most directly affects foreign cloud providers and data center operators that buy or deploy Nvidia’s advanced AI processors. It is especially important for companies whose revenues depend substantially on Chinese technology businesses. These providers may need to demonstrate stronger demand from customers outside China before entering future commercial arrangements with Nvidia.

Chinese AI companies are affected as customers rather than as the direct recipients of the rule. Many have used foreign cloud providers to access processors they cannot buy directly. Those providers offer rented computing capacity, allowing Chinese companies to train large language models and perform other demanding AI operations.

The pressure extends beyond Asian facilities. Nvidia reportedly applies the requirement to customers worldwide, while also reducing the number of approved Asian chip buyers and adding stricter compliance checks. Foreign operators serving Chinese firms now face a choice: find enough non-Chinese business, reduce Chinese exposure, or risk more difficult access to Nvidia hardware.

Chinese AI companies could lose an important route to advanced American processors. They have increasingly depended on foreign cloud providers because direct purchases of certain powerful Nvidia chips are restricted. If those providers must reduce their Chinese exposure, Chinese companies may face fewer places to rent the computing power needed for AI development.

The key mechanism is indirect access. A foreign cloud or data center can deploy Nvidia processors outside China and rent computing capacity to Chinese technology businesses. But Nvidia’s reported 51% revenue condition makes that business model harder for providers that rely heavily on Chinese customers. Stricter checks also aim to stop chips reaching China through unauthorized suppliers and intermediaries.

The likely result described in the article is tighter access to advanced computing resources. Chinese companies could face reduced availability and higher costs, although the article does not quantify those effects. Foreign operators may seek more international customers, while Nvidia’s restrictions could further narrow the routes available to Chinese AI developers.

The reported price gap is about $2.2 million per rack. An Nvidia B300 rack sells for approximately $600,000 internationally, while one reportedly sold for about $2.8 million in China in October 2026. The Chinese price is therefore roughly 4.7 times the international price, based on the figures provided.

The difference reflects the article’s description of limited availability in China. The B300 is among Nvidia’s most advanced processors, but US export restrictions prohibit its direct sale into China. That constraint makes access scarce and creates strong financial incentives for alternative supply routes, including foreign computing facilities and intermediaries.

These figures illustrate why the market matters to cloud providers and Chinese AI companies. Providers can earn much more from scarce hardware in China, but Nvidia is tightening controls and reducing approved Asian buyers. The new revenue rule could make it harder for foreign operators to depend on Chinese demand, even when the price premium is substantial.

US export controls are government rules that restrict the sale, transfer, or provision of specified goods and technology to particular countries, companies, or users. They are used for national-security and policy goals. In the article, those controls prohibit Nvidia from directly selling its B300 processors into China, limiting Chinese companies’ direct access to an advanced American technology.

The controls matter because powerful processors can support demanding AI operations, including training large language models. US authorities are also examining indirect routes, such as intermediaries, foreign computing facilities, and commercial arrangements that could circumvent existing restrictions. Nvidia has responded with stricter compliance checks and fewer approved Asian chip buyers.

The article presents export controls as part of a broader effort to restrict advanced American processors from reaching Chinese technology businesses. They do not necessarily eliminate every indirect route, but they make direct purchases unavailable and increase scrutiny around alternatives. Nvidia’s new cloud-provider condition is intended to reduce those backdoor channels further.

Chinese companies may access American AI chips indirectly by renting computing capacity from foreign cloud providers or data centers. Instead of buying processors directly, they use systems operated elsewhere for AI workloads. The article says Chinese technology companies have increasingly depended on this arrangement because some advanced Nvidia processors cannot be purchased directly in China.

The mechanism is commercial access rather than ownership. A foreign operator obtains Nvidia hardware, installs it in a data center, and provides remote computing services to a Chinese customer. Intermediaries or unauthorized suppliers can also create routes for restricted processors to reach Chinese businesses. US authorities are trying to restrict these arrangements, and Nvidia has added compliance checks.

Nvidia’s reported 51% rule targets this exposure. Foreign providers must show that at least 51% of revenue comes from customers outside China for future agreements. That could reduce the number of operators willing or able to serve Chinese AI companies, narrowing their access to advanced computing.

AI processors are specialized chips built to handle the large number of calculations required by artificial intelligence. Cloud computing delivers access to computers and processors through remote data centers, so users can rent capacity instead of owning all the hardware. The article links these resources to demanding AI operations, including training large language models.

Training a large language model requires powerful computing systems to process vast amounts of information and adjust the model repeatedly. Nvidia’s advanced processors provide that computing capability. A cloud provider can place the processors in a data center and let customers use them remotely. This arrangement is why Chinese technology companies have used foreign providers to access chips unavailable through direct purchases.

The article describes American-chip data centers and clouds as the foundation for American model and application makers competing worldwide. Nvidia’s restrictions matter because they affect who can build and rent these systems. If providers reduce Chinese business, Chinese AI companies may find advanced training capacity harder to obtain.