Nvidia is reportedly asking its data centre and cloud-computing partners to reduce their business exposure to Chinese customers, introducing another obstacle for Chinese AI companies seeking access to advanced computing power.
According to the Financial Times, the chipmaker has instructed cloud providers to ensure that at least 51% of their customers are based outside China. The guidance reportedly applies to future contracts rather than existing leasing agreements and has been communicated verbally through Nvidia’s sales teams. The move could affect overseas data centres that rent Nvidia-powered computing capacity to Chinese AI developers, even when the servers themselves are located outside China.
Key Takeaways
- Nvidia has reportedly asked cloud-computing partners to reduce their exposure to Chinese customers.
- The reported guidance requires at least 51% of cloud customers to be based outside China.
- The instructions were reportedly communicated verbally through Nvidia’s sales and regional teams.
- The guidance is not described as a mandatory restriction at this stage.
- The reported policy focuses on future contracts, not existing leasing agreements.
- Nvidia is also asking cloud providers to identify the end users of its AI chips.
- Chinese AI developers could face greater difficulty renting advanced computing capacity from overseas providers.
- The development comes amid growing US scrutiny of indirect access to advanced AI chips.
What Has Nvidia Asked Its Data Centre Partners to Do?
The Financial Times reported on November 3 that Nvidia had instructed its data centre partners to reduce their exposure to Chinese customers.
The reported guidance includes a requirement for cloud-computing providers to ensure that at least 51% of their customers are non-Chinese.
The Information separately reported that Nvidia had communicated the guidance to some cloud-computing customers, including companies operating specialised AI infrastructure businesses known as neoclouds.
These providers purchase or operate large quantities of GPUs and rent computing capacity to businesses developing artificial intelligence models.
According to the reporting, Nvidia’s instructions are intended partly to encourage cloud providers to diversify their customer bases rather than depend heavily on Chinese businesses.
The guidance has reportedly been communicated verbally through sales representatives and regional teams instead of through a formal written policy.
Importantly, the reported requirement is not currently described as a mandatory rule. Nvidia has not publicly announced a new global prohibition on cloud providers serving Chinese customers.
Why Is Nvidia Targeting Overseas Cloud Providers?
The development highlights a challenge facing US authorities as they attempt to restrict China’s access to advanced semiconductor technology.
US export controls limit the shipment of certain high-performance AI chips to China.
However, companies do not necessarily need to physically import GPUs to access their computing power.
They can rent processing capacity from overseas data centres that already operate Nvidia hardware.
For example, a Chinese AI company could potentially lease GPU capacity from a cloud provider operating servers in Singapore, Malaysia or another international location.
The Chinese company would access the computing resources remotely rather than physically importing the restricted hardware.
This creates a potential gap between restrictions on physical chip exports and access to computing services.
Nvidia’s reported guidance appears designed to address this indirect route by encouraging cloud providers to diversify their customers and disclose who ultimately uses the company’s hardware.
How Chinese AI Companies Could Be Affected
Chinese AI developers rely on substantial computing resources to train large language models, run experiments and operate commercial AI applications.
Companies such as Moonshot AI, Alibaba, DeepSeek and other developers compete in a market where access to advanced GPUs can influence model development speed and performance.
If overseas cloud providers become more cautious about serving Chinese customers, those companies could face several challenges.
Reduced access to overseas GPUs: Cloud providers may become less willing to allocate large amounts of Nvidia computing capacity to Chinese businesses.
Higher computing costs: A smaller pool of available providers could reduce competition and increase prices for GPU rentals.
Slower model development: Limited access to advanced hardware could make large-scale training and experimentation more difficult.
Greater dependence on domestic chips: Chinese developers may increasingly turn to domestic semiconductor suppliers and alternative computing architectures.
However, the impact will depend on how widely Nvidia’s guidance is adopted and whether providers interpret it as a commercial recommendation or a strict condition for future business.
Why the 51% Customer Requirement Matters
The reported 51% threshold is particularly significant because it focuses on the customer composition of cloud providers rather than simply the physical location of servers.
A data centre could operate entirely outside China but still generate a substantial portion of its revenue by serving Chinese AI companies.
Under the reported guidance, such a business could need to expand its customer base in other markets.
For example:
| Customer composition | Chinese customers | Non-Chinese customers |
|---|---|---|
| Example A | 60% | 40% |
| Example B | 49% | 51% |
| Example C | 30% | 70% |
Under the reported threshold, Example A would fall below the suggested non-Chinese customer requirement, while Examples B and C would meet it.
These are illustrative examples, not actual Nvidia customer figures.
The reported policy also focuses on future contracts. Existing GPU leasing agreements are not automatically affected by the guidance described in the reports.
Nvidia Wants Greater Visibility Into Who Uses Its Chips
Another important element of the reported instructions concerns customer identification.
According to The Information, Nvidia is asking cloud providers to submit the identities of the end users of its chips.
This could give the company greater visibility into how its AI hardware is being used after it has been installed in overseas data centres.
The distinction matters because a cloud provider may purchase Nvidia GPUs and offer computing services to multiple businesses.
Without information about the final customers, Nvidia may have limited visibility into whether its hardware is ultimately supporting companies in markets subject to additional restrictions.
Greater customer transparency could help Nvidia assess commercial and regulatory risks associated with its global distribution network.
However, the company has not publicly explained how it intends to verify customer identities or enforce the reported guidance.
US Export Controls Are Creating Pressure on Nvidia’s Global Business
Nvidia occupies a central position in the global AI infrastructure market.
Its GPUs are widely used for training and running large AI models, making the company an important supplier to cloud providers, technology companies and research organisations.
At the same time, the US government has increasingly treated advanced AI computing as a strategic technology.
Export restrictions are intended to limit China’s access to certain high-performance semiconductor products that could support advanced military and technological capabilities.
However, the development of overseas data centres has complicated enforcement.
A company can operate computing infrastructure in another country and provide remote access to customers across multiple jurisdictions.
This has increased scrutiny of cloud providers, semiconductor suppliers and the relationships between hardware manufacturers and their customers.
Nvidia’s reported move suggests that commercial relationships involving AI computing capacity are becoming an additional area of concern.
What Does This Mean for Nvidia’s Business?
Nvidia’s strategy creates a potential trade-off.
On one side, the company wants to maintain its position as a leading supplier of AI infrastructure to global cloud providers.
On the other, it must manage regulatory and geopolitical risks associated with the use of its technology.
Reducing exposure to Chinese customers could make some cloud partnerships more aligned with US policy concerns.
However, it could also limit the commercial opportunities available to Nvidia’s partners.
Overseas data centre operators that have invested heavily in Nvidia GPUs may need to find additional customers in the United States, Europe, India and other markets to maintain high infrastructure utilisation.
The effect on Nvidia’s own revenue remains unclear. The reported guidance does not establish that existing chip purchases or leasing contracts have been cancelled.
The Bigger Picture
The development reflects a broader change in the global AI industry.
Competition is no longer limited to which company can manufacture the most powerful processors or develop the most capable AI models. Access to computing infrastructure has become a strategic issue involving governments, cloud providers, chipmakers and international customers.
The traditional focus on controlling physical semiconductor exports may increasingly expand towards monitoring who can remotely access advanced computing resources.
For China, this could accelerate efforts to develop domestic AI chips and alternative computing infrastructure. For Nvidia, it introduces another layer of complexity in managing international demand while complying with US restrictions.
Looking Ahead
The next important question is whether Nvidia turns its reported customer-diversification guidance into a formal requirement and how widely cloud providers adopt it. The identities of affected providers, enforcement mechanisms and the treatment of future contracts will determine the policy’s practical impact.
For Chinese AI developers, the development could make overseas GPU access more complicated and expensive. For the wider industry, it signals that access to AI computing power is becoming increasingly influenced by geopolitical considerations, not just commercial demand.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
