Nvidia Reportedly Sets 51% Non-Chinese Revenue Rule for Global Cloud Partners

A majority revenue rule could narrow China's access to AI computing

Nvidia is reportedly requiring its cloud computing partners worldwide to earn at least 51% of their service revenue from customers outside China, a change that could limit Chinese technology companies' access to powerful artificial intelligence chips through overseas data centers.

The requirement, described in reporting dated November 3, 2026, would apply to future contracts rather than existing agreements. Five people familiar with the matter described the restriction. Nvidia neither confirmed nor denied the reported policy in its public response.

The proposed threshold would reach beyond the direct sale of chips. Chinese companies rent computing capacity from cloud providers abroad to carry out demanding AI work, including training large models. Those arrangements give companies access to processors without requiring them to buy and install the hardware themselves.

Under the reported revenue rule, a provider would need to earn more from customers outside China than from Chinese customers. For every $100 in service revenue, at least $51 would have to come from non-Chinese customers, leaving no more than $49 from Chinese customers.

The restriction reportedly applies globally, including to data centers outside Asia. Its reach would therefore depend on the customer mix of a cloud business, extending the commercial pressure beyond the country where its servers operate.

What would change for cloud providers?

Nvidia's sales and regional teams reportedly communicated the requirement verbally, without issuing written guidance. The company previously had no comparable requirement, according to the reporting.

The public description leaves several practical details unresolved, including how Nvidia would classify a Chinese customer, which services would count toward revenue, how compliance would be measured and when the requirement would begin.

A cloud provider with a large Chinese customer base could face several choices when arranging future business: attract more customers from other countries, reduce the proportion of revenue coming from Chinese companies or reconsider how much new capacity to build around Nvidia hardware.

The reported protection for existing agreements matters for both providers and their customers. Current leases would continue under their existing terms, while new contracts would face the requirement. Companies planning additional AI projects could encounter different commercial conditions from those attached to work already contracted.

A revenue threshold also differs from a limit on the number of customers. A provider with many small customers outside China could still receive most of its income from a few large Chinese companies. Under the reported rule, the share of revenue would be the relevant measure.

Why overseas cloud access matters

Chinese technology companies have used overseas cloud leases for intensive AI applications, including the training of large models. Training is the process in which a model learns patterns from large amounts of data by repeatedly performing calculations and adjusting its internal settings.

Specialist AI chips can carry out many calculations at once. Data centers combine those processors with memory, networking equipment, electricity and cooling so they can work together on much larger tasks than a typical personal computer can handle.

Cloud providers sell access to that equipment as a service. A customer can rent computing capacity instead of paying to build and operate an entire facility. For a company developing AI, access to rented processors can therefore be an important part of its ability to start or expand a project.

The reported Nvidia requirement would affect that rental route by changing the conditions facing the provider. A Chinese company seeking a new lease could find that a cloud operator's ability to accept the business depends partly on how much revenue it earns from customers elsewhere.

Nvidia argues American infrastructure must compete worldwide

Nvidia's statement focused on the connection between American computing hardware and the companies that develop AI models and applications. The company presented the international use of American chips as a foundation for American participation in the wider AI market.

Nvidia said in its statement:

Data centers and clouds built on American chips are the foundation that American model and application makers need to compete worldwide.

The company also described international cloud business as consistent with longstanding American policy:

Longstanding U.S. policy encourages every cloud to build on American infrastructure and to win business everywhere they can.

Nvidia argued that American AI infrastructure needs to succeed worldwide for the United States to win the AI race. It also warned that reliance on chips made elsewhere would encourage the development of foreign models and applications.

That position highlights the commercial stakes for Nvidia. Its processors sit at the hardware layer of an industry that also includes cloud services, AI models and applications used by businesses and consumers. Restrictions affecting one layer can influence purchasing and development decisions across the others.

China calls for fair treatment as Washington considers controls

Chinese foreign ministry spokeswoman Mao Ning said she was unaware of the report when asked about it. She responded with a statement of China's general position on the treatment of companies operating internationally.

Mao, speaking in her role as ministry spokeswoman, said:

As a matter of principle, we consistently hope that all parties can provide a fair, transparent and nondiscriminatory business environment for companies from all countries, including Chinese companies.

Washington is reportedly considering tighter controls on Chinese access to advanced American chips through overseas cloud providers. Nvidia has also reportedly reduced its approved customer base in Asia following stricter compliance reviews.

The reported company requirement and Washington's consideration of further controls are separate developments. Nvidia would be setting conditions for its commercial partners, while any new American government controls would depend on the measures Washington adopts.

China and the United States are competing for leadership in AI, and both have accused the other in recent months of stealing AI technologies and putting security at risk. The cloud access issue brings that competition into decisions about who can rent computing power and under what conditions.

China's chip market is already shifting

IDC figures cited in the reporting put Nvidia's share of China's AI accelerator market at 34% in the first half of 2026, compared with Huawei's 28%. Nvidia's shipments fell 44% from a year earlier to 678,000 units.

The two market shares placed Nvidia six percentage points ahead of Huawei during that period. Together, they accounted for 62% of the market measured by IDC, showing the scale of both companies' positions as Chinese buyers sought processors for AI work.

UCLA professor Christopher Tang described Nvidia's share as having effectively fallen to zero in his assessment of the more recent situation, while IDC's 34% figure covered the first half of 2026. IDC did not specify which Nvidia chips its estimate included or whether it counted restricted B200 and B300 chips reportedly smuggled into China.

Tang identified high-bandwidth memory, or HBM, as China's biggest immediate bottleneck. HBM is memory designed to move large amounts of data rapidly to and from a processor. Powerful AI chips need a fast supply of data to keep their calculating capacity busy.

American restrictions on HBM and chipmaking equipment are limiting China's production of advanced AI chips. Tang also warned that restrictions could accelerate efforts to develop domestic alternatives. Chinese companies therefore face pressure on both access to foreign computing services and the equipment needed to expand production at home.

How the developments fit together

The developments span market changes earlier in 2026 and the subsequent report about cloud contracts. The main points in that sequence are:

• First half of 2026: IDC put Nvidia at 34% of China's AI accelerator market and Huawei at 28%, with Nvidia shipments down 44% from a year earlier.
• October 2026: IDC's figures were reported, documenting the earlier market position.
• November 3, 2026: Reporting described a global requirement for Nvidia cloud partners to earn at least 51% of service revenue from non-Chinese customers.
• Future contracts: The reported requirement would apply to new agreements, while existing leases would remain unaffected.

For Chinese AI developers, the immediate business issue is access to additional computing capacity. For cloud operators, it is whether their future customer revenue would satisfy Nvidia's reported threshold. For Nvidia, it is how to maintain the international reach of its hardware while operating amid tighter scrutiny of Chinese access.

The combination of a commercial revenue requirement, possible government controls and China's push for domestic chips could shape different parts of the same investment decision: which processors to use, where to rent them and which customers a provider can serve.

Key Points

• Nvidia reportedly wants cloud partners worldwide to earn at least 51% of service revenue from non-Chinese customers.
• The requirement would cover future contracts, leaving existing agreements unchanged.
• Nvidia neither confirmed nor denied the report and argued that American AI infrastructure needs a global market.
• China's foreign ministry called for a fair, transparent and nondiscriminatory business environment.
• Washington is considering tighter controls on Chinese access to advanced American chips through overseas clouds.
• IDC put Nvidia's China AI accelerator share at 34% in the first half of 2026, ahead of Huawei's 28%, while Nvidia shipments fell 44%.