AI News
NVIDIA Tells Cloud Partners to Cap Chinese Customers at 49% of New Contracts, FT Reports
NVIDIA has told its data-centre and cloud partners that Chinese customers should make up no more than 49% of business under future contracts, the Financial Times reported on November 3, citing five people familiar with the matter. Under the guidance, cloud providers must sell at least 51% of their services to companies that are not Chinese, which caps the Chinese share at 49%.
The rule, as reported, applies only to future contracts. Existing computing-rental agreements are untouched, and the guidance covers partners worldwide, not only in Asia. The Financial Times said the instruction has been passed on verbally by NVIDIA's sales and regional teams since the summer, without being put in writing, and that it is the first time NVIDIA has imposed this kind of sales restriction on its partners.
What is confirmed and what is not
Keep the status plain. What is confirmed is the reporting itself: the Financial Times reported the guidance, citing five people familiar with the matter, and Reuters carried the report while saying it could not independently verify it. What is not confirmed is the policy itself. NVIDIA neither confirmed nor denied the report when asked by AFP, and the reports give no effective date and no enforcement mechanism for the 51% threshold.
A separate report by The Information on the same day, citing three sources, said the guidance is not currently mandatory or strictly enforced, and it remains unclear whether NVIDIA will make it binding. Another open question is how the ratio would be calculated: whether it is judged contract by contract, or against a provider's overall revenue or service volume, changes which providers would actually be affected.
What NVIDIA said instead
NVIDIA's statement to AFP did not mention the 51% figure or China directly. The company said data centres and clouds built on American chips are the foundation American model and application makers need to compete worldwide, and stressed the importance of United States AI infrastructure staying competitive in global markets. China, for its part, called for non-discriminatory treatment, and its foreign ministry said it was not aware of the reported guidance.
The context behind the guidance
The reported cap did not come out of nowhere. The United States Commerce Department issued related guidance in May, with Blackwell chips among the products covered, and Reuters reported that NVIDIA had already cut its list of approved Asian customers by more than 50% in July. The reported 49% cap extends the same logic from which chips can be sold to whom rented computing capacity can be sold to.
China remains a meaningful but minority share of NVIDIA's business. According to NVIDIA's Form 10-Q for the second quarter of fiscal 2027, revenue from China, including Hong Kong, for the three months ended July 26, 2026 was $7.88 billion, about 8.2% of total revenue of $96.221 billion. NVIDIA's forecast for the third quarter of fiscal 2027 offers no estimate of data-centre computing revenue in China, and the company projects revenue of $108 billion, plus or minus 2%, without quantifying any effect on partners' bookings.
Who feels it first
The providers most exposed are those whose business centres on leasing dedicated GPU clusters to a single Chinese company, since a 49% cap applied contract by contract would force them to rethink their structure. Two large usage contracts disclosed by Japan's Data Section alone total about $1.297 billion in usage fees over three years, which shows the scale of the deals now sitting under the reported rule.
What to watch next is whether NVIDIA makes the guidance binding, how the ratio is measured, and what the company says on its third-quarter fiscal 2027 earnings call on November 17. Until then, the 49% figure is a reported instruction from unnamed sources, not a confirmed NVIDIA policy, and it should not be quoted as one.