Nvidia has instructed cloud computing clients to sell at least 51% of their services to non-Chinese customers, according to five people familiar with the matter cited by the Financial Times. The directive applies only to new contracts and was communicated verbally by sales teams rather than through formal documentation. Existing leasing agreements remain unaffected by the requirement, which targets overseas cloud providers renting Nvidia chips to Chinese AI firms.
What Nvidia Reportedly Told Its Cloud Partners
The instruction requires data center operators to ensure that a majority of their service revenue or capacity goes to customers outside China. Two sources told the FT that this mandate covers Nvidia clients globally, not just those based in Asia. The chipmaker’s regional teams and sales force delivered the message orally over recent months without issuing written compliance guidelines. No official document outlines how the 51% threshold is measured or verified. Nvidia neither confirmed nor denied the report when contacted for comment.
Verbal Delivery Creates Compliance Uncertainty
The absence of written guidance leaves cloud providers interpreting an unwritten rule with significant business consequences. One Singapore-based provider told the FT that finding enough non-Chinese clients outside the United States is difficult. Strict implementation would hurt business for smaller GPU clouds that built models around Chinese demand. Reports disagree on whether the 51% floor measures revenue, customer count, or computing capacity. Nvidia has not explained its verification methodology or enforcement timeline for partners signing new deals.
Earnings Context and Excluded China Revenue
Nvidia reports third-quarter results on November 17, 2026, with consensus revenue estimates near $108.98 billion, according to Yahoo Finance. The company’s own guidance assumes zero data center compute revenue from China for the current quarter. This forecast was set before the verbal instruction became widely known among partners. Analysts project sequential revenue growth of approximately 12.3% from the prior quarter’s $96.2 billion. Gross margin guidance stands at 74%, plus or minus 50 basis points, despite rising memory costs.
Pre-Emptive Move Amid Export Enforcement Pressure
Nvidia implemented the requirement over the summer as a pre-emptive effort to reduce criticism from U.S. lawmakers, two sources told the FT. Washington is targeting both chip smuggling intermediaries and overseas leasing arrangements that give Chinese firms access to advanced semiconductors. U.S. export rules already bar direct sales of Nvidia’s most advanced chips, such as the B300, inside China. Chinese AI companies have responded by renting compute power from clouds in Singapore and Southeast Asia. The 51% floor aims to close that remote-access pathway without restricting existing leases.
Capacity Constraints Drive Allocation Logic
Nvidia CEO Jensen Huang stated in October that demand remains extraordinary and capacity is the primary constraint. The company lined up six suppliers that month to expand production capability. When supply is constrained, chips not allocated to one partner go to the next buyer in line. Capacity previously serving Chinese renters now flows to hyperscalers and governments waiting for allocation. Smaller cloud providers bear the adjustment cost while Nvidia continues selling every chip it manufactures. The company treats partner data centers as American infrastructure that should win global business.
Do the new rules apply to existing contracts? No, the 51% non-China service requirement affects only future contracts signed after the verbal instruction began circulating this summer. Existing leasing agreements remain valid and are not subject to retroactive compliance demands. This grandfathering protects current revenue streams for partners already serving Chinese clients.
What exactly did Nvidia instruct its cloud partners to do? Nvidia told clients they must sell at least 51% of their services to non-Chinese customers under new agreements. The instruction was delivered verbally by sales and regional teams without written documentation. Partners must now demonstrate majority non-China exposure to secure new chip allocations.
Were official documents issued with the new requirements? No, five sources told the FT that the directive was communicated orally rather than through formal compliance paperwork. Nvidia has not published written guidelines defining measurement standards or audit procedures. This lack of documentation creates interpretation risk for partners negotiating new deals.
Does the 51% floor apply worldwide or only to Asia? Two sources confirmed the requirement covers Nvidia clients globally, not just data centers in Asia. Partners in Europe, the Middle East, and other regions face the same allocation conditions as Asian providers. The geographic scope reflects U.S. policy targeting overseas leasing regardless of partner location.
How does this relate to Nvidia’s upcoming earnings? Nvidia reports November 17 with guidance already excluding China data center compute revenue. The verbal instruction reinforces that exclusion operationally ahead of the print. Investors will watch whether management addresses partner compliance or allocation shifts during the earnings call.
Nvidia’s verbal directive reshapes cloud partner economics without altering its own revenue trajectory ahead of the November 17 earnings report. The unwritten rule functions as an allocation mechanism that redirects constrained capacity away from Chinese end-users while preserving existing lease structures.
