Nvidia has told its cloud computing partners that at least 51 percent of their services must go to non-Chinese customers, the Financial Times reported on November 3, citing people familiar with the instruction. The rule has not been put in writing. Nvidia's sales teams have passed it along verbally since the summer, and the company has neither confirmed nor denied it.
The requirement applies to future business rather than lease agreements already signed. What it actually measures remains unclear. Some accounts describe a threshold on revenue; others note that Nvidia has not said whether the 51 percent refers to revenue, computing capacity or the number of customers, nor how it defines a Chinese company in the first place.
Who it hits
The providers most exposed are the smaller specialist operators, the so-called neoclouds, based in Singapore, Malaysia and Japan. Many of them built their businesses partly on Chinese demand after tighter US export rules made it harder for firms in China to buy advanced chips directly. One Singapore-based provider told the FT that finding enough non-Chinese customers outside the United States is genuinely difficult, and that some operators may have to scale back data-center plans if they cannot.
In a statement, Nvidia said "longstanding US policy encourages every cloud to build on American infrastructure," but stopped short of addressing the specific instruction.
A tightening pattern
The move fits a year of steadily narrowing access. Over the summer Nvidia cut its list of approved Asian buyers by more than half, and the US Commerce Department issued guidance in May aimed at curbing the flow of chips to China through third countries. Washington has been pressing on the same point from several directions this autumn, from the sanctions it imposed over alleged trade-secret theft to the order that DeepSeek be purged from federal systems.
There is a tension running underneath all of it. Chinese firms remain eager buyers, as the $35 billion in Nvidia chips ordered by ByteDance, Alibaba and Tencent made plain, and the intermediaries in Asia have every commercial reason to serve them. Nvidia, caught between that demand and US policy, appears to be managing the risk through its partners rather than waiting for the next formal rule.
Because the instruction is informal and unverified by the company, its exact force is hard to judge. But if the reporting holds, it marks a notable shift. Export controls have mostly governed what Nvidia can sell and to whom. A revenue test imposed on its customers' customers reaches a step further down the chain, and asks the cloud operators themselves to police who ultimately uses the hardware.
Commentarii · 0