Nvidia has told cloud computing clients they must ensure at least 51% of the services they sell go to non-Chinese companies, a move that could force global providers to rethink where they build and who they serve. The instruction, delivered verbally by sales teams and regional staff rather than in writing, extends worldwide and applies only to future contracts, according to people familiar with the matter. The timing is delicate: Nvidia shares have fallen more than 13% over the past month, even as they remain up 7% year-to-date, and the chipmaker has been overtaken by Apple as the world’s most valuable listed company.
The new rule lands at the center of the AI supply chain, where demand for Nvidia’s chips remains fierce and access to China is increasingly constrained. It also adds a fresh layer of pressure on cloud providers outside the US, which now face a hard commercial choice: recruit more non-Chinese customers or scale back their exposure to China to keep buying Nvidia hardware. Nvidia did not confirm or deny the report to AFP, and Reuters said it could not immediately verify it. Still, the market message was clear enough: Nvidia is drawing a sharper line around who gets access to its AI infrastructure.
The Financial Times, citing five people with knowledge of the matter, reported that Nvidia’s sales force and regional teams told cloud computing clients that a majority of their customers must be non-Chinese. The company had not imposed such a restriction before. The rule is said to apply globally, not just in Asia, and only to future contracts, meaning existing leases are not affected. That detail matters because it suggests Nvidia is not ripping up current business, but instead trying to shape the next round of deals as the AI market reorganizes around Washington’s export limits and Beijing’s counterpressure.
Nvidia’s own framing is about American infrastructure and global competition. In a statement to the Financial Times, the company said: “Datacenters and clouds built on American chips are the foundation that American model and application makers need to compete worldwide. Longstanding US policy encourages every cloud to build on American infrastructure and to win business everywhere they can.” In a separate statement carried by AFP via RTL Today, Nvidia said: “America cannot win the AI race unless America’s AI infrastructure wins worldwide — if the world’s clouds rely on foreign-made chips, the model and application layers will be foreign-built as well.” That is not a small rhetorical shift. It casts the rule as policy-aligned rather than China-specific.
The problem for cloud providers is practical, not theoretical. A Singapore-based cloud provider told the Financial Times it would be “difficult to find enough non-Chinese clients outside the US… If strictly implemented, it will definitely hurt business.” That warning gets to the heart of the issue: many cloud companies in Asia and beyond have built growth plans around regional demand, including Chinese customers. If Nvidia enforces the threshold aggressively, some buyers may have to change their customer mix before they can keep purchasing the most sought-after AI gear.
The backdrop is a market already strained by scarcity. The Financial Times reported that a rack of Nvidia B300 chips, which costs about $600,000 internationally, fetched Rmb18.5mn, or about $2.8mn, in China last month because of tight supply. That kind of price gap shows how limited access can distort the market. When a product is scarce, every new rule around eligibility matters more, because it changes not just who can buy, but who can compete. Nvidia’s move may therefore hit hardest where demand is strongest and supply is most constrained.
The company has already been tightening access in another way. The Financial Times reported Nvidia has more than halved the number of Asian customers authorized to buy its AI chips after creating a compliance “white list” this year. The new cloud rule fits that broader pattern. Rather than a one-off restriction, it looks like part of a larger compliance system meant to control who sits on the receiving end of Nvidia’s hardware. That would make sense in a world where US export policy, customer screening and geopolitical risk are increasingly tangled together.
For Nvidia, that also means a more complicated sales pitch. The company is still selling the same basic message: its chips are essential to the AI race. But now it is doing so with guardrails that go beyond simple buyer qualification. If cloud customers must show that at least 51% of their services go to non-Chinese companies, then access to Nvidia chips becomes linked to the business model of the buyer, not just its balance sheet or purchase order. That is a meaningful change in how the market is being segmented.
The exact mechanics are unclear. The Financial Times report does not specify whether the 51% threshold is measured by revenue, computing capacity or customer count, and it does not define how Nvidia classifies a Chinese company. Those gaps matter, because different definitions could produce very different outcomes. A provider with a broad international customer base might be fine under one test and fail under another. Until those details are clear, the rule functions more like a strategic constraint than a fully legible formula.
Still, the direction of travel is obvious. Cloud providers outside the US may need to grow faster in non-Chinese markets, or trim their China exposure, if they want continued access to Nvidia’s chips. That could alter sales strategies, data-center investment decisions and partner relationships across Asia and beyond. It also raises the stakes for a company already dealing with the reality that the world’s biggest AI hardware maker is no longer just selling silicon; it is helping define the commercial boundaries of the AI ecosystem.
For investors, the immediate read-through is not about one headline order but about the durability of Nvidia’s market power. The company remains central to AI infrastructure, yet its position is increasingly shaped by policy friction and customer filtering. That combination can support pricing power, but it can also complicate growth. With shares down more than 13% over the past month and still up 7% this year, Nvidia is in a zone where every strategic move is being judged not just as a sales tactic, but as a signal of how the AI market will be governed next.
Beijing, for its part, is signaling opposition. Chinese foreign ministry spokeswoman Mao Ning said: “As a matter of principle, we consistently hope that all parties can provide a fair, transparent and non-discriminatory business environment for companies from all countries, including Chinese companies.” That response suggests the policy tension is not going away. Nvidia may present the rule as support for American infrastructure, but from China’s perspective it looks like another barrier. The market consequence is a more fragmented AI economy, where access to the best chips depends not only on demand, but on where a cloud provider’s customers come from.
