AI-assisted research analysis by FutureKnowledge. Sources checked 3 November 2026. This is not investment advice.
Reviewed by FutureKnowledge: figures, dates and attributions were checked against the linked sources before publication.
Nvidia has told its cloud and data-centre partners that at least 51% of their services on new contracts must go to non-Chinese customers, effectively capping Chinese business at 49%, the Financial Times reported on 3 November, citing five people familiar with the matter. The rule, reported by Investing.com and TNW, targets a key route for Chinese AI developers: renting chips from cloud providers abroad.
Nvidia has neither confirmed nor denied the report. In a statement to AFP, carried by Malay Mail, the company said America “cannot win the AI race” unless its AI infrastructure wins worldwide. Reuters said it could not immediately verify the report, according to The Standard.
What Nvidia reportedly told cloud partners
The main points, as reported by outlets citing the FT:
- The ratio: cloud computing clients must sell at least 51% of their services to non-Chinese companies, according to AFP via Malay Mail.
- Scope: the requirement applies to future contracts only. Existing agreements are not affected, TNW reported.
- Delivery: the instruction was given verbally through Nvidia’s sales force and regional teams, not in writing, and Nvidia had no such restriction before, Investing.com said.
- Reach: the Seoul Economic Daily, also citing the FT, said the rule applies globally, not only in Asia.
No companies were named in the reports. Chinese foreign ministry spokeswoman Mao Ning said she was not aware of the report and that Beijing hopes all parties provide a “fair, transparent and non-discriminatory” business environment, AFP reported.
Nvidia shares were higher on the day. They were up about 1.4% before the US open, according to GuruFocus via Yahoo Finance, and the market snapshot on Investing.com’s report showed a gain of about 3%. The stock had been caught in the chip sell-off that pushed the Nasdaq 100 into a correction.
Why Nvidia is tightening China access through the cloud
The cloud loophole
The US has restricted sales of Nvidia’s most advanced chips to China since 2022. Many Chinese AI firms have responded by leasing capacity from data centres abroad, including for model training, Investing.com noted. Renting computing power does not export the chip itself, and Washington has tried to close this route for years.
Compliance pressure
Nvidia hardware has still reached China, partly through smuggling. Earlier this year several senior employees at server maker Super Micro were charged over smuggling about $2.5 billion worth of Nvidia AI servers into China, TNW and Investing.com reported. The Seoul Economic Daily said Nvidia had already more than halved the number of Asian customers approved to buy its AI chips, using a whitelist drawn up this year after tighter compliance reviews.
Strong Chinese demand
The price gap shows why the issue matters. A rack of Nvidia B300 chips that sells for about $600,000 internationally traded in China last month for 18.5 million yuan, about $2.8 million, the Seoul Economic Daily reported, citing the FT. The B300 cannot be sold to China under US export controls. Demand for legal chips is also high, as covered in our report on Nvidia’s $35 billion of China chip orders.
What the Nvidia China cap means for Nvidia’s revenue
Nvidia’s own guidance already treats China as close to zero for its main data-centre business. In its second-quarter fiscal 2027 results, released on 26 August, the company reported revenue of $96.2 billion and data-centre revenue of $89.0 billion, up 117% from a year earlier. Its third-quarter outlook of $108.0 billion, plus or minus 2%, assumes no China data-centre compute revenue.
So the reported rule does not cut any China revenue that Nvidia is counting on. The open question is indirect: whether cloud providers outside the US can fill their capacity without relying heavily on Chinese renters. If some cannot, they may slow new orders.
How it affects businesses and India
The biggest effect falls on non-US cloud providers that built capacity for Chinese AI developers. A Singapore-based provider told the FT it is hard to find enough non-Chinese customers outside the US, and data centres that fail to attract them may have to scale back, the Seoul Economic Daily reported.
For India, no direct link has been reported. If the rule is global, as the reports suggest, Indian data-centre and GPU-cloud operators buying Nvidia systems for new contracts would in principle face the same customer-mix test. Our assessment, not a reported finding, is that the constraint would be smaller for operators that already sell mainly to domestic and Western clients, though competition for those clients could rise.
Who is more or less exposed
- More exposed: cloud and colocation providers in Asia whose GPU clusters rely on Chinese renters; Chinese AI developers that train models on overseas cloud capacity.
- Less exposed: US hyperscalers and AI labs whose capacity serves mostly non-Chinese customers; Chinese domestic chip makers, which may gain if overseas access narrows.
Scenarios to watch
The table below sets out conditional scenarios, not predictions.
| If this happens | Possible consequence | Signal to watch |
|---|---|---|
| Nvidia confirms the rule or puts it in writing | Cloud contracts are likely to add customer-origin checks and audit clauses | Nvidia statements or partner disclosures |
| Washington turns similar limits into formal regulation | The cap would apply to all US chip suppliers, not just Nvidia | Commerce Department notices on cloud access |
| Asian cloud providers struggle to find non-Chinese customers | Some capacity expansions could slow, affecting GPU orders | Utilisation and capex comments from regional providers |
| Beijing responds with its own measures | Faster push toward domestic AI chips and data centres | Chinese ministry statements and procurement rules |
What to watch next
- Whether Nvidia addresses the report in its third-quarter fiscal 2027 results and whether it keeps assuming no China data-centre compute revenue.
- Comments from Asian and Middle Eastern cloud operators on customer mix.
- Any formal US rule on remote access to advanced chips, and any Chinese response beyond the foreign ministry’s comment.
- Investor mood on AI hardware after recent deals such as Marvell’s $115 billion Lumentum acquisition.
Bottom line
If the FT report is accurate, Nvidia is now acting as a gatekeeper on who can rent its chips, not just who can buy them. The 49% cap on Chinese business does not hit revenue Nvidia is counting on, because its guidance already excludes China data-centre compute. The pressure falls instead on overseas cloud providers that built capacity for Chinese AI developers, and on those developers. Because the rule is verbal, unconfirmed and limited to new contracts, its real effect will show up in contracts and orders over the coming quarters.
Method note: we compared reports from Investing.com, TNW, AFP (via Malay Mail), the Seoul Economic Daily, GuruFocus (via Yahoo Finance) and Reuters (via The Standard), all citing the Financial Times, with Nvidia’s second-quarter fiscal 2027 results release. Nvidia has not confirmed the rule. Reported share moves differed by timing; we give both. Details on the customer whitelist and B300 prices come from the Seoul Economic Daily’s account of the FT report.