Nvidia is ordering its data centre partners to reduce exposure to Chinese customers, the Financial Times reported on 3 November. The FT said the chipmaker has told cloud computing providers that “a majority of their clients must be non-Chinese”. Its opening adds that they “must now sell at least 51 per cent of”, a sentence that continues behind the paywall.
Nvidia’s own filing shows what ties it to such partners. Its latest 10-Q lists $36 billion of cloud service commitments to AI cloud partners and up to $3.5 billion of guarantees on their data centre leases: about $39.5 billion in all, or 3.2 times its first-half revenue from China-headquartered customers, by our arithmetic.
Business Pill · Concentration
An evergreen Business Engineer explainer on concentration: when a few big pieces make the total. It teaches the general idea only and does not discuss Nvidia, the Financial Times report or any company in this story.
The key insight: The FT reports that Nvidia has told cloud computing providers that a majority of their clients must be non-Chinese. By our arithmetic from its 10-Q, Nvidia has about $39.5 billion of commitments and guarantees tied to AI cloud partners, 3.2 times its first-half revenue of $12.4 billion from China-headquartered customers.
Update, 3 Nov: Bloomberg’s account says the 51 per cent applies to services sold.Bloomberg reported at 05:39 UTC that, according to the FT, Nvidia told its cloud computing customers to make sure they sell at least 51% of their services to companies outside China. Bloomberg said the FT cited unidentified people. Reuters, which also carried the FT report, said it could not immediately verify it. The parts of the Bloomberg and Reuters reports we could read do not name the providers, give a deadline or say how the rule would be enforced, and Nvidia’s newsroom listed no statement on it when we checked at about 06:05 UTC.
The report, by Zijing Wu, Demetri Sevastopulo and Eleanor Olcott, was published at 05:00 UTC. Its opening says Nvidia “is ordering its data centre partners to reduce exposure to Chinese customers in a move that further restrains China’s ability to keep up with US rivals”.
It says the instruction came “in recent months” and was given to “cloud computing clients”. We could read the headline, the summary line and the opening only. The sentence carrying the 51 per cent figure is cut off at that point, so what the 51 per cent is measured on is not in the text we read.
If the floor is 51 per cent, the complement is 49 per cent: by that arithmetic, no more than 49 per cent of whatever the rule measures could be Chinese.
Nvidia’s cloud service commitments to AI cloud partners ($36.0 billion) and its maximum gross exposure under guarantees of their data centre leases ($3.5 billion), against its revenue from customers headquartered in China, including Hong Kong ($12.4 billion in the first half of fiscal 2027; $7.9 billion in the quarter to July). The $39.5 billion total, the 3.2x ratio, the 8.2% share and the 20% escrow share are our arithmetic; commitments and guarantees are balances, China revenue is a flow. Source: NVIDIA Form 10-Q for the quarter ended 26 July 2026.
The Money Behind the Partners
Nvidia’s Form 10-Q for the quarter to 26 July 2026 describes a business model introduced that quarter with “certain select AI cloud partners”. Under it, “AI clouds procure our data center infrastructure products and we commit to cloud service agreements”.
Those commitments, “typically six years in duration”, “totaled $36 billion as of July 26, 2026”. The filing says the AI clouds can stop providing that capacity to Nvidia and sell it to third-party customers “at more advantageous rates”.
Nvidia has also guaranteed select AI cloud partners’ data centre lease obligations in case of default. The maximum gross exposure is $3,529 million, and the partners have placed $712 million in escrow, about 20 per cent of that exposure, by our arithmetic.
Where China Sits in Nvidia’s Revenue
The same 10-Q counts revenue by where direct customers are headquartered. China, including Hong Kong, was $7,880 million of $96,221 million in the quarter to July, or 8.2 per cent, by our arithmetic. For the first half of fiscal 2027 it was $12,430 million of $177,837 million, or 7.0 per cent.
The filing adds a caveat that bears on cloud partners: “The end customer and shipping location may be different from our direct customers’ headquarters location”. It also says Nvidia was “effectively foreclosed from competing in China’s data center computing/compute market” at the end of that quarter.
The Rules Nvidia Cites
On compliance, the 10-Q says that, “as we do not have physical control of our products after sale, we must also rely on the compliance programs of our customers and partners”.
It names a bill aimed at cloud access. Congress is considering the Remote Access Security Act, “which could prohibit the provision of cloud services to any company with an ultimate parent headquartered in China”. The filing says that if enacted it “could impose new restrictions on cloud service providers and OEMs”.
What Came Before
In July the FT reported that Nvidia had cut its list of approved Asian customers after a new vetting process, according to Techzine’s account of that report. More than half of existing customers reportedly failed the first round, and so-called neocloud providers were particularly affected.
Techzine said the stricter criteria covered customers in Singapore, Malaysia and Japan, among other countries, and that Nvidia now inspects customers’ data centres, reviews contracts and speaks with end users. Nvidia told the FT then that export compliance is its top priority.
As we read the 10-Q, Nvidia is a supplier to its AI cloud partners, a buyer of their capacity and a guarantor of some of their leases. The part of the FT report we could read does not say whether any of those terms carry the rule.
As we read the filing, the reported rule sits where Nvidia says its control is weakest: after the sale, where it says it must “rely on the compliance programs of our customers and partners”.
NVIDIA Form 10-Q, quarter ended 26 July 2026
“as we do not have physical control of our products after sale, we must also rely on the compliance programs of our customers and partners”
Three Implications
FOR CLOUD PROVIDERS The FT says providers were told they must sell at least 51 per cent of a measure its readable text does not name; Bloomberg’s account of the report says it is their services. By that arithmetic, up to 49 per cent of those services could still go to Chinese companies.
FOR NVIDIA’S REVENUE Customers headquartered in China were 8.2 per cent of Nvidia’s revenue in the quarter to July, by our arithmetic from the 10-Q, which says Nvidia was “effectively foreclosed from competing in China’s data center computing/compute market”.
FOR WASHINGTON The 10-Q says Congress is considering the Remote Access Security Act, which could prohibit cloud services to any company with an ultimate parent headquartered in China. The readable FT text does not link the rule to that bill.
The framework describes Layer 9, governance, this way: “Capability gating inside labs, procurement frameworks inside enterprises, export controls between nation-states, litigation over contested deployments. No longer a passive layer; an active priced variable.”
As we read it, the FT report describes that layer reaching Layer 5, compute capacity, through a supplier’s terms with its cloud clients rather than through a government rule. The framework was published in June 2026 and does not discuss this report.
The parts of the FT, Bloomberg and Reuters reports we could read do not name the cloud providers, give a deadline or say how Nvidia would enforce the rule. Bloomberg’s account of the FT report says the 51 per cent applies to the services providers sell (see the update above). None of them gives Nvidia’s reasons or a company comment.
The 10-Q’s AI cloud partners may or may not be the clients the FT describes, and the filing does not link its commitments or guarantees to any China rule. Our $39.5 billion adds balances; China revenue is a flow over six months.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
The FT reports that Nvidia has told cloud providers that most of their clients must be non-Chinese, with a floor of at least 51 per cent. Nvidia has about $39.5 billion of commitments and guarantees tied to AI cloud partners, by our arithmetic, against $12.4 billion of China revenue in the first half.
A note on sourcing. The Financial Times report is paywalled; we read only its headline, summary line and opening, and quote those. Figures on Nvidia’s AI cloud partners and China revenue are from its Form 10-Q for the quarter ended 26 July 2026. The July report is quoted as Techzine described it. The $39.5 billion total and all ratios are our arithmetic.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
