SANTA CLARA, 29 AUG 2026 — Less than two months after launching it, Nvidia has paused parts of a programme that lent credit support to AI cloud companies buying its chips. The reported reason is not a lack of demand, but internal warnings from staff that the structure could attract antitrust scrutiny.
The pause landed days after the company reported quarterly revenue of US$96.2bn, up 106 per cent year on year.
What the programme did
The scheme offered credit support to AI cloud companies so they could afford Nvidia hardware. In return Nvidia earned revenue on the hardware sale and then took a share of the cloud revenue those customers generated from Nvidia-powered capacity.
The structure is unusual: a supplier financing its customers' purchases of its own product, then taking a share of the income those purchases produced. Nvidia confirmed on 27 August that it had paused parts of the programme.
The concern came from inside
Reporting says Nvidia employees told current and potential customers they were worried the programme could draw antitrust scrutiny, citing how much control the company could come to exert over customers' businesses.
The source of the concern is the interesting detail. Companies often withdraw products after a regulator asks a question. It is much rarer for one to pull back because its own sales staff flagged the risk to customers, before any regulator was reported to have opened anything.
It suggests the control the structure created was obvious to the people running it, and obvious enough that they warned customers rather than only their own managers.
Why control, rather than price, is the issue
A supplier discounting for volume is routine. One that finances the purchase and then takes a revenue share is doing something else, acquiring an interest in the customer's pricing, its product roadmap and its capacity planning.
For a company with Nvidia's position in AI accelerators, that reaches into competition law territory in at least two ways. It can foreclose rivals, because a customer whose economics are entangled with Nvidia has less reason to buy AMD or build its own silicon. And it can create the conditions for coordination, because a supplier with revenue-share visibility into many competing clouds sees things a supplier selling hardware does not.
None of that is a finding of wrongdoing. It is the kind of argument a competition authority would test, and the one Nvidia's employees appear to have anticipated.
The circularity problem this belongs to
The pause is one move within a pattern this publication has been tracking for months, and it is the first move in the other direction.
Nvidia's US$105bn for OpenAI's Ohio campus is a guarantee rather than a cheque — a promise about resale value that lets someone else raise the money. Broadcom used the same instrument when it guaranteed the resale value of its own chips to get a US$35bn loan priced. In each case a chip vendor's balance sheet underwrites the purchase of that vendor's chips.
The concern investors have raised is that arrangements like these inflate apparent demand. If a supplier funds the buyer, the resulting order tells you less about end-market appetite than an unfinanced order would. That does not make the demand fake. It does make it harder to measure, which is a real problem when the measurement is what a valuation rests on.
What this means for buyers in this region
Vendor financing is not an American peculiarity, and Southeast Asian operators are being offered versions of it as the regional buildout accelerates.
The structure that was paused had two separable parts, and only one of them is unusual. Credit support for a hardware purchase is ordinary commercial practice, offered by every large equipment vendor in every capital-intensive industry. Taking a share of the revenue the equipment subsequently earns is not. It converts a supplier into something closer to a partner with an interest in the operating business, and it does so without any of the governance a partner would normally accept.
An operator in Johor or Batam evaluating a similar offer should separate those two things in the term sheet, because they carry different consequences. Financing affects the cost of capital. Revenue sharing affects who has a claim on the business, and it is far harder to unwind three years in than it looks at signing.
What the quarter actually showed
The results themselves were not ambiguous. Revenue of US$96.2bn for the quarter ended 26 July beat both Nvidia's own guidance and analyst estimates. Gross margin widened to 75 per cent from 72.4 per cent, operating income rose 124 per cent to US$63.7bn and adjusted earnings per share rose 120 per cent to US$2.22.
Margin expansion at that revenue scale points to pricing power rather than volume growth alone. A company that could not sell its output would not be posting it.
Which is the point about the financing programme. Nvidia did not need it to move product this quarter. Since Nvidia did not need the programme to move product this quarter, the decision to pause was cheap. The more interesting question is not why it stopped, but why it was started.
What to watch
Nvidia has paused parts of the programme, not cancelled it, and no detail has been published on which parts or for how long. A quiet resumption in a modified form is as likely as a withdrawal.
Two things would clarify the situation. First, whether any competition authority confirms an inquiry, which would turn an internal worry into an external process. Second, whether the guarantee structures continue. If residual-value guarantees are still written while revenue-sharing is paused, Nvidia has drawn a line between underwriting an asset and taking a share of a customer's income. That line, rather than the pause, would be the durable decision.
