Vendor financing is ordinary in capital goods. It stops being ordinary when it becomes hard to tell customer-funded demand from supplier-arranged demand. A framework for reading Wednesday's Q2 FY2027 results, and the four disclosures that would settle it.
Nvidia reports second-quarter fiscal 2027 results after the close on Wednesday 26 August 2026. Management has guided to roughly $91 billion in revenue for the quarter.
Every preview will ask the same question: did demand hold. That question has been answered the same way for eight consecutive quarters, and it is not the interesting one. The interesting question is the one the market itself asked on 10 August.
That day Nvidia announced it was partnering with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to establish financing platforms intended to mobilise more than $500 billion of third-party capital for AI compute infrastructure. The stock fell about 2.5%, taking roughly $130 billion off the company's market capitalisation, and briefly traded down 3% during the session.
Sit with what that reaction means. This is a company whose shares have risen for three years on news of more AI spending. It fell on news of more AI spending that it had helped arrange. The reaction was the information.
It is worth being exact here, because the shorthand circulating around this announcement is misleading in a way that matters.
Nvidia is not lending $500 billion. The structure, per the company's own announcement, mobilises third-party capital — money from asset managers and their investors — to finance AI data-centre buildouts. The stated intent is to treat compute infrastructure as a financeable asset class, borrowed against in roughly the way commercial real estate, toll roads or aircraft fleets are. The financing sits off Nvidia's balance sheet by design.
That is a materially weaker version of the concern than 'Nvidia is funding its own customers,' and anyone arguing the strong version should be held to the structure rather than the headline.
But it does not dissolve the concern either, and the reason is worth spelling out. Whoever supplies the capital, the chipmaker is now an active participant in solving its customers' funding problem — and funding, not silicon supply, is what analysts have identified as the binding constraint on the buildout. A supplier that removes the constraint on its own demand has changed its relationship to that demand, even if no dollar of its own is at risk. The question is no longer whether the demand is real. It is what the demand is a function of.
The framework readers need here is older than the AI trade and considerably less exotic than the discourse suggests.
Capital-goods industries have always financed their customers. Commercial aircraft are sold with manufacturer-arranged financing, guarantees and residual-value support. Telecom equipment vendors extended credit to carriers for decades. Agricultural and construction equipment makers run finance arms that are, in accounting terms, substantial lenders. In every case the logic is the same: the product costs more than the customer can pay from cash flow, the useful life is long, and the manufacturer knows the collateral better than a bank does.
None of this is improper. It is not aggressive accounting, it is not a red flag on its own, and treating it as inherently suspicious would mean treating most of industrial history as suspicious.
What it does do is make one specific thing harder to see: whether reported revenue reflects demand a customer chose to fund from its own resources, or demand that exists because the supplier arranged the funding. Both are real revenue. They are not equally durable, because the second depends on the financing continuing to be available.
Supplier-arranged demand becomes a problem under conditions that are identifiable in advance.
The historical precedents are specific rather than vague. Telecom-equipment vendor financing in the late 1990s produced several years of reported growth that looked like end demand and turned out to be the vendors' own credit extension — and it unwound when the credit stopped, not when the demand did. That is the failure mode: not fraud, but a demand signal that was partly a financing signal, discovered only when financing conditions changed.
It is worth noting that the current structure is meaningfully different from that precedent in one respect: the capital is third-party, so the credit risk sits with sophisticated institutional investors rather than on the manufacturer's own books. Whether that makes the system more robust — real underwriters with their own money at stake — or simply distributes the fragility more widely is the genuine open question, and reasonable people are on both sides of it.
A results call will not resolve this. It can move it, in either direction, and the disclosures that would do so are specific.
First: commitments, guarantees and off-balance-sheet exposure. Any language about supply commitments, purchase guarantees, backstops or residual-value support connected to the financing platforms. The absence of such language is itself informative.
Second: receivables and days sales outstanding. If DSO is rising materially faster than revenue, reported sales are converting to cash more slowly. This is the single most mechanical indicator available and it appears in the filings without needing management to characterise it.
Third: customer concentration. Nvidia discloses revenue concentration among its largest direct customers. The question is whether concentration is rising, and whether the concentrated names overlap with the entities the financing platforms are intended to serve.
Fourth: equity and credit exposure to buyers. Nvidia has taken equity positions in a number of AI companies that are also customers. The relevant figures are the size of that portfolio and whether any of it is being marked in a way that flows through reported results.
Context for why this has become the live debate rather than an accounting footnote: Microsoft, Amazon, Alphabet and Meta have collectively guided 2026 capital expenditure to roughly $720–745 billion, an increase on the order of 77% from 2025, with some analysts projecting combined hyperscaler spending above $1 trillion annually by 2027.
At that scale, the funding question becomes unavoidable arithmetic. Operating cash flow at even the largest technology companies does not cover it indefinitely, which is why the financing structure exists at all. The buildout has reached the point where how it is paid for is a more consequential variable than whether the chips are wanted.
“Investors have started to distinguish between demand a customer pays for and demand a supplier arranged. That distinction is what moved the stock on 10 August, and it is the frame worth carrying into Wednesday.”
This piece is not an argument that Nvidia's demand is illusory. There is no evidence for that claim and the burden would be substantial. It is an argument that 'is demand holding' has stopped being a useful question, and that the more informative one — what is the demand a function of — is answerable from disclosures rather than from a headline revenue number. On Wednesday the revenue figure will lead every summary. The receivables line will not, and it will tell you more.
Vendor financing is when a manufacturer helps its customers pay for its products — through direct credit, arranged third-party capital, guarantees or residual-value support. It is entirely permitted and completely ordinary in capital goods: aircraft, telecom equipment and industrial plant have been sold this way for decades. The accounting question is not whether it is allowed but whether the arrangements are disclosed clearly enough for an outsider to see how much reported demand depends on them.
DrafterDaily covers business the way it actually works — structure, incentives and the disclosures that settle the argument.
Read more Business coverageThe headline was the 94% drawdown from a $3.44 billion peak. The more useful arithmetic is the cash sitting inside the thing being bought — and the fact that everyone except the angels was paid in another private company's paper.
The biography is the least interesting part. What the board actually did was unbundle the CEO job — keeping Tim Cook on with a defined policymaker-engagement remit — and that separation says more about Apple's next decade than the appointment does.
Every outlet reported the size of Meta's teen-safety settlement and the screen-time limits. Almost nobody read the payment structure — where roughly $5.3bn is released only if two competitors act.