Series 01 · Equity Research
NVDA
NASDAQ · NVIDIA Corporation
Earnings Preview — No Position

Count the days, not the dollars.

Nvidia reports its fiscal third quarter tonight, five days after its founder resigned and with the stock a third below its October high. The revenue number will almost certainly be fine. This note is about the three things we will read before it: whether the guide still accelerates, how long customers are taking to pay, and whether the neoclouds that drove most of last quarter's growth can still fund their orders.

17 November 2026 · AMA Invest · Equity Research · ~11 min read

01 / The setupWhere the stock is going into the print.

Nvidia reports the quarter ended 25 October 2026 after the US close on Tuesday 17 November, with the call at 2 p.m. Pacific. It is the first report since Jensen Huang resigned as chief executive and director on 12 November. Colette Kress, the CFO who issued the guidance in August, will now report against it as interim CEO while keeping the CFO role. Harvey Jones has been named the company's first board chair, and the board says it is searching inside and outside the company for a permanent successor.

The stock closed at a record $239.24 on 6 October and at about $156 on Monday — roughly 35% lower, near a $148.80 52-week low, at a market value of about $3.75 trillion. Estimates have barely moved over that period. What has fallen is the multiple, to roughly 17 times forward earnings as the financial press calculates it. Options imply a move of roughly 7–9% on the print.

Q3 Guide
$108B
±2% · no China data-center compute
Q3 Consensus
~$109B
~$2.47 adj. EPS · ~+91% y/y
Q4 Consensus
~$125B
~+15% q/q — the number that matters
Stock
~$156
−35% from 6 Oct record close

For reference, the July quarter: revenue of $96.2B (+106% y/y, +18% q/q), data center $89.0B, gross margin 75.0%, non-GAAP EPS $2.22. On the income statement it was a clean beat. The cash-flow statement and the footnotes were where the questions started, and that is where we will be reading tonight.

02 / The slowdown testWhat a slowdown would actually look like.

A company guiding to $108B a quarter does not slow down by missing. Nvidia has beaten its own revenue guide by roughly 3–5% in each recent quarter, so a "normal" print is $111–113B, and $109B would already be the weakest beat in some time. The slowdown, if there is one, will show up in four quieter places.

1. The January-quarter guide. Consensus is about $125B, roughly 15% sequential growth. Nvidia's pattern is to guide a few points above consensus. A guide at $125B is in line and has tended to be sold; a guide near $118–121B would be the first deceleration of the cycle that the company chose to put in writing. Anything close to flat on the October quarter changes the debate entirely.

2. Whether fiscal 2028 is still supply-limited. In August Ms Kress said next year's growth of about 70% was constrained by supply, with customer forecasts pointing to roughly double that. We want to hear that sentence again, in those words. If the constraint quietly shifts from "what we can build" to "what customers have scheduled," that is the tell.

3. Vera Rubin's share and the gross margin. Rubin was guided to about 20% of data-center revenue in its first quarter, with gross margin stepping down a point to 74% (±50 bps). A lower Rubin share means the ramp slipped; a margin below 73.5% means memory and system costs are biting harder than planned. Nvidia's supply commitments jumped from $119B to $279B in a single quarter, mostly for memory, so the margin guide for January is a direct read on how that purchase is pricing.

4. Who is growing. Last quarter, hyperscalers grew 13% sequentially to $48.7B, while "AI clouds, industrial and enterprise" — the bucket that holds the neoclouds, AI-native labs, sovereigns and enterprises — grew 25% to $40.3B. On our arithmetic that second bucket supplied about 59% of the quarter's incremental data-center revenue. The mix matters more than the total: hyperscaler dollars come out of operating cash flow, and a growing share of the rest comes out of capital markets.

What is not slowing The largest customers are not cutting. After their September-quarter reports, the four big US hyperscalers are guiding to roughly $730B of combined 2026 capital spending. Alphabet said 2027 will rise significantly; Amazon reaffirmed about $220B; Meta raised the low end of its range. Microsoft trimmed to about $170B from about $175B — and attributed it to third-party capacity arriving late, not to demand. Late capacity is the theme to carry into section 04.

03 / The payment clockCustomers are taking longer to pay. Nvidia is not.

This is the line item we care most about. In the July quarter, Nvidia's days sales outstanding went from 45 to 60 in three months. Receivables rose by $22.3B to $63.1B — up 55% in a quarter in which revenue grew 18%. Management's explanation was "extended payment terms" on large, multi-quarter purchases by "certain investment-grade customers." As reported from the 10-Q, some of those terms now run from 90 days to as long as a year, and five direct customers account for about 70% of the receivable balance.

It is worth being precise about which side of the ledger stretched. Nvidia's own payments to its suppliers have barely moved: on our calculation, days payable went from about 52 at the January year-end to about 57 in July. The fifteen extra days are on the other side. Nvidia, as the vendor, is being paid later — and at the same time it is paying earlier for supply, through larger inventories and prepayments. Both effects pull cash in the same direction.

LineJan → Jul 2026Read
Receivables$38.5B → $63.1B$22.3B of the rise came in the July quarter alone. Five customers hold ~70% of the balance.
DSO45 → 60 daysApril to July, as disclosed. Company cites extended terms for investment-grade buyers of multi-quarter orders.
Inventory$21.4B → $31.6BAbout 119 days of cost of revenue. Raw materials tripled; management ties the build to Vera Rubin.
Payables$9.8B → $15.1BDays payable ~52 → ~57 on our estimate. Suppliers are being paid on roughly the same clock.
Supply commitments$119B → $279BApril to July. Primarily memory; about $92B falls due in the remainder of fiscal 2027.
Free cash flow$48.6B → $21.3BQ1 to Q2. Roughly 36% of GAAP net income of $59.7B, which itself included ~$7.8B of gains on equity stakes.

One quarter of this is explainable. A product transition needs inventory, and offering a hyperscaler longer terms on a multi-quarter order is a commercial choice rather than a credit event. The question for tonight is whether July was a step or the start of a slope. Because the company will report a DSO figure in the CFO commentary within minutes of the release, this is one of the few parts of the debate that gets a number.

DSOReceivables at ~$109B revenueWhat it would tell us
~45 days~$54BJuly was timing. The receivable build reverses and free cash flow rebounds sharply.
~55 days~$66BPartial normalisation. Terms were a one-off reset for a handful of large orders.
~60 days~$72BThe new terms are the terms. Not alarming alone, but cash conversion stays structurally lower.
70+ days$84B+A second consecutive stretch. Growth is being bought with the balance sheet, and the "investment-grade" description needs testing.
Revenue tells you what was shipped. Receivables tell you who has paid for it. In July those two numbers stopped moving together.

Alongside the DSO figure we will check three things in the 10-Q when it is filed: whether the customer share of receivables has concentrated further, whether the language on terms still says "investment-grade," and whether the supply-commitment total has risen again from $279B. A company that is paid later by its customers and commits earlier to its suppliers is carrying the cycle's financing on its own balance sheet. With $56.6B of cash and debt securities and $25B of notes issued in June, Nvidia can afford to. We simply want to know how much of the growth depends on it.

04 / The neocloudsThe customers that have to borrow to buy.

The neoclouds reported in the past two weeks, and their message was consistent: demand is not the problem, delivery and funding are.

CoreWeave raised full-year revenue guidance on a $3.66B quarter and a backlog of about $132.5B. It also carries $41.6B of corporate debt, paid $846M of net interest in the quarter, guides to $37–39B of 2026 capex, and its CFO said the company will not issue unsecured debt at current yields. Nebius moved 2026 revenue to the low end of its $3.0–3.4B range because third-party colocation sites were delivered late, raised capex to about $27B, and lost nearly half its market value in a month. IREN told investors that about $6B of this fiscal year's build is not yet funded. In each case management said no capacity had been cancelled.

For Nvidia this matters in two ways. The first is timing. A GPU cannot be installed in a data hall that has not been handed over, and Microsoft's capex trim, Nebius's guide and the analyst downgrades that followed all cite the same bottleneck. Late sites mean later shipments, or shipments that sit in a customer's receivable for longer. That would connect section 03 to this one. The second is funding. Neocloud equity is down sharply and debt is more expensive, which is the condition under which an order book stops converting on schedule.

The third link is Nvidia itself. By July it held roughly $94B of equity securities in public and private companies, up from about $35B in January, with a further $25B committed. It had guaranteed up to $108.5B of data-center lease obligations — $105B of that for a single campus leased to OpenAI — and had $29B of cloud-service agreements, including capacity rented from companies that also buy its systems. Press reports since the leadership change say Ms Kress is expected to slow new investments and credit support, while honouring anything contractually binding.

The question we would ask on the call If Nvidia provides less financing to its customers from here, how much of the fiscal 2028 order book assumed that it would? A more conservative balance sheet is good for the quality of earnings and, at the margin, bad for the quantity of orders. Tonight is the first time the person who made that trade-off as CFO has to answer for it as CEO.

05 / The other sideThe bull case, stated properly.

We have listed a good deal that could go wrong, so it is worth being equally clear about what the stock already discounts.

Nothing in the reported numbers has weakened. Revenue doubled year on year last quarter, the guide implies another 12% sequential step, and the largest buyers are raising capital budgets rather than cutting them. Neocloud management teams describe qualified demand at multiples of the capacity they can bring online, and several raised prices during the quarter.

The de-rating has been about people and politics. Earnings estimates have been roughly flat while the stock fell a third. Seventeen times forward earnings is a multiple the market gives to a business it expects to stop growing. If Ms Kress delivers an ordinary beat and an ordinary guide, the stock does not need a new narrative; it needs the absence of a bad one.

The working-capital build may simply reverse. Receivables on multi-quarter orders are collected as those orders ship. Inventory built for Rubin becomes revenue when Rubin ships. A quarter in which DSO falls and free cash flow exceeds net income would retire the most quantifiable part of the bear case in one release.

The interim CEO wrote the guidance. Whatever else has changed, the person reporting tonight is the person who set the numbers in August and has signed every set of accounts for thirteen years. Of all the ways a succession can begin, a first report from the former CFO is the least likely to contain a surprise in the financials.

06 / The checklistWhat we will read, in order.

We find it useful to write the thresholds down before the release rather than after it:

ItemReassuringConcerning
Q3 revenue$111B+Below ~$109.5B — the smallest beat of the cycle, even if technically above the guide.
Q4 guide$128B+$121B or lower, or a range wide enough to be a hedge.
DSO≤ 55 daysAbove 60 for a second quarter; receivables above ~$72B.
Free cash flow$45B+Below ~$30B again, with working capital as the reason.
Gross margin74%+, flat guideUnder 73.5%, or a Q4 guide that steps down again on memory.
Rubin~20% of data centerMaterially lower, with the shortfall pushed into the January quarter.
Customer mixBig cloud leadsGrowth again led by the AI-cloud bucket while its receivables lengthen.
SupplyFlat near $279BAnother large step up in supply commitments, guarantees or equity stakes.
FY28 languageSupply is the limitAny softening on the ~70% growth frame, or on customer schedules.
CEO searchA timelineNew disclosure on export-control inquiries or the pending shareholder suit.

07 / ScenariosWhere this clears.

Our probability-weighted view of where the stock settles in the days after the report, against ~$156:

Prob.Clearing priceScenario
25%$172 – $185 Clean beat and cash repair. Revenue above $111B, a Q4 guide near $128–130B, DSO back toward 50–55 days and free cash flow above $45B. The balance-sheet debate closes and the multiple recovers part of what it lost.
40%$150 – $165 Good income statement, unresolved balance sheet. A normal beat and an in-line guide, with DSO still near 60 and commitments higher. Nothing breaks and nothing is settled.
25%$135 – $148 The slope. A thin beat, a Q4 guide of $118–121B attributed to customer site timing, and DSO above 60 again. The stock tests and breaks its 52-week low.
10%$115 – $130 Push-outs acknowledged. A guide close to flat on the October quarter, explicit neocloud or sovereign delays, or a reserve against receivables. Estimates are cut for the first time this cycle.

That weights to roughly $155 — almost exactly the current price. The distribution is wide, with about 14% of upside to the middle of the best branch and about 21% of downside to the middle of the worst. We do not think the expected value pays for the range.

Why we are watching rather than acting The single most important input to this stock for the next year — whether receivables and commitments keep growing faster than revenue — will be published tonight and cannot be estimated from outside. We would rather pay a higher price for a known answer than today's price for a guess.

08 / PositionWhat we are doing.

AMA Invest holds no position in NVIDIA (NVDA), long or short, at the time of publication, and does not intend to initiate one ahead of the report.

We would become interested on the long side if the release shows DSO at 55 days or lower, free cash flow back above $45B, a January-quarter guide at or above $128B and no further growth in guarantees or supply commitments — in other words, if the growth is shown to be funded by customers rather than by Nvidia. We would stay away, or look at the short side, on a second consecutive rise in DSO alongside a guide at or below consensus, because that combination says the receivable balance is doing work the order book used to do.

The discipline is simple to state: for a company growing at this rate, the income statement is the last place a slowdown appears. The payment clock is the first. We will publish a follow-up after the 10-Q is filed, scoring each line of the checklist above against what was reported.

AMA Invest
Private Investment Fund · Cayman Islands
Position disclosure. AMA Invest holds no position, long or short, in NVIDIA Corporation (NASDAQ: NVDA) at the time of publication. Positions may change without notice.

Sourcing. July-quarter results, guidance, balance-sheet and cash-flow figures are taken from NVIDIA's Q2 fiscal 2027 earnings release and CFO commentary (26 August 2026) and its press release on the CEO transition (12 November 2026). Receivable concentration, payment-term ranges, inventory composition, commitment and guarantee detail are from the Form 10-Q for the quarter ended 26 July 2026 as summarised by third-party analyses, and should be checked against the filing. Consensus figures (~$108.8–109.2B revenue, ~$2.47 EPS, ~$125B January-quarter revenue), the forward multiple, the options-implied move and share-price data are as reported by financial press and data providers on 13–17 November 2026 and differ by source. Neocloud and hyperscaler figures are from those companies' third-quarter 2026 reports as covered by the press. Days payable, the ~59% incremental-revenue share, receivable levels at each DSO, free cash flow as a share of net income, and all scenario prices and probabilities are AMA Invest estimates, not company figures. Reports regarding NVIDIA's future investment policy are unconfirmed by the company.

This note reflects the views of AMA Invest as of the publication date and is provided for informational purposes only. It does not constitute an offer to sell or a solicitation of an offer to buy any security or investment product, nor investment advice. Statements regarding markets, valuation, positioning, and outlook are the manager's opinions and are subject to change without notice. Forward-looking statements involve risk and uncertainty; actual results may differ materially. Past performance is not indicative of future results. Investments in the fund are available only to eligible investors who meet applicable suitability and accreditation requirements under relevant jurisdictions. All investments involve risk, including the possible loss of principal.