Michael Burry has bought Nvidia calls as a hedge and, in the same breath, said he still expects the company to "not distribute enough to shareholders." Those eleven words reach us the way this kind of thing usually does — relayed by a market-data account summarising his disclosed positions on 26 August, not from a filing and not from a post of his we have read. Treat it as a claim someone made, because that is what it is.
It is worth testing anyway, because both halves of "distribute enough" are filed quarterly — and Nvidia had filed them earlier that same afternoon. In the July quarter it returned a record $26 billion — $19.732 billion of buybacks and $6.047 billion of dividends — against $21.4 billion of free cash flow. It paid out 120% of what the quarter generated, and the dividend line alone was 24.8 times what it had been a year earlier.
On the plain reading, the claim is wrong. On the reading a holder actually experiences it — cash arriving against money invested — that record $26 billion is worth 1.8% a year, and the reason has almost nothing to do with the payout ratio.
The points
- $25.779 billion returned in the July quarter, $19.732 billion of repurchases and $6.047 billion of dividends. Disclosed; CFO Colette Kress called it "a record $26 billion."
- $21.400 billion of free cash flow in the same quarter — operating cash flow less capital expenditure. Disclosed. The payout ratio is therefore 120.5%, which is ours.
- Dividends of $6.047 billion against $244 million a year earlier. Disclosed on both sides. Per share that is a penny going to twenty-five cents, and the next one is $0.25 payable 1 October.
- "Relative to our plan to return 50% or more of free cash flow, we have returned 60% on a year-to-date basis." Kress, on the 26 August call. Disclosed, and it is a policy rather than a quarter.
- $127.0 billion of free cash flow over the last four quarters, ours from the reported figures, against about $5.59 trillion of market value at the 4 September close of $230.36. That is 44 times free cash flow.
- Shareholder yield is the payout ratio divided by the price-to-free-cash-flow multiple. At 44×, a 60% payout is 1.36% and a 100% payout is 2.27%. Ours, and it is the article.
The payout ratio says whatever quarter you pick
The 120% looks decisive and it is not, because the denominator moved. Nvidia's free cash flow over the last five quarters ran $13.5bn, $22.1bn, $34.9bn, $48.6bn and then $21.4bn — a 56% sequential fall in a quarter when net income rose to $59.7 billion. The gap was a $22.3 billion receivables build as days sales outstanding went from 45 to 60, and it is the single reason the ratio spiked. Our analysis of the print covered that, and the $25 billion of notes Nvidia issued in the same quarter, which is how a company pays out more than it collected.
Run the same sum on the April quarter and the payout ratio is about 33%. Run it on the six months and Kress gives you the answer herself: 60%, against a stated floor of 50% or more. Three periods, three numbers, one programme. A ratio whose denominator swings 56% quarter to quarter cannot settle an argument about whether a company distributes enough, and anyone quoting a single quarter's version of it — in either direction — is picking a number.
The policy is the durable statement, and it is unambiguous: half of free cash flow is the floor, 60% is where the year is running, and Kress said Nvidia intends "to increase and return excess free cash flow net of strategic uses."
The number that does not move
Convert the same programme into what a holder receives and the picture stops being ambiguous. At $230.36 on 4 September, against the 24.285 billion diluted shares reported for the July quarter, Nvidia is worth about $5.59 trillion — 44.0 times the $127.0 billion of free cash flow it produced over the last four quarters. The card at the top of this piece is the table below.
| Payout basis | Cash a year | Yield |
|---|---|---|
| Dividend at $0.25 a share | $24.2B | 0.43% |
| Buyback at July's rate | $78.9B | 1.41% |
| Both, at July's rate | $103.1B | 1.84% |
| Policy floor: 50% of free cash flow | $63.5B | 1.14% |
| Year to date: 60% of free cash flow | $76.2B | 1.36% |
| Every dollar of free cash flow | $127.0B | 2.27% |
That last row is the ceiling, and it is the point of the piece. There is no distribution policy available to Nvidia that produces a yield above about 2.3%, because the identity is fixed:
shareholder yield = payout ratio ÷ price-to-free-cash-flow
Nvidia controls the numerator and not the denominator. It has already pushed the numerator about as far as a company with a capital programme can — 60% of free cash flow, a 25-fold dividend increase, a buyback running at $79 billion a year, and a quarter in which it borrowed to pay out more than it took in. All of that bought a 1.8% yield. Doubling the dividend again would add 0.43 points. Getting the total to 2.5% would require paying out more than every dollar of free cash flow, permanently.
The mechanism is worth stating plainly because it works against the company as it succeeds. Every extra dollar of annual free cash flow adds at most a dollar to what can be distributed — and, if the market keeps paying 44 times for it, adds $44 to the market value that dollar is measured against. Growth that the market capitalises faster than the payout ratio can rise is a yield that falls while the cheques get bigger. The distribution curve and the yield curve are not the same curve, and only one of them is Nvidia's to move.
For what it is worth, the yield has risen. The same quarter a year earlier returned $9.965 billion, or $39.9 billion annualised, against a market value near $4.4 trillion at the end of July 2025 — roughly 0.9%. It has doubled in a year, and it is still under two.
The clause that does the work
If Burry's sentence has a defensible reading, it is not "Nvidia returns too little." It is the four words Kress attached to the forward commitment: "net of strategic uses."
Those uses are disclosed and they are not small. Nvidia has put nearly $50 billion into the frontier AI labs, discloses $25 billion of committed equity investments among its future commitments, and paid $2.944 billion on the Groq arrangement in the July quarter alone. Its non-marketable securities went from $22.3 billion in January to $51.2 billion in July — a $28.9 billion increase over the same six months in which about $42 billion went to shareholders, that being the company's 60% applied to the half-year's free cash flow. Nvidia does not split the $28.9 billion between fresh cash and mark-ups on stakes it already held, so it is a ceiling on what the investing cost rather than the cost itself. We are not going to guess at the split. What that money is buying is a separate argument, and we have made it separately.
The honest summary is that Nvidia has two claims on its free cash flow and has told you the order: strategic uses first, then 50% or more of what is left. A shareholder who wants the order reversed is making a governance argument, not an arithmetic one.
What it does to our model
Nothing, and the shape of the nothing is the point. Our Nvidia model prices racks and margins, holds the share count flat at 24.285 billion across the whole horizon, and carries a single net-cash figure of $24.2 billion struck on 31 August. It has no buyback driver and no dividend line, and adding one would not move the fair value much in either direction: a repurchase takes a dollar out of net cash and a share out of the denominator at the same moment, and at fair value those cancel. A distribution policy changes who ends up owning the compounding, not what the compounding is worth. That is precisely why no valuation model settles a "distribute enough" argument — and why this piece is arithmetic about a yield rather than a case about a price.
What would change this
- Whether the receivables build reverses. The policy is stated as a share of free cash flow, so the denominator of the ratio and the denominator of the ceiling both depend on 60-day terms going back toward 45. Two more quarters at 60 and the cash line and the revenue line are telling different stories.
- What "strategic uses" costs next. There is $25 billion of committed equity investment in the table and no cap disclosed on the category. It sits ahead of the 50% floor by the company's own wording.
- The repurchase authorisation. About $99 billion remained after the July quarter — a little over five quarters at the rate Nvidia bought in July. A top-up would say the pace is meant to hold.
- The multiple. It is the one term in the identity Nvidia does not set. At 30 times free cash flow the same 60% policy yields 2.0%; at 60 times it yields 1.0%. Nothing the company does to its payout matters as much as what the market charges for the cash flow.
Every July-quarter figure here — repurchases, dividends paid, operating cash flow, capital expenditure, the securities balances, the committed equity investments and the remaining repurchase authorisation — is as Nvidia reported for the three months ended 26 July 2026. The 50%-or-more payout plan, the 60% year-to-date figure, the "net of strategic uses" wording, the nearly $50bn in frontier AI labs and the $0.25 dividend are Colette Kress's, on the 26 August earnings call. The Burry quotation is claimed, not disclosed: it reaches us through a third-party market-data account summarising his positions, and we have not seen the underlying statement. Free cash flow is operating cash flow less capital expenditure, and the trailing $127.0bn, every payout ratio, every yield and the 44× multiple are ours. Market value uses the 4 September close of $230.36 and the 24.285bn diluted shares reported for the quarter, so it moves with the price; the year-ago 0.9% yield rests on a market value implied by Nvidia's trailing earnings multiple at end-July 2025 and is approximate.
