TD SYNNEX Q3 FY26 — Amazon Got the Servers, the Warrant, and the Working Capital

A record the tape refused

Set the beat aside for a moment, because it is not in dispute. Revenue came in $2.558B above the top of the company's own $18.2B to $19.0B outlook. Non-GAAP EPS of $5.68 was 22.15% over consensus, the fourth straight quarter of beating it ($4.73 against $3.20 in Q1, $4.85 against $4.10 in Q2). The chief executive, Patrick Zammit, opened with "another record quarter, with Distribution and Hyve both performing above our expectations and growing above market." All of that checks out against the filing.

The stock still fell 9.88%, from $287.89 to about $259.44 (unverified — last regular-session trade, market data rather than a filing figure). I don't read that as the market disbelieving the demand. It read the other two statements in the release, the cash flow statement and the balance sheet, and saw that "growing above market" had become a question of who carries the inventory. For a distributor that has always been the real question. It is just usually too quiet to notice.

The income statement is the thin part

Gross margin fell 61 basis points to 6.61%, and it has fallen every quarter this year: 7.30%, 6.84%, 6.61%. What rescued operating income was SG&A, which rose only $36M to $783M, so the 2.98% operating margin widened even as gross margin narrowed. That operating leverage is genuine. But it sits on a business where 93.4% of every revenue dollar is cost of revenue before overhead touches it, which leaves very little room for error.

The mix explains the slide. By the presentation's numbers (unverified against the filing), Hyve billed about $7.0B, up 117%, and is now 22% of billings. Its operating income rose only 56%, to $253M, so its margin on billings fell to 3.61% from 5.04%. The chief financial officer called the new programs "neutral to accretive" (unverified pending the official transcript). What that phrase leaves out is that the programs are accretive in dollars and dilutive in rate, and the rate is what a distributor's multiple is built on. Distribution carried the profit: $483M of operating income, up 55%, on billings up 27%, with Europe's operating income up 116%. The part of the business that earns the best returns grew slowest. The part that grew fastest is where the balance sheet went.

Where the $416M went

Net income was $416M and operating cash flow was negative $917M, against positive $246M a year earlier. Receivables absorbed $1,993M and inventory $1,420M; suppliers funded $1,310M through payables. Free cash flow for the nine months is negative $2,237M, against negative $33M a year ago. This is the warehouse is the position in its plainest form: the earnings are real, and they are sitting on the balance sheet as goods and invoices that have not turned into cash yet.

Management's explanation was specific, and I'll take it at face value: roughly $1 billion of free cash flow consumed by "increased inventory in Hyve's supply chain business in addition to new customers and new programs with existing customers" (unverified pending the official transcript). Read what that sentence actually says. It isn't a channel stuck with product nobody wants. It is inventory held on behalf of named programs. That rules out the bear case where demand is the problem. It also rules out the idea that the build clears when demand cools, because nothing about demand is cooling. The only way it clears is if the programs start converting cash faster than new ones consume it.

Two flat years, then a new floor

This is the part that changes the reading, and it comes from the longer series rather than the quarter. Take trade working capital, meaning receivables plus inventory minus payables. From May 2024 to November 2025 it sat between $3.55B and $4.74B, through a period when inventory rose $2.40B over eighteen months. Then it went to $4.86B in February, $5.71B in May and $7.80B in August. Inventory added $5.79B in nine months, up 60.95%, while nine-month revenue grew 29.17%. Year over year, inventory is up 67.29% and payables 43.39%. Suppliers are still funding most of the build, but a shrinking share of it.

The same series corrects the headline I nearly wrote. Cash falling from $2.44B to $0.75B looks like a collapse. It isn't. The $2.44B November balance was the high point of the series; a year ago cash was $0.87B, and in February 2025 it was $0.54B. The cash level is ordinary. What changed is the borrowing needed to hold it there: $1,500M of net revolver draws in the quarter, current borrowings up to $1.93B from $1.02B, total borrowings $5.52B against $4.61B, and interest expense up 32.16%. So the problem isn't a cash shortage. It is that growing Hyve now needs debt. A one-period change is a phase, not a decision, until you have the series — and in this case the series makes the finding stronger. The build isn't one quarter. It has run for three consecutive quarters from a two-year floor, and each quarter added more than the one before.

The composite, and what it averages away

Management gave one composite: a gross cash conversion cycle of 22 days, up 5 days sequentially and 6 days year over year, "reflecting incremental mix of Hyve," on net working capital of $6.5B (unverified pending the official transcript). The composite hides the component. Six days is a modest-sounding number, but the parts behind it are not modest: inventory 0.64 times quarterly revenue at the end of Q1, 0.71 at the end of Q2 and 0.709 at the end of Q3, with payables growing more slowly than inventory. A cycle that lengthens because the fastest-growing segment carries more stock per dollar of revenue won't shorten as that segment keeps growing. It shortens only if Hyve's own days fall. The phrase "reflecting incremental mix of Hyve" is the company telling you the six days are structural to the mix, whether or not it meant to.

The rebate that indexes to your own stock

The part of the quarter the release mentions only in footnotes is the Amazon warrant. On May 30 TD SYNNEX gave Amazon.com NV Investment Holdings a warrant for up to 3,238,066 shares, about 4.08% of the diluted count. 215,871 of those vested immediately at $0.01. The remaining 3,022,195 vest at $191.10 "upon the attainment of specified payment thresholds based on aggregate payments by" Amazon affiliates. The warrant expires in 2033, vests in full if the company is acquired, and the 8-K adds that the company "expects to continue its multi-year collaboration with Amazon."

This is the rebate that indexes to your own stock, the Marvell–Google structure applied to a distributor. Each tranche is paid in shares, not a fixed discount, so its cost rises with the stock price, and at about $259.44 the unvested strike is well in the money. The 8-K never sizes the payment thresholds, so a reader can't convert the warrant into a cost per dollar of Amazon revenue. That missing disclosure is the finding. It is already showing up in the numbers: a $15M provision for common stock warrants in Q3, participating securities taking about 1.1% of net income against 0.9% a year earlier and a forecast 1.5% in Q4, and a Q4 share-based compensation line of $40M against $16M reported in Q3. Combine it with the cash statement and the arrangement is clear. Amazon gets the servers, a claim on the equity that grows as it buys more, and a supplier that funds the inventory for those programs. It is the vendor becomes the credit with a hyperscaler as the counterparty. That is the safest version of it, and it is still a balance sheet doing work the income statement is credited for.

The guide says the same thing twice

Q4 guidance puts billings at $31.4B to $32.4B, flat on Q3's $31.8B at the midpoint, and GAAP net income at $368M to $408M. The $388M midpoint is below Q3's $416M. Non-GAAP EPS of $5.65 to $6.15 looks fine, but the gap to GAAP ($4.58 to $5.08) is where the warrant charge sits. Flat billings are the best case for cash: it's the first quarter this year when growth isn't adding to the working-capital load. Management says Q4 cash flow turns positive and that every business will be a "sustainable cash generator in fiscal 2027" (unverified pending the official transcript). A flat-billings quarter is exactly when that promise costs the least to keep.

The strongest case against me

The best bull argument is seasonal, and the company's own history supports it. Fiscal Q4 is when TD SYNNEX releases working capital: trade working capital fell between August and November in both of the last two years (the chart above shows both dips), and cash reached its $2.44B peak at the last year end. Put that pattern together with flat Q4 billings, management's statement that "a significant portion of the investments to support expected growth is now in place," and a buyer with every reason to pay on time, and the $7.80B could fall back quickly. On that reading this quarter was the peak of a program ramp, the 9.88% drop overshot, and I'm confusing a lump with a new level. It is a serious argument, and I take it seriously. I still come down on the other side, for one reason: the seasonal release has never started from a base this high. Giving back what a normal Q4 gives back still leaves the balance near the May level. To reach the old band, Q4 would have to reverse more than two quarters of Hyve build in one quarter, while the CFO is describing mix as the reason the cycle lengthened.

What decides it

One number, on one date: trade working capital on the November 30, 2026 balance sheet, published with Q4 results in January 2027. If it is above $5.71B, the build is the new operating level for a Hyve-weighted TD SYNNEX. That means a distributor multiple on a business that now needs debt to grow. If it is at or below $4.74B, the ramp has converted, management's cash language was right and I'm wrong. Anything in between is a partial result, and I'd call it that. Two secondary checks: Hyve's margin on billings against the 3.61% printed here, which tests "neutral to accretive", and the warrant provision line, which is the only public view of how fast Amazon's payments are climbing the vesting schedule.

The trade (analysis, not advice)

I wouldn't pay for this quarter's EPS beat until the November balance sheet is out. The demand is real, the distribution franchise is earning more, and the warrant ties TD SYNNEX to the largest buyer in the market. None of that is in question. What's in question is how much the growth costs, and the 9.88% move priced exactly that. The asymmetric setup is to wait for the Q4 print. A fall back toward the old band would re-rate the name on better cash quality with the beat already confirmed. A number above the May level would mean the market's reaction today was just the start of the de-rating, not an overshoot. Anyone who wants the AI server trade without the working capital can get it one layer up, from the component and accelerator vendors, through picks-and-shovels capture. Anyone buying TD SYNNEX should know they are buying Hyve's inventory along with it.

Bottom line

TD SYNNEX printed its best quarter and its worst cash quarter together, and they come from the same place. Hyve is growing at 117% on Amazon-scale programs, and TD SYNNEX pays for that growth three ways: working capital, debt and equity. The income statement gets credit for all of it, and the balance sheet carries all of it. My view is that November shows the carrying cost as the new normal, not a one-quarter lump. The number that decides it is trade working capital above $5.71B or at or below $4.74B on November 30.