Dell's margin recovered and its cash did not

Disclosure. Convexity builds research tools for public-market data. This note is analysis, not investment advice, and nothing in it is a recommendation to buy, sell, or short any security. We hold no position in Dell Technologies, in any company named below, and no relationship with any of them. Every reported figure comes from the earnings release furnished with Dell's Form 8-K of September 1, 2026, or from management statements on the earnings call held the same day, both linked at the bottom. Where a number is our arithmetic on Dell's disclosed figures rather than a figure Dell reported, the sentence using it says so. This note corrects an argument we published on the morning of September 1, before the quarter it discusses was reported; that correction is the first section below and is also recorded on the original note.

On the morning of September 1 we published a note arguing that Dell's growth structurally dilutes its gross margin. Dell reported that afternoon. Gross margin was 20.9% of revenue against the 17.8% we had called a trend, and management attributed the improvement to precisely the segment we said was causing the damage.

We were wrong, and the note said in advance what would make it wrong, which is the only reason anyone can tell. Its first falsifier read: gross margin flat or up while AI revenue grows means the structural claim fails. AI-optimized server revenue doubled to $16.4 billion and the rate went up 260 basis points. That is the test, met on the first attempt.

What follows is the quarter as it actually printed, including the part of the original argument that survived, and a different problem that the margin recovery has obscured.

20.9%

gross margin, against 18.3% a year earlier

19.4%

gross margin for the half, against 19.5%

-13%

operating cash flow, on revenue up 58%

2.0x

inventories, in six months

Second quarter of fiscal 2027, ended July 31, 2026. Revenue, margin and cash flow figures are Dell's as filed. The inventory multiple is our arithmetic on the balance sheet, which compares July 31, 2026 with January 30, 2026 and is therefore a six-month change rather than a year-over-year one.

The quarter that beat everything

Revenue of $46,971 million, up 58%, against guidance of $44.0 billion to $45.0 billion. Diluted earnings per share of $6.34 against $4.48 guided at the midpoint. Net income of $4,133 million against $1,164 million. The full-year outlook went up by $25 billion to $192 billion, which is roughly 70% growth on a company that did $95.6 billion two years ago.

Underneath that, the Infrastructure Solutions Group did $31.8 billion, up 89%, with AI-optimized servers at $16.4 billion and traditional servers and networking at $10.5 billion, up 122%. Storage, which has been the persistent disappointment, did $4.9 billion and grew 26%. ISG operating income was $4.8 billion, up 225%.

That last pair is the number that undoes our argument. ISG operating margin reached 15%, which Dell says is a record and which is up 620 basis points year over year. The segment we described as structurally low-margin improved its own profitability by more than six points while nearly doubling in size.

Chief financial officer David Kennedy put the consolidated version on the call plainly: gross margin dollars grew 78% to $9.9 billion, and the rate of 21.1% was driven by an improvement in the ISG margin rate and a higher mix of ISG revenue. Both halves of that sentence point the opposite way from our note. The AI segment got better, and more of it made the blend better rather than worse.

Guidance says the same thing about the next quarter. Kennedy expects the ISG operating income rate up just over a point year over year, and said so in the same breath as AI server revenue more than tripling. A company that believed its own growth was margin-destructive would not guide that way.

What we actually got wrong

It is worth being precise about the error, because "we were too bearish" is not a diagnosis.

The original argument had two premises. The first was arithmetic and holds: through the first quarter of this fiscal year, blended gross margin had fallen as the AI mix rose, and the products line carried a materially lower rate than services. That was true and remains true of those quarters.

The second premise was causal, and it is the one that failed. We treated the AI server business as structurally low-margin, on the reasoning that assembling somebody else's accelerators into a rack is close to contract manufacturing, and that the margin therefore could not improve much with scale. If that were right, the only way the blend could recover would be for the mix to stop shifting.

The mix did not stop shifting. AI-optimized server revenue doubled, ISG rose to 68% of revenue, and the blend improved anyway, because the segment's own rate improved by 620 basis points. Dell was not stuck at a manufacturing margin. It had room, and the room came from somewhere our note did not consider: storage attaching to AI deployments at a much better rate, services attaching alongside, traditional servers growing 122% inside the same segment, and a pricing environment where supply constraints work in the seller's favour rather than the buyer's.

That last point is the uncomfortable one for a bearish reading. When DRAM is scarce, the company holding inventory and a backlog has pricing power it does not have in a normal market. The same constraint we framed purely as a cost risk is also, right now, a margin support. Both things are true and we only wrote down one of them.

The half tells a quieter story

There is a number in Dell's own percentage table that will not appear in the coverage, and it belongs in any honest reading of the quarter.

Gross margin for the six months was 19.4% of revenue. For the same six months a year earlier it was 19.5%.

The quarter expanded. The half did not. The first quarter of this fiscal year, the one our original note dissected, ran 17.8%, and the second quarter's 20.9% pulls the two-quarter average back to almost exactly where it sat a year ago. That is not a refutation of the recovery, which is real and which the segment detail explains. It is a caution about the size of the claim the recovery supports. One quarter of 20.9% after one quarter of 17.8% is a company whose margin moves several points from quarter to quarter on mix, not a company that has established a new level.

The honest summary is that the structural decline we described is not visible in the data, and neither is a structural improvement. What is visible is volatility, in a business that changed shape faster than most companies do in a decade.

Where the earnings actually came from

Consolidated operating income was 11.5% of revenue against 6.0%. That is a 550 basis point improvement, and only 260 basis points of it came from gross margin.

The rest came from operating expenses. Selling, general and administrative costs were 7.1% of revenue against 9.7%. Research and development was 2.4% against 2.6%. Total operating expenses fell to 9.5% of revenue from 12.3%. Kennedy described the full-year operating expense rate of roughly 8% as the lowest in the company's 42-year history.

This is a real achievement and it is also a different achievement from the one the headline margin implies. Dell grew revenue 58% while growing SG&A 15%. The operating leverage in that gap is enormous, and it is the largest single contributor to the earnings beat. It is also, by construction, finite: an expense base cannot fall as a share of revenue forever, and the easiest year to show that leverage is the year revenue rises 58%.

Two smaller contributors are worth naming for completeness. The diluted share count fell 5% to 652 million, so per-share growth outruns dollar growth. And Dell returned a record $4.3 billion to shareholders in the quarter through buybacks and dividends, with the board declaring a $0.63 quarterly dividend on September 1.

The half of Dell nobody is discussing

Client Solutions did $15.0 billion, up 20%, with commercial client at $13.2 billion, up 22%, and consumer at $1.8 billion, up 7%. Operating income was $1.1 billion, up 42%.

That is a third of Dell's revenue growing at twenty percent with operating income growing at twice the rate of sales, and almost nobody will write about it, because it is a PC business in a quarter where the same company booked $60.9 billion of AI orders. It deserves two observations anyway.

The first is that commercial growing at 22% while consumer grows at 7% is the shape of a corporate refresh, not a consumer recovery. Windows 10 support ended in October 2025 and the enterprise replacement cycle that followed is the most plausible driver. Refresh cycles are pull-forwards. They end.

The second is that CSG operating income growing 42% on revenue growing 20% is the same operating leverage story as the consolidated numbers, in the segment where it is least likely to be about AI. That supports the reading in the previous section: a meaningful part of what improved this quarter is expense discipline applied across a company that got much larger, rather than anything specific to the AI business.

Our original note treated CSG mostly as the stable thing that AI was crowding out of the mix. On this quarter it grew 20% and expanded margin. It is not being crowded out. It is just being outgrown, which is a different claim and a much less alarming one.

The other place the cash went

Inventory and payables are the loud part of the working capital story. The financing arm is the quiet part.

Short-term financing receivables were $12,805 million against $8,458 million at the fiscal year end, up 51%. Long-term financing receivables were $7,625 million against $5,822 million, up 31%. Together that is $20.4 billion against $14.3 billion, an increase of roughly $6.2 billion in six months by our arithmetic on the two pairs of balance sheet lines.

Dell Financial Services exists to let customers buy Dell equipment on Dell's balance sheet, and in a quarter where the equipment in question runs to hundreds of millions of dollars per deployment, a fast-growing financing book is what a fast-growing order book looks like from the lender's side. It is a normal and long-standing part of the company.

It is also cash that has left, and it is the reason the distinction between GAAP operating cash flow and Dell's adjusted free cash flow measure is not merely presentational. Dell's adjustment removes the effect of the financing business, on the reasoning that lending growth is an investment rather than an operating drain. That reasoning is defensible. The consequence is that the measure Dell directs attention to is the one that excludes a $6.2 billion six-month use of cash, and the measure it does not emphasize is the one that includes it.

Neither presentation is wrong. Which one is informative depends entirely on whether you think a rapidly expanding captive lending book against AI infrastructure buyers is a routine cost of doing business or a growing claim on the balance sheet. That is a judgment, and this note does not have the disclosure to settle it: Dell reports the receivables and the allowance against them, currently $186 million short-term and $128 million long-term, but does not break the book down by customer type in the release.

The number that got worse

Revenue rose 58%. Net income rose 255%. Cash flow from operating activities was $2,225 million against $2,543 million, which is a decline of 13%.

Put those together and the conversion is stark. Dell turned 54% of net income into operating cash this quarter. A year ago it turned 218% of a much smaller net income into operating cash. That is our arithmetic on the two figures as filed, and it is the cleanest single statement of what changed.

For the six months the picture is better and still not good: $6,306 million of operating cash flow against $5,339 million, up 18%, on revenue up 71%. Cash generation is growing. It is growing at roughly a quarter of the rate of the business that generates it.

Dell reports adjusted free cash flow as a non-GAAP measure and directs attention there, which is a legitimate presentation for a company with a captive financing arm whose receivables move with sales. We are using the GAAP operating cash flow line because it is the one that reconciles to net income without judgment, and because the gap between the two is the subject.

Where the cash went

The balance sheet answers it directly. These figures compare July 31, 2026 with January 30, 2026, so they describe six months rather than a year.

Inventories were $21,290 million against $10,437 million. They more than doubled. Over the same six months cost of revenue rose 71%. Inventory grew half again as fast as the cost of goods it represents, which is the definition of building stock ahead of sales rather than in proportion to them.

Accounts receivable were $22,918 million against $17,585 million, up 30%. Accounts payable were $49,723 million against $33,630 million, up 48%, and now equal 39% of total assets by our arithmetic on the filed balance sheet.

So the working capital build is being financed substantially by suppliers. That was true when we wrote the original note and it is more true now. Payables of $49.7 billion against total assets of $127.4 billion is an unusual structure for a hardware company, and it is the mechanism that lets Dell double inventory without raising equivalent debt. Long-term debt rose to $25,985 million from $23,513 million over the same six months, which is a $2.5 billion increase against a $10.9 billion inventory build.

Why the inventory line is the memory story

The original note argued that memory cost inflation was a risk of unknown magnitude sitting inside a price-locked backlog. The margin half of that note was wrong. This half was not, and management confirmed the mechanism on the call in terms that leave little room.

Asked about constraints, chief operating officer Jeff Clarke named DRAM first, then NAND, then noted spotty CPU shortages and said that just about every product going through a leading node is constrained. He also acknowledged, on the composition of growth, that input costs continue to go up and that there is an element of inflation inside the reported growth rate.

A doubling of inventory in six months, growing faster than cost of sales, in a company whose operating officer says memory is the binding constraint and that input prices are rising, is what buying ahead looks like on a balance sheet. That is an inference rather than a disclosure: Dell does not break inventory into components and does not disclose its purchase commitments in the earnings release. The Form 10-Q that follows carries a purchase obligations note, and that is where this stops being an inference.

The reason it matters is the same reason it mattered before. If Dell is pre-buying memory at rising prices to serve a backlog priced earlier, the margin effect arrives later than the revenue does, and the quarter that shows it will not be the quarter that booked it. The 620 basis point ISG improvement is evidence against that concern having bitten yet. It is not evidence about the timing.

The backlog got larger and the customer base got wider

Dell booked $60.9 billion of AI server orders in the quarter and recognized $16.4 billion of AI server revenue, exiting with a record $95 billion backlog. Clarke said the pipeline continues to grow sequentially and remains multiples of the backlog.

Two things in that are worth separating. The first is that $60.9 billion of orders in a single quarter, against a full-year AI server revenue expectation of $74 billion, means the backlog is being built far faster than it is being drawn down. The second is that Dell still does not quantify how much of the backlog is non-cancelable. Its own definition leaves the split undisclosed, and that was one of the open items in the original note. It remains open.

On concentration, which was the other open item, the direction is favorable. Clarke said demand is broadening across neoclouds, sovereigns and enterprise customers, and that the AI customer count has surpassed 6,500. That is up from the figure above 5,000 given a quarter earlier. A wider customer base is the specific thing that would make the concentration concern in the original note less pressing, and the count moved the right way. The audited customer concentration disclosure updates annually, so the 12% single-customer figure will not be refreshed until the Form 10-K.

What this note is claiming, and what it is not

It is not claiming that Dell has a cash problem in the sense of liquidity. Cash and equivalents were $11,569 million, essentially flat over six months. The shareholders' deficit narrowed to $(1,427) million from $(2,470) million as retained earnings grew from $3,325 million to $10,065 million. The company returned $4.3 billion in the quarter and raised its dividend. None of that describes strain.

It is claiming that the quality of the earnings and the size of the earnings are moving in different directions, and that the second is much easier to see than the first. A quarter that converts 54% of net income into operating cash, while inventory doubles and payables fund it, is a quarter where the income statement and the cash flow statement are telling different stories about the same business. That gap is not a problem by itself. It is the thing to watch, in the way the gross margin line was the thing to watch three months ago and turned out to resolve favorably.

The lesson we take from getting the last one wrong is not to make a smaller claim. It is to be clearer about which line settles it. For the margin argument, the settling line was the next quarter's gross margin percentage, and we said so, and it settled against us in ninety days. For this one the settling line is operating cash flow against net income over the next two quarters, and the purchase obligations note in the Form 10-Q.

What would prove this wrong

Cash conversion normalizes next quarter. The single cleanest refutation. If operating cash flow tracks net income in the third quarter, this was a working capital timing effect inside one unusually large quarter and nothing more.
The inventory build is deliberate and disclosed. If the Form 10-Q shows purchase commitments that match the inventory growth and management frames it as securing supply at known prices, then the build is a hedge that worked rather than an exposure, and the margin effect is already in the guidance.
Payables are structural, not a stretch. Dell has run very high payables for years, and days payable outstanding on this quarter's cost of revenue is roughly in line with where it has been. If suppliers are financing Dell because that is the industry's normal shape rather than because Dell is leaning harder, the funding observation is ordinary.
The half-year comparison is the wrong frame. Fiscal first quarters at Dell carry a different mix from second quarters. If the seasonal pattern explains the 17.8% and 20.9% pair, then the flat six-month figure is an artifact of comparing halves rather than evidence about the level.
AI margin keeps expanding through the tripling. Guidance says the ISG rate goes up again in the third quarter while AI revenue more than triples. If that holds, the mix concern is dead in every form, including ours, and the margin question should be closed rather than restated.
The backlog converts at disclosed prices. If Dell quantifies the non-cancelable portion and the pricing terms, the memory exposure inside the backlog becomes measurable, and it may well be small.
Operating leverage continues. An 8% operating expense rate is remarkable and there is no rule that says it stops. If Dell holds it while revenue grows again next year, the earnings quality question narrows to working capital alone.

Dell reports its fiscal third quarter on November 24, 2026. The Form 10-Q for this quarter follows within days of this note and carries the two disclosures that bear on it most: the purchase obligations note, and the detail behind the inventory line. Neither appeared in the release.

Primary sources

Disclosure, again. Convexity builds research tools for public-market data. This note is analysis, not investment advice, and nothing in it is a recommendation to buy, sell, or short any security. We hold no position in Dell Technologies and no relationship with the company. Every reported figure is from the filings and the call cited above. The following are ours, derived by arithmetic on those figures and not reported by Dell: the 54% and 218% cash conversion ratios and the comparison between them, the observation that inventories grew faster than six-month cost of revenue, the payables share of total assets of 39%, the total assets figure itself, the split of the 550 basis point operating margin improvement into its gross margin and operating expense components, and the comparison of the six-month gross margin percentages against each other. Dell does not break its inventory into components, does not disclose purchase commitments in the earnings release, and does not quantify the non-cancelable portion of its backlog; the section discussing memory pre-buying is explicitly framed as an inference from the inventory line and management's stated constraints, not as a finding. The customer concentration figure referenced from our earlier note is Dell's audited annual disclosure and updates only with the Form 10-K. The earnings call is cited from a published transcript rather than from Dell's own webcast archive, and quotations are attributed to the speaker as that transcript records them.