23 American companies filed reports on 27 August, and six of them received a standalone article here. Most reported a quarter; IREN and Affirm reported a full fiscal year ended 30 June.
This piece does not summarise them. It is about three things visible only when you read them in sequence - and none of them appears in any report's headline.
Errors or inaccuracies are possible. Spotted something that looks wrong? Write to me and I will correct it.
Six That Received Their Own Article
| The central figure | |
|---|---|
| Marvell | Record revenue, +37% - and the non-GAAP gross margin fell |
| IREN | The crossover: AI overtook mining. And financing at 6% against 9% |
| Rubrik | +38%, the loss narrowed - and guidance is almost flat sequentially |
| SentinelOne | "Record profitability" and a widening GAAP loss, in the same quarter |
| Best Buy | The margin rose, and product margins fell |
| Dollar General | A fifth consecutive quarter of traffic growth |
And three more are covered inside this piece: Affirm, Ulta Beauty and Gap.
The First Pattern: The Gross Margin Fell at Three Technology Companies
Three companies with almost nothing in common - a chipmaker, a backup company and an endpoint security company - reported the same thing on the same night: cost of revenue grew faster than revenue.
| Revenue growth | Cost of revenue growth | The margin | |
|---|---|---|---|
| Marvell | +36.6% | +28.9% | Non-GAAP: 59.4% → 58.9% |
| Rubrik | +37.9% | +45.0% | GAAP: 79.5% → 78.4% |
| SentinelOne | +20.6% | +34.9% | GAAP: 75% → 72% |
At Marvell the mechanism differs from the other two - there cost of goods sold grew more slowly than revenue, and the GAAP margin actually rose. What fell is the non-GAAP margin, the one that strips out amortisation of intangibles and describes the product economics. And guidance for next quarter is 57.5%-58.5% - another step down.
And at the other two it is direct: cost of revenue grew at almost twice the pace of revenue.
And at a software company, that departs from the basic logic of the sector. The classic software model says the hundred-and-first customer costs almost nothing, so the margin should widen with scale, not narrow.
What could explain it? The reports do not say, so I do not assert. The obvious hypothesis is compute cost: products running models in real time consume compute, and that is a cost that grows with usage and sits in cost of revenue.
If that is indeed the reason, it is a very significant development - because it says AI is not only a source of new revenue in software, but also a new cost item that was not there before.
And that is exactly the number I will track over coming quarters - not the growth rate, but what happens to the line beneath it.
The Second Pattern: Two Chains, the Same Tailwind, and Both Said It Is Leaving
Best Buy and Dollar General do not compete, do not sell the same thing, and do not serve the same customer.
On the same night, both reported a margin improvement sourced in tariff refunds.
| The benefit | What the company said | |
|---|---|---|
| Best Buy | about $34 million, IEEPA tariff refunds | Partially offset by "lower product margin rates" |
| Dollar General | about 81 basis points of gross margin, about 66 of operating, about $0.25 of EPS | Does not anticipate a material impact going forward |
And that teaches two things.
The first, about the reports themselves: at Dollar General, earnings per share rose 33.3%. Without the tariffs it would have risen about 20%. Both numbers are correct, and only one describes the business.
And the second, broader: trade policy enters the accounts. Not as an idea and not as a future risk - as a line in one quarter's gross margin, which vanishes the next.
And anyone comparing quarter to quarter without stripping it out is measuring a trend that does not exist.
The Third Pattern: The Bottom Line Was Created Below the Operating Line
At three of the night's companies, net income does not describe what happened in the business - because it was created in the lines beneath.
And the most striking is Affirm.
| Affirm, fiscal 2026 | |
|---|---|
| Revenue | $4,261.1 million (+32.2%) |
| Pre-tax income | $492.7 million (against $61.5 million) |
| The tax line | a benefit of $1,437.1 million |
| Net income | $1,929.8 million |
| Diluted earnings per share | $5.53 (against $0.15) |
In other words net income is 3.9 times pre-tax income, and 74.5% of it originates in the tax line.
This happens when a company recognises a deferred tax asset - an accounting recognition that accumulated past losses will be offset against future profit. And it is one-off in character, usually occurring when a company reaches sustained profitability.
So the real story at Affirm is not $1.93 billion. It is $492.7 million - pre-tax income that grew eightfold, on a pre-tax margin of 11.6%. And that is a good story in its own right, with no need for the large number.
And the other two companies demonstrate the same principle in two other ways:
At Marvell, non-GAAP net income - $865.9 million - is 2.81 times GAAP net income of $308.0 million. The gap, $557.9 million, is stock compensation and amortisation of intangibles.
And at SentinelOne the direction is reversed: the non-GAAP operating margin rose from 2% to 10%, while the GAAP net loss widened from $72.0 to $93.4 million. The reason is a $24.4 million restructuring expense, stripped out of one measure and not the other.
And what unites all three is not an accounting trick. All three companies published every number, clearly, in the same release. The difference is only which line the reader stops at.
And the Consumer, Split in Two
Two retailers without a separate article together give the sharpest picture of the quarter:
| The quarter | |
|---|---|
| Ulta Beauty | Sales +8.9% · comparable sales +3.8% · EPS +13.3% to $6.55 |
| Gap | Sales -2% · comparable sales -1% |
Ulta raised guidance and increased its buyback programme to $1.8 billion, from $1.5 billion.
Gap, by contrast, reported sales below expectations but gross margin strength that carried it past profit expectations, and also raised its earnings-per-share outlook. Management notes another quarter of double-digit comparable sales at the Gap brand, and "work to do" at Old Navy - where a new chief executive, Michael Francis, has been appointed.
And the two of them, together with Dollar General and Best Buy, paint a consumer who is neither collapsing nor booming: they are coming to the store - traffic at Dollar General rose 2.0%, a fifth consecutive quarter - and buying cosmetics and electronics, and less basic apparel.
And the Rest That Filed the Same Night
Fourteen more companies filed the same day. I name them without figures - I have not read their reports in depth, and I do not present a number I have not verified:
Banco BBVA Argentina · Baozun · Bilibili · Build-A-Bear · C3is · Canadian Solar · Gaotu · HealthEquity · Materialise · Medirom · Newegg · PagerDuty · Standard Nuclear · WesBanco.
What I Take From This Night
| Cost of revenue in software | Three companies, the same direction - the measure for coming quarters |
| Tariffs in the accounts | They enter and leave; they must be stripped out |
| Below the operating line | Tax, stock compensation, restructuring - three ways to change a bottom line |
| Financing priced by customer | 6% against 9% at IREN - the most direct pricing of AI risk I have seen |
| The consumer | Comes to the store, chooses the category |
הזווית שלי
דעה אישית של אילן אברמוב - לא ייעוץ ולא המלצה
What makes a night like this useful is not any single report, but what you see when you lay them side by side.
Because separately, each of them sounds excellent. Marvell: record revenue. Rubrik: 38% growth. SentinelOne: record profitability. Affirm: $1.9 billion of profit. Best Buy and Dollar General: raised guidance. And all those sentences are true.
Together they say something else: this quarter was better on the top line than in the lines in the middle.
And what I try to hold after a night like this is one simple question, which I ask of every report: where was the profit created? In cost of goods sold, in operating expenses, in financing, in tax, or in a decision about what to strip out. And only the first answer describes a business.
That does not make these reports bad - some are genuinely excellent. IREN executed a real crossover. Dollar General has brought people into the store for five straight quarters. Rubrik crossed into meaningful positive cash flow. Those are real things.
But if I had to pick one figure from this night to keep, it would be the three cost-of-revenue lines. A falling gross margin at three fast-growing technology companies, all at once, is exactly the kind of thing that looks negligible in a single quarter and very significant after four.
(It is important to stress: this is my personal opinion only, and nothing herein constitutes a recommendation to take any action.)
