Marvell published its second-quarter report for fiscal 2027, and revenue is a record: $2.739 billion, up 37%.
Management raised the outlook for the next two years, and data centre revenue accelerated to 46% growth. And yet two lines in the report tell a second story - and they are the two lines I read first.
Errors or inaccuracies are possible. Spotted something that looks wrong? Write to me and I will correct it.
What Marvell Does
Marvell makes chips for data infrastructure. It does not sell to the end customer and does not manufacture itself - it designs, and hands the design to a foundry.
And its two product families differ in their economics, which explains the whole report.
The first - connectivity. Standard chips that move data inside a data centre and between it and the world: network controllers, optics, storage interfaces. One product sells to many customers, so the margin is high.
The second - custom silicon, what the industry calls ASICs. A giant customer - a hyperscaler - orders a chip designed for it alone, and Marvell designs it end to end.
And that is an entirely different business: the revenue is large and predictable, but the margin is lower - because a substantial part of the value sits in manufacturing and memory bought from outside, not in Marvell's intellectual property.
So when management talks about "a significant acceleration in our Custom business in the second half", it is describing both growth and pressure on the margin. Both come from the same place.
The Quarter
| $ millions | The quarter | Year ago | Change |
|---|---|---|---|
| Revenue | 2,739.3 | 2,006.1 | +36.6% |
| Cost of goods sold | 1,283.7 | 995.5 | +28.9% |
| Gross profit | 1,455.6 | 1,010.6 | +44.0% |
| Research and development | 741.1 | 519.0 | +42.8% |
| Selling, general and administrative | 257.6 | 192.8 | +33.6% |
| Operating income | 459.7 | 290.1 | +58.5% |
| Interest expense | -61.6 | -51.9 | |
| GAAP net income | 308.0 | 194.8 | +58.1% |
| GAAP earnings per share | $0.33 | $0.22 | |
| Non-GAAP net income | 865.9 | ||
| Non-GAAP earnings per share | $0.94 | ||
| Cash flow from operations | 605.5 | 461.6 | +31.2% |
The First Line I Read: The Gross Margin
The GAAP gross margin rose - 50.4% to 53.1%. The non-GAAP gross margin fell - 59.4% to 58.9%.
Both numbers are correct, because they measure different things.
The gap between them is amortisation of acquired intangible assets - $142.7 million this quarter, which is 5.2% of revenue. And that amortisation shrinks over time by its nature, as the asset is written down. So the GAAP margin rises even when the business itself has not improved - simply because an old amortisation burden is fading.
The non-GAAP margin is the one that measures the economics of the product, and it fell.
And what matters more than the fall itself is the guidance: for next quarter Marvell guides a non-GAAP margin of 57.5% to 58.5% - a further step down, and at the midpoint a drop of almost a full point from this quarter.
This is happening while revenue grows 37% and accelerates. And those two facts go together rather than contradict each other: the growth is coming from the business with the lower margin.
And this is not a Marvell-only phenomenon. The same night, Rubrik and SentinelOne reported too - and at all three, without exception, cost of revenue grew faster than revenue. I wrote about this at length in the day's summary.
The Second Line: The GAAP to Non-GAAP Gap
GAAP net income: $308.0 million. Non-GAAP net income: $865.9 million.
The gap is $557.9 million - that is, the non-GAAP figure is 2.81 times larger, and the gap alone equals 20.4% of revenue.
And what sits inside that gap? Mainly two things:
| The quarter | Year ago | |
|---|---|---|
| Stock-based compensation | $326.2m | $153.6m |
| Amortisation of intangibles | $214.9m | $243.7m |
And note the opposite directions. The amortisation is falling - an old asset being written down. Stock compensation more than doubled - from $153.6 to $326.2 million, which is 11.9% of revenue.
Stock compensation is the expense easiest to argue about. On one hand it is not cash: the company did not pay $326 million out of the till. On the other it is a real cost to the shareholder - it dilutes them, and when you strip it out of earnings without stripping the dilution out of the share count, you count the same benefit twice.
And Marvell, to its credit, buys back stock - $200 million in the quarter. But stock compensation is 1.6 times the buyback, so the buyback offsets part of the dilution and not all of it.
What Is Good in This Report
And three things really are good, and the criticism above must not obscure them.
The first - the acceleration is real. Revenue rose 13.3% in a single quarter, and next quarter's guidance is $3.150 billion - another 15% sequentially. A company of this size accelerating is an event.
The second - the cash flow. $605.5 million from operations in the quarter, which is 22.1% of revenue and nearly twice accounting net income. Cash flow is unaffected by the argument over stock compensation - that is money coming in.
And the third - data centre accelerated to 46%. That is the division that decides, and the acceleration in it is the important number in the report.
And the Balance Sheet, Which Is the Hard Part to See
| $ millions | 1 August 2026 | 31 January 2026 |
|---|---|---|
| Cash | 3,932.8 | 2,638.8 |
| Goodwill | 13,873.9 | 11,062.2 |
| Intangible assets | 2,346.6 | 1,754.7 |
| Total assets | 27,554.6 | 22,285.3 |
| Long-term debt | 4,962.9 | 3,970.8 |
| Shareholders' equity | 18,531.6 | 14,308.4 |
Goodwill and intangible assets together: $16.2 billion.
That is 58.9% of the balance sheet, and 87.5% of shareholders' equity.
In other words almost all of Marvell's equity is an accounting record of past acquisitions, not tangible assets. And that is not a flaw - this is what a chip company that grew by acquisition looks like, and so do its competitors. But it does say something about what happens if one of those acquisitions fails to meet expectations: a goodwill write-down costs no cash, but it erases equity.
And in this half year goodwill grew by $2.8 billion - an acquisition of $1.27 billion in cash, alongside a preferred stock issuance that brought in $2.0 billion in March 2026. The company is growing on new money too, not only on profits.
What I Will Check Next Quarter
| The non-GAAP gross margin | 58.9% today, guided to 57.5%-58.5% - whether the decline stops |
| The custom silicon mix | The acceleration there is both the growth and the margin pressure |
| Stock compensation | 11.9% of revenue, and more than double last year |
| The data centre run rate | 46% today - that is the division that decides |
| Buyback versus dilution | $200m a quarter against $326m of compensation |
| Investor Day on 6 October | Where the long-term strategy will be presented |
הזווית שלי
דעה אישית של אילן אברמוב - לא ייעוץ ולא המלצה
This report is an excellent example of how a company can be both a great growth story and an expensive share - and that those are two separate questions.
What Marvell showed here is demand. Record revenue, sequential acceleration, a raised outlook two years forward, and management talking about "exceptionally robust" AI bookings. There is no demand weakness here.
What it did not show is pricing. The real gross margin is falling, and the guidance points to a further decline. And that is exactly the pattern of a supplier growing inside very large customers: hyperscalers buy in enormous volume, and whoever buys in enormous volume gets a price.
So the question I hold about Marvell is not "will demand continue" but "how much of that demand reaches the shareholder". The answer this quarter is 11.2% - that is accounting net income as a share of revenue. And anyone reading only the $0.94 non-GAAP figure sees an entirely different business from whoever reads $0.33.
And what I would look for on the earnings call is not the growth rate but the split between connectivity and custom. That is the number that predicts where the margin goes, and it does not appear in the press release.
And one last thing on context: Nvidia posted a gross margin of 75.0% this week. Marvell operates in the same supply chain, sells to the same customers, and grows at a similar rate - and its non-GAAP margin is 16.1 points lower. The difference is not in execution quality. It is in who holds the position of power opposite the buyer.
(It is important to stress: this is my personal opinion only, and nothing herein constitutes a recommendation to take any action.)
