Marvell Shares Close Down 10.3% After Fiscal Q2 Beat and Raised Fiscal 2028 Revenue Outlook

The Number Everyone Agrees On, and the One Nobody Can Check Yet

Marvell Technology beat its own forecast. It beat what Wall Street expected. And its stock dropped 10.28% anyway, closing at $216.62 on Friday, August 28^[6]. That combination sounds like a contradiction, but both halves of it are true, and the gap between them is the whole story.

The company reported fiscal second-quarter revenue of $2.739 billion, a record, up 37% from a year earlier^[1]. That came in $39.0 million above the middle of the range Marvell itself had promised back in May^[1]. Adjusted profit was $0.94 a share, two cents ahead of what analysts expected^[5]. Marvell even raised its own long-range forecast, pushing its fiscal 2028 revenue target up to about $18 billion from roughly $16.5 billion^[2]. None of that is in dispute. What's in dispute is why a company that just did all of that lost a tenth of its market value in a single day.

Why a Chip That Sells for Less Can Still Be Worth More

Start with the part everyone agrees on but few outlets explained well: gross margin, and why it's falling even as sales grow. Marvell's non-GAAP gross margin was 58.9% last quarter, and the company guided the current quarter down to a range of 57.5% to 58.5%^[10]. Gross margin is simply the share of each sales dollar left after the direct cost of making the product. A shrinking margin on rising revenue sounds like a red flag, but here it isn't necessarily one.

The reason is what Marvell is actually selling more of: custom chips built for one buyer, in this case Google. A custom chip is designed to one company's specifications, and much of its final price is memory and packaging that Marvell itself buys and resells rather than manufactures. Big buyers with that kind of leverage negotiate hard on price. So as custom chips become a bigger share of Marvell's business, the average margin naturally drops, even though total dollars sold keeps climbing^[10].

Both sides accept that mechanism. They disagree on whether it's a fair trade. Management's view is that a lower percentage on a much larger and more predictable pile of sales is a good deal, because these are multi-year commitments, not one-off orders. Skeptics counter that when a stock is priced for perfection, a shrinking margin is exactly the kind of detail that tips the scale, because there's no cushion left to absorb bad news^[3][10].

The $120 Billion Number That Isn't a Forecast

The other flashpoint is Marvell's July 29 agreement with Google, which covers custom chips for Google's in-house AI processors, known as TPUs^[9]. As part of the deal, Marvell gave Google a warrant, which is simply the right to buy stock later at a price fixed today. Google's warrant lets it buy up to 58,970,907 shares at $206.58 each, worth up to about $12.2 billion at that price^[9].

Here's the catch that headlines tend to skip: Google doesn't get all of it automatically. The warrant vests in slices, one slice for every $500 million in chips Google actually buys, with purchase targets running out through fiscal 2033^[9]. Multiply the full run of tranches out and you get the number that shows up in bullish headlines, including one Benzinga cited from an analyst putting long-term revenue potential at $120 billion^[12]. That figure is a ceiling if every milestone is hit for the next seven years, not a number Marvell is currently earning.

This is also why Marvell isn't just gaining a customer. It's giving that customer a growing ownership stake, tied directly to how much the customer spends. Fully exercised, the warrants would make Google Marvell's fifth-largest shareholder^[9]. Google gets a cheaper path to AI chips than buying from Nvidia, whose processors are pricier and often backordered; Marvell gets a locked-in buyer building its whole AI chip stack around Marvell's parts^[5][9].

A Company That Can't Say Yes Fast Enough for the Market

The genuinely contested question is timing, and it's where the earnings call got interesting. Investors wanted the Google deal to show up in that new fiscal 2028 number, roughly $18 billion. Instead, CEO Matt Murphy told investors the fiscal 2028 target already includes some Google revenue, but that the deal would contribute "much more significantly" starting in fiscal 2029^[2]. Marvell's fiscal year runs ahead of the calendar, so its fiscal 2029 begins in 2028 — this is a one-year shift in when the big money lands, not the two-year gap it can look like if you drop the word "fiscal."

Bulls hear a management team being conservative about a deal it's calling "game-changing," with what Murphy described as "a lot of upside bias" for the years ahead^[11]. Skeptics hear a huge headline number that keeps sliding into a year nobody can verify yet. Both readings use the same sentence from the same call. Neither can be proven wrong until fiscal 2029 actually arrives.

There's a third piece worth sitting with: how far the stock had already run. Marvell was up about 188% for the year and roughly 30% in August alone, much of that gain arriving after the Google deal was first announced^[8]. Some analysts pointed out that at roughly 84 times trailing earnings, Marvell's price was already betting heavily on future growth, using a measure called the PEG ratio — price-to-earnings divided by expected growth — that came in well above the level where a stock is considered fairly priced for its growth rate^[3]. When a stock has already priced in a rosy future, even genuinely strong results can disappoint simply because they were the baseline, not the surprise.

How Newsrooms Told Three Different Versions of the Same Afternoon

Coverage of the drop itself splintered, partly because outlets measured the stock at different moments. CNBC's headline read "Marvell stock tumbles 6% as outlook underwhelms despite Google deal," capturing an intraday snapshot rather than where shares ended the day^[2]. The word "despite" quietly assumes a beat should have produced a rally — an assumption the stock's enormous run-up this year makes questionable. The Motley Fool published two separate pieces the same day, one citing an 8% fall and another headlined "Marvell Slides 10%," for the identical trading session^[3][7]. Only the closing number, down 10.28%, is fully settled^[6].

Retail-trading outlets leaned into the upside case. Benzinga's headline led with the $120 billion long-term figure before any mention of the vesting conditions attached to it^[12], while TheStreet's framed those same conditions as "fine print" — even though Marvell disclosed the warrant structure publicly when the deal was announced^[9]. An Indian retail-investing platform, INDmoney, covered the story for its own U.S.-stock-holding readers as a straightforward explainer, notably without any of the U.S.-China chip rivalry angle that often colors Asian coverage of the semiconductor industry^[15].

What Gets Settled Next, and When

Strip away the framing choices, and what actually happened is this: Marvell sold $2.739 billion worth of chips, 37% more than a year ago, and generated $605.5 million in cash from its operations^[1]. It has a signed, binding contract with Google covering several categories of custom silicon^[9]. None of that changed on August 28. What changed was how much investors were willing to pay for it, from around $241 down to $216.62 in one trading session^[6].

The industrial fact underneath all of it isn't disputed either: the biggest cloud companies are building their own AI chips because it's cheaper and faster than waiting in line for Nvidia's^[5]. That demand is real, and it's under contract. What isn't yet knowable is whether Google's spending lands on the schedule Murphy described, in fiscal 2029, or slips further out — a question that won't be answered by anything Marvell says on a call. It will be answered by what Google actually buys.