Marvell Just Printed a Record Quarter and the Stock Barely Moved — That Tells You Everything

Marvell Just Printed a Record Quarter and the Stock Barely Moved — That Tells You Everything

Let me tell you something that's been sitting with me all week. Two chip companies, two record quarters, two days apart. One made ninety-six billion dollars in a single quarter — the most profitable company on earth by some measures — and its stock fell anyway. The other just posted its best quarter ever, raised its full-year guidance, and the market's reaction was basically a shrug.

That second company is Marvell. And the shrug tells you more about where this AI cycle actually is than any single number in the release.

The Record Quarter Nobody Cheered

Marvell reported fiscal second-quarter 2027 results Thursday after the close, and the headline numbers were genuinely good. Record revenue of $2.739 billion, up 37% year over year. Data center revenue growth accelerated to 46% — that's the engine, and it's getting stronger, not weaker. Non-GAAP earnings came in ahead of the 93 cents a share the street was modeling.

Then came the part that should have made the stock pop: guidance. Marvell guided fiscal Q3 to $3.15 billion in revenue, plus or minus 5%, with non-GAAP EPS of $1.10. That's a 15% sequential jump — the first time this company is knocking on the door of $3 billion in a single quarter. And management said they raised their revenue outlook for fiscal 2027 and fiscal 2028 compared with what they told us last quarter.

CEO Matt Murphy said the quiet part out loud: "Marvell delivered record second-quarter fiscal 2027 revenue of $2.739 billion, up 37% year over year, driven by continued strong demand across our Data Center portfolio, where revenue growth accelerated to 46% year over year."

So what did the stock do? Up about a percent in after-hours — a little over $244, up from a close around $241. Meanwhile the options market had been pricing a $22 billion swing on this print. That's a double-digit move on a company this size. The market loaded the dice for a huge reaction and then basically shrugged.

Same Disease as Nvidia

You saw the same thing the night before with the big guy. Nvidia printed $96.2 billion in revenue, up 106%, with data center revenue at $89 billion, up 117%. It guided Q3 to $108 billion. And the stock fell anyway — that's five straight post-earnings declines now.

Let me be clear about what that means. It doesn't mean these companies are broken. It means the market has already priced in the future, and now it's demanding proof the future is arriving on schedule. When a stock trades at 84 times trailing earnings after a 231% one-year run — which is exactly where Marvell was sitting — the quarter itself almost doesn't matter. The question isn't "did you beat?" It's "did you beat by enough to justify what I already paid?"

Marvell beat. It raised. It accelerated. And it got a percent. That's the same disease Nvidia has, just at a different scale. The market is no longer rewarding good news. It's only punishing bad news — and punishing it brutally.

The $120 Billion Fine Print

Now here's the part that actually matters, and it's the reason this quarter was about more than the numbers. Three weeks ago Marvell handed Google a warrant to buy up to 58.97 million shares at $206.58 apiece — roughly $12.2 billion worth of stock, about 7% of the company. If fully exercised, it makes Google Marvell's fifth-largest investor.

And what did Google give in return? The right to up to $120 billion in cumulative revenue through fiscal 2033. Let me put that number in context: Marvell's entire revenue last fiscal year was about $8.2 billion. The Google deal, if it fully lands, is roughly fifteen years of today's Marvell, compressed into seven.

But here's the fine print Wall Street glosses over: that warrant doesn't vest just because time passes. It vests in stages, tied to qualifying product revenue. Google doesn't pay Marvell $120 billion because it likes the logo. Google pays it because Marvell keeps shipping chips Google actually wants to buy. The warrant is an option on a relationship, not a check that cleared.

That's the key. The market is pricing Marvell like the $120 billion is already in the bank. The company is guiding like it's on the way. But between now and fiscal 2033, every single quarter has to deliver. One slip — one Google internal decision to slow custom silicon, one TPU generation that doesn't hit, one budget review — and the warrant math gets ugly fast.

What the Warrant Actually Changes

Because let's not pretend this is just a Marvell story. This is a Broadcom story wearing a Marvell nametag.

Broadcom is the elephant in custom silicon — a $1.95 trillion company that has been Google's go-to ASIC partner for years, with TPU shipments projected to hit 7 million units a year by calendar 2028 and AI revenue estimated to grow from roughly $44 billion this calendar year to $78 billion next. Google handing Marvell a warrant-linked custom chip deal is the first time any hyperscaler has tied equity upside to silicon delivery — and it's a direct shot across Broadcom's bow.

The message is unmistakable: Alphabet doesn't want one custom-silicon supplier holding the keys. It's splitting the work. It's using Marvell to build a second lane so Broadcom can't hold it hostage on price, on capacity, on roadmap. That's the exact same playbook Amazon used with Trainium to loosen Nvidia's grip, and the same playbook Google used with TPUs in the first place.

For anyone who runs infrastructure, this is a structural shift, not a stock story. The AI buildout is no longer a one-vendor GPU game. It's becoming a multi-vendor custom-silicon game, with warrants and financing and long-term supply agreements standing in for handshake deals. And every one of those vendors is now financially welded to the hyperscaler it serves.

What This Means for Independent Hosting Providers

First — watch the custom-silicon split like a hawk. When hyperscalers diversify chip suppliers, the availability picture for GPUs changes. Google leaning on Marvell means Broadcom capacity moves elsewhere, and Nvidia allocation pressure shifts. That affects lead times for the rest of us, even if we never touch a TPU.

Second — don't confuse announced with realized. The $120 billion is potential. The warrant is conditional. Every contract in this industry now has an "if the buildout keeps going" clause buried in it, and you should model your own capacity plans on the realized numbers, not the press-release numbers.

Third — the warrant is the tell. When a chip vendor hands a customer equity to lock in delivery, that's not a partnership, that's a bet on the buildout's continuity. Marvell just bet the farm that AI infrastructure keeps growing for the next seven years. Your supplier relationships are implicitly making the same bet. Diversify them.

Fourth — use chip earnings as your pricing signal. When a semiconductor company raises guidance two fiscal years out, the equipment and connectivity pricing pressure you'll feel in 12 to 18 months is being set right now. Lock what you can while the window is open.

The Structural Reality — Silicon Is Now a Subscription With Warrants

Step back and look at the pattern. Nvidia is financing $500 billion of compute purchases. Marvell is giving Google warrants to guarantee a decade of revenue. CoreWeave gets Meta as an anchor tenant. OpenAI leases a data center with Nvidia guaranteeing $105 billion of the lease. The entire AI supply chain is now wiring itself together with cross-equity stakes, vendor financing, and performance-based options.

That's what a real structural buildout looks like — not a bubble of hype, but a web of contracts where every player has locked itself to every other player. The problem is that webs like this cut both ways. When they work, they compound. When they break, they break all at once, because everyone is holding everyone else's paper.

Marvell's quarter was good. The guidance was better. But the market's shrug is the honest signal. We've reached the phase of the cycle where even record quarters get priced like disappointments — and that's exactly when the real risk shows up, not in the numbers, but in the gap between what's promised and what gets delivered.

The Bottom Line

Marvell didn't get punished for a bad quarter. It got ignored for a good one. That's what a market that has already priced in the future looks like — and it's the most dangerous moment in this cycle, because it means every future miss will be paid for in full.

Run your business on realized revenue, not promised revenue. Watch the custom-silicon lanes. And remember: the warrant is the whole game now. The AI buildout doesn't need more cheerleaders — it needs more delivery.

— Allan Ali, Founder

This article was produced with AI-assisted research and editorial support. Sources: Marvell Technology Q2 FY2027 press release and 8-K, Reuters, The Motley Fool, Barron's, TechPowerUp, 247wallst, StockTitan.