Nvidia CEO Jensen Huang called Marvell Technology the next trillion-dollar company on June 2, and the stock jumped 33% the same day—its biggest one-day gain ever. By Thursday it hit a new high near $329. Friday it cratered 16% in a sector sell-off. Monday it bounced on S&P 500 inclusion news, then gave it all back Tuesday. A week's worth of conviction evaporated and returned and evaporated again, and the only constant was the line from Computex that started it all.
The move is a textbook case of the market pricing the outcome before anyone's executed the work. Marvell's market cap sits at $232 billion; a trillion-dollar target implies roughly 4x upside from here. The stock trades at a price-to-earnings ratio above 90 on a trailing basis. The company posted $2.4 billion in revenue last quarter, which annualizes to less than $10 billion—meaning the valuation is pricing not just growth but a transformation in scale the company hasn't delivered yet. Management expects $16.5 billion in fiscal 2028 revenue, impressive but still a fraction of what a trillion-dollar valuation would traditionally demand. The gap between the current business and the price is a bet that Huang is right and that Marvell will capture it all.
What makes Huang's call credible is that he's not guessing—he's already written the check. Nvidia committed $2 billion to Marvell in March, and Huang explained the thesis onstage: disaggregated computing runs across enormous clusters, and connectivity—Marvell's specialty—is what makes it possible. Marvell designs high-performance chips for cloud computing, AI, enterprise networking, 5G, and automotive systems, and the company is solving a critical bottleneck in AI infrastructure, with demand for optical networking components set to exceed supply substantially. If the hyperscalers keep building at the current pace, Marvell sits at a choke point in the architecture that has to scale or the whole thing stalls.
The problem is the market's pricing the scale-up as if it's already happened. Marvell expects a 70% jump in its data center interconnect business this year, $600 million in switching-product revenue in fiscal 2027 rising to $1 billion in fiscal 2028, and custom silicon revenue growth accelerating from 20% in fiscal 2027 to over 100% in fiscal 2028. Those are real numbers, and they're big. But they're also guidance, not results—and the stock's pricing them in at a multiple that assumes no execution risk, no hyperscaler capex slowdown, no competitive encroachment, and no margin pressure as volumes ramp. Three-quarters of Marvell's revenue now comes from data centers, so its fortunes rest on a small group of large cloud customers and their willingness to keep spending on custom AI chips; if that spending slows or a major customer designs a chip without Marvell in it, the growth story cools fast.
The tape is telling you exactly who's holding the risk. A third of the float changed hands in a week, the stock whipsawed 49 percentage points peak-to-trough, and the only new information was an endorsement and an index inclusion. The buyers at $329 on Thursday are underwater if they held through Friday's close. The sellers into Friday's panic left money on the table when Monday's S&P news pushed it back up. Nobody's trading fundamentals here—they're trading the narrative, and the narrative is one man's view of what happens if everything goes right for the next several years.
Huang may be the single most credible voice in AI infrastructure, and his $2 billion stake says he's not posturing. But credibility doesn't eliminate execution risk, and a trillion-dollar market cap four times higher than today's price is not a trade—it's a multi-year thesis that requires Marvell to keep every hyperscaler design win, scale production without margin compression, and grow revenue faster than the market's already pricing. The stock's at 90 times earnings. That's not paying for what Marvell is. It's paying for what Huang says it will become—and hoping nobody else gets there first.