Jensen Huang walked onto Matt Murphy's Computex stage, spent roughly 10 minutes talking, and added $75bn to Marvell Technology's market capitalisation.
Shares surged as much as 30% on Tuesday, pushing Marvell's valuation towards a quarter of a trillion dollars, after Huang declared the semiconductor and networking company will be "the next trillion-dollar company."
The market did not pause to ask questions. It just bought.
Talking his own book?
Nvidia took a $2bn stake in Marvell three months ago as part of a broader partnership centred on NVLink Fusion, the interconnect technology that links processors inside AI data centres. When Huang calls Marvell a future trillion-dollar company, he is talking about a business in which Nvidia is a significant shareholder.
That $2bn investment is now worth considerably more than it was on Monday morning. Huang's endorsement is not disinterested analysis. It is the CEO of the world's most valuable public company promoting a company in which he has a material financial position.
Jim Cramer said the quiet part out loud on Tuesday: "These are big moves, and they're not based on anything other than one person saying it." He asked whether the reaction suggested things were "a little too euphoric."
The question answers itself.
The substance behind the showmanship
Strip away the theatre and Marvell's transformation, well, it's been significant. In 2016 the company generated $2.3bn in annual revenue, with less than 10% from data centres. Revenue is projected to reach $16.4bn next year, with data centres accounting for more than 75% of the total.
The stock had already gained more than 158% in 2026 before Tuesday's move.
Marvell designs the networking and connectivity chips that move data between processors inside AI data centres. As models grow larger and training clusters scale to tens of thousands of GPUs, the interconnect fabric becomes as important as the processors themselves. Marvell also designs custom ASICs for hyperscalers including Amazon, Microsoft and Google.
Huang's argument is that AI computing is shifting towards disaggregation and distribution, requiring the kind of high-speed connectivity that Marvell specialises in. The transition from copper cabling to silicon photonics, which Murphy highlighted during the keynote, is a structural tailwind for the company.
That is a credible investment case. It did not need Jensen Huang's personal endorsement to stand up.
Under-appreciated, or overhyped?
When the CEO of the most influential company in technology makes a public price prediction about a company in which he holds a $2bn stake, it matters whether he is sharing genuine insight or performing investor relations for his own portfolio.
Both things can be true simultaneously. Marvell may well reach a trillion-dollar valuation. The networking layer of AI infrastructure is genuinely underappreciated relative to the GPU layer. The fundamentals support a premium.
But a 30% single-session move driven by one man's onstage comment is not price discovery. It is a market that has decided Jensen Huang cannot be wrong, and is pricing accordingly.
That kind of faith has a name. It is not usually "investing."
