During the AI boom years, the Computex convention in Taipei had a unique atmosphere: filled exhibition halls, the sound of demonstration servers, executives wearing black T-shirts giving keynote addresses to rooms full of people simultaneously peering at their laptops and phones. In that context, Jensen Huang and Marvell CEO Matt Murphy took the stage together and declared Marvell Technology to be “the next trillion-dollar company.”
Huang did this with the serene assurance of someone who has been correct about this industry enough times to deserve it. Traders had heard enough in a matter of hours. In one of the most remarkable single-day movements a company of that magnitude has produced in recent memory, Marvell’s shares surged more than 30% in a single session, reaching all-time highs and pushing the company’s market capitalization above $250 billion.
It wasn’t an unexpected recommendation. Nvidia has previously declared a $2 billion strategic investment in Marvell, with the goal of working together to develop the bespoke processing units and optical interconnects that are becoming increasingly important to large-scale AI infrastructure. Once you comprehend the architecture of a contemporary AI cluster, the reasoning becomes clear:
Nvidia manufactures the GPUs that perform the computation, but when thousands of those chips are used simultaneously, as is now necessary to train the largest models, the bottleneck shifts to the speed at which data can flow between them. The networking chips that manage such movement are manufactured by Marvell. In that context, Huang’s description of Marvell as essential is more of a technical evaluation made in public than a compliment.
Investors appear to have faith in Huang. The man has shape, so that’s noteworthy. For a number of years, his public support for businesses and goods in the AI supply chain has served as a sort of unofficial quality stamp, and the market has learnt to factor that in. At least in comparison to the typical churn of analyst upgrades and conference soundbites, the signal-to-noise ratio is abnormally high when Nvidia builds a $2 billion investment in a supplier and then puts its CEO on stage to suggest the supplier is likely to be worth a trillion dollars. Following the Computex session, Stifel reiterated its buy recommendation for Marvell and increased its price objective to $321. Others went in the same direction.
Watching this unfold gives the impression that Marvell has spent years performing genuinely challenging technical work in an area of the semiconductor stack that doesn’t make many headlines, such as high-speed optical networking, custom XPUs, and the infrastructure plumbing rather than the glamorous processing cores, and is now having that work acknowledged in the most public way possible.
The business is not brand-new. It has long been producing networking silicon for data centers. However, the AI scaling wave has transformed what was once a strong niche industry into what appears to be a key dependency, and Nvidia’s investment formalizes that link in a way that modifies the stock’s valuation.
It’s still uncertain when a trillion-dollar valuation will be reached or if the cycle of spending on AI infrastructure will continue to expand at a rate that will support it. Huang’s forecasts are given a serious hearing instead of being dismissed due to his track record, although a serious hearing is not the same as a certainty. The structural truth he described appears to be more resilient: as AI clusters expand, it becomes more difficult to overlook the businesses addressing the data mobility issue. For now, Marvell is figuring it out.
