Jensen Huang walked onstage at Computex 2026 in Taipei on Monday, stood next to Marvell CEO Matt Murphy, and said six words that added more than $47 billion to Marvell's market value overnight. "The next trillion-dollar company, ladies and gentlemen."
Marvell Technology shares surged approximately 25 percent in premarket trading on Tuesday June 2, pushing toward $273 from a prior close in the $219 range. The company's market capitalization before the move stood at just under $192 billion.
At $273, the implied cap approaches $240 billion. Huang's target of a trillion dollars is approximately five times away from where the stock is trading this morning, and investors are treating the distance as an invitation rather than a warning.
The stock had already gained more than 130 percent year-to-date before Tuesday's move. It is now one of the single best-performing semiconductor stocks in the world in 2026, and Jensen Huang, whose endorsements have been among the most reliably powerful stock catalysts of the AI era, has now staked his credibility on its trajectory.
What Did Jensen Huang Say?
The Computex trade show in Taipei is one of the technology industry's largest annual gatherings, and Jensen Huang has used the platform before to make announcements that move markets.
His appearance onstage with Matt Murphy on Monday was not a casual fireside chat. Nvidia had already invested $2 billion in Marvell on March 31, a move that sent Marvell shares up 11 percent on the day of that announcement.
Huang appearing in person on Marvell's stage at the world's biggest tech trade show and calling the company the next trillion-dollar chip company is a different category of endorsement entirely.
He explained the reasoning in specific terms that go beyond cheerleading. "When you take a computing problem, and you disaggregate it into a lot of parts, and you distribute it across the entire data center, what's necessary is connectivity," Huang said. "That's the reason why Matt's doing so well. That's the reason why Marvell is so essential."
The connectivity argument is the specific insight that makes Marvell's business model something more than a custom chip story. As AI systems scale, from a few thousand GPUs to hundreds of thousands, from a single data center to a distributed computing fabric spanning multiple facilities, the bottleneck is not the chips themselves. It is the speed at which data moves between them.
A hundred thousand Nvidia GPUs working on a single training run need to communicate with each other thousands of times per second. The networking fabric, the interconnects, the optical links that allow all of those chips to share state and synchronize computation, that is the infrastructure that Marvell builds.
Huang's framing was also a statement about where the AI industry stands. "The AI industry has entered a stage of full commercialization," he said at Computex. "Computing power is revenue, and computing power is profit."
That sentence, computing power is profit, is the summary of why companies like Marvell, which sits between the chip and the network, are becoming indispensable to the hyperscalers who are converting AI compute into commercial products.
What Marvell Actually Makes And Why It Matters Now
Marvell Technology is not a household name. It does not make the Nvidia GPUs that appear in news coverage of the AI boom. It does not make the DRAM that Micron sells or the CPUs that AMD and Intel compete over.
What it makes are the pieces that allow all of those components to work together at the scale that modern AI requires.
The first category is custom AI chips, what the industry calls XPUs or custom accelerators. The major hyperscalers, Amazon, Microsoft and Google, do not want to be entirely dependent on Nvidia hardware for their AI infrastructure.
Each of them has invested heavily in custom silicon designed specifically for their own AI workloads. Marvell designs and produces those custom chips for Amazon's Trainium and Inferentia programs, for Microsoft's Maia accelerator and for Google's TPU family.
The custom chips business is the fastest-growing segment in Marvell's portfolio and is on track to exceed $10 billion in annual revenue by fiscal 2029, a number the company confirmed in its most recent guidance.
The second category is networking and connectivity. As data centers fill with AI accelerators that need to communicate continuously at extraordinary bandwidth, the switching, routing and interconnection fabric becomes the rate-limiting factor in the system's performance.
Marvell's networking chips are deployed in the infrastructure that connects GPUs within a server, servers within a rack, racks within a data center and data centers across a network.
The third category, and the one that Huang specifically highlighted at Computex, is optical interconnects. Moving data between chips using light rather than electricity is more efficient at the distances and data rates that modern AI data centers require. Marvell is one of the leaders in silicon photonics, the integration of optical components with standard chip manufacturing processes, and Nvidia's investments in photonics technology are directly connected to the Marvell partnership.
Nvidia has also invested $2 billion each in Lumentum and Coherent, two other optical technology companies, as it builds out an interconnected ecosystem of AI infrastructure suppliers.
The NVLink Fusion Partnership
The structural relationship between Nvidia and Marvell is not simply a financial investment. NVLink Fusion is the technical architecture that makes them partners at the product level.
NVLink Fusion is Nvidia's rack-scale interconnect technology, the system that allows many different types of AI accelerators, not just Nvidia's own GPUs, to work together inside a Nvidia-enabled data center infrastructure.
By integrating with NVLink Fusion, Marvell's XPUs can be deployed alongside Nvidia GPUs in the same rack, using the same networking fabric, managed by the same software stack.
For hyperscalers who want to use a mix of Nvidia hardware and custom silicon in their data centers, NVLink Fusion is what makes that architecturally coherent.
For Marvell, being the designated custom silicon partner inside Nvidia's ecosystem is approximately the best competitive position available in the AI chip market, one that provides design wins at scale with the world's largest cloud providers while operating under the protective umbrella of the most influential infrastructure company in AI.
The $2 billion Nvidia investment in Marvell, combined with Huang's trillion-dollar endorsement at Computex, is Nvidia publicly signaling that it intends the Marvell relationship to be central to its ecosystem strategy for the foreseeable future. That signal is what investors are responding to on Tuesday morning.
The Numbers Behind The Trajectory
Marvell reported record revenue of $2.418 billion in Q1 FY2027, with operating cash flow of $638.8 million.
Data center revenue now accounts for 75 percent of the company's total revenue, up from approximately 50 percent just two years ago. The revenue mix shift is the clearest possible indicator of where the growth is coming from and where it is going.
In full year FY2026, Marvell generated $8.2 billion in total revenue, a record. With the custom chips business alone forecast to surpass $10 billion in fiscal 2029, and the networking and optical interconnect businesses also growing, the trajectory from $8.2 billion to something meaningfully higher within three years is supported by contracted hyperscaler relationships rather than aspirational projections.
The gap between $192 billion and $1 trillion is real and substantial. A company does not move from one to the other in a single calendar year.
Jensen Huang does not make trillion-dollar predictions about companies without a specific thesis, and his thesis for Marvell is the thesis of the AI infrastructure buildout itself: that the world is going to need more connectivity, more custom silicon and more optical efficiency as AI scales from its current state to something orders of magnitude larger, and that the company best positioned to provide those three things in an integrated fashion with Nvidia's blessing is Marvell Technology.
The market believes him. By approximately $47 billion this morning.
Please leave a comment: Your opinion is important to us!
Add Your Comment