NVIDIA CEO, Jensen Huang, the black-jacket guy, recently highlighted Marvell Technology as a company that could potentially become the next trillion-dollar company, which obviously got people’s attention. Because apparently, once a company starts sniffing around the trillion-dollar club, everyone suddenly becomes an expert on it. So, what does Huang see in Marvell? Before getting into that, let’s first understand the semiconductor industry without making your brain file for bankruptcy.
The semiconductor industry can be broadly divided into two categories. The first consists of fabless semiconductor companies, such as NVIDIA, AMD, and Marvell, which focus mainly on designing chips. The second category includes foundries, where companies manufacture chips designed by others. Industry leaders such as TSMC and GlobalFoundries operate in this segment. Basically, one company designs the brain and another company actually builds the damn thing. Chip-designing companies can outsource the actual manufacturing of their chips to specialized manufacturers. In our case, NVIDIA and Marvell both rely heavily on TSMC for manufacturing.
Some companies, like Intel, Micron, and Texas Instruments, have much more control over both chip design and manufacturing, although the exact business models differ across these companies. As you can see, both Marvell and NVIDIA operate in the fabless semiconductor world. So how the hell is Marvell different from NVIDIA?
NVIDIA is primarily known for its GPUs, which are used to train and run AI models. Marvell, on the other hand, focuses heavily on custom silicon, networking, connectivity, and other infrastructure technologies that help AI systems communicate and operate efficiently. In simple terms, NVIDIA builds a lot of the machinery doing the AI calculations, while Marvell builds some of the plumbing that helps all this machinery communicate with each other.
And when you have thousands of ridiculously expensive chips sitting in a data center, you probably don’t want them sitting there like a bunch of rich people at a party who refuse to talk to each other. NVIDIA also has a massive software ecosystem built around its hardware, making its products extremely attractive to AI developers. Marvell’s strength lies more on the infrastructure side of the business.
Marvell, with a market capitalization far smaller than NVIDIA’s, is still comparatively tiny next to the AI giant. But that is exactly what makes Huang’s prediction interesting. If Marvell eventually becomes a trillion-dollar company, it would need to grow several times from its current size. Even though NVIDIA and Marvell operate in different parts of the semiconductor industry, they overlap in areas such as AI networking, optical connectivity, and custom AI infrastructure. And as AI data centers continue to expand, both companies are fighting for a larger piece of the rapidly growing AI infrastructure pie.
This growing importance of AI infrastructure is where Jensen Huang appears to see the opportunity. Modern AI systems aren’t just one giant chip sitting in a room doing all the work. They can involve thousands of processors working together simultaneously. And for all these chips to function efficiently, massive amounts of data need to move between them at ridiculously high speeds. Even if you have the world’s fastest AI chip, it isn’t going to be very useful if it spends half its life waiting around for data.
This is where Marvell comes in. Marvell specializes in networking, connectivity, and custom silicon technologies that help AI data centers operate at scale. Its products help move data between servers, storage systems, and AI accelerators, allowing enormous AI clusters to work together as a unified system. At COMPUTEX 2026, Marvell itself emphasized the growing importance of connectivity as AI infrastructure scales.
In other words, the future of AI isn’t just about who builds the fastest processor. It is also about who can move all that data around without turning the entire data center into a traffic jam. And this is where Marvell could potentially benefit.
Another factor working in Marvell’s favor is the rise of custom AI chips. Large cloud providers are increasingly developing specialized chips designed around their own workloads. Marvell has built a business around helping customers develop custom silicon, giving it exposure to another potentially huge part of the AI semiconductor market. NVIDIA and Marvell have also partnered through NVIDIA’s NVLink Fusion platform, with Marvell providing custom XPUs and compatible networking technologies.
This helps explain why the trillion-dollar prediction gets people excited. Huang isn’t simply looking at Marvell’s current size and saying, “Yep, trillion-dollar company.” He’s betting that the company could become an important piece of the infrastructure required to keep the AI machine running.
Of course, reaching a trillion-dollar valuation isn’t going to be easy. Even at a market capitalization of roughly $250 billion, Marvell would need to increase its value by more than four times to reach $1 trillion. That’s a pretty fucking big jump. Nevertheless, Huang’s comments suggest that he believes Marvell’s position in AI infrastructure gives the company a realistic path toward that milestone over the long term.
Whether he’s right is another question entirely. Because in the semiconductor industry, predicting the future is basically trying to predict which horse wins a race while the horses are still being genetically engineered
