In This Article:
Marvell Technology is being touted as the third largest AI chipmaker after Nvidia (NASDAQ:NVDA) and Broadcom (NASDAQ:AVGO). To give readers a context of what the company does, here’s how it stands in comparison to the two competitors. Nvidia makes standard GPUs that all AI applications need to run on. These GPUs are general purpose and when used for repetitive tasks, a lot of resources on the GPU go unused. To improve on this lack of efficiency, Application Specific Integrated Circuits (ASICs) are used. These are GPUs that are built to perform the same task over and over again, like Large Language Model training that Big Tech is currently utilizing them for.
Broadcom and Marvell both make these custom chips, but Broadcom has a much larger portfolio of products. Marvell on the other hand is focused on data center infrastructure. In other words, where it lacks in GPU strength, it makes up for in infrastructure.
To put these two custom chipmakers in perspective, where Broadcom brings in over $50 billion in revenue, Marvell brings in just $5. Interestingly, Marvell also trades at a market cap that is one-tenth of Broadcom’s!
Being the smallest of the three chipmakers and showing the intention of really getting into the AI game, we believe there is a lot more upside in MRVL. The company pivoted to the data center and related end-markets in the third quarter of fiscal 2025. It shifted focus to R&D work to develop its data center chips. This increased focus and spending on R&D should help the company leverage its technology to generate incremental revenue, just like Nvidia and Broadcom are doing.
But how does the company expect to gain market share? After all, if things were this easy, every company would pivot to AI, wouldn't it? Here’s what Marvell is doing to eat into Broadcom’s lunch. The company will first enhance its electrical and optical SerDes. The electrical SerDes helps convert parallel data into serial data so it can be sent over a single wire. The optical SerDes lowers signal loss and provides a higher bandwidth (allowing more of the data to pass through at one time).
Then the company is moving on to improving its packaging technology, which is all about getting more and more transistors onto a silicon wafer. Think of transistors as on/off switches. The more switches you can get on a wafer, the more tasks you can perform, thus reducing your cost per transistor.
Semiconductor companies in the past have shown that this is a difficult task. However, companies with good management, like AMD in the past, have shown that a consistent focus on R&D can make this possible. Intel has also shown that lowering your focus on R&D even when having a technological lead can have disastrous consequences. If Broadcom slips up at some point in the future, it could be Marvell’s ‘Lisa Su’ moment. Looks like the company is getting ready and waiting for that slip-up. Investors should too.