We have not written about Marvell in a while, and so we tuned into their earnings call yesterday. Marvell reported good numbers and guided above Street expectations. Most critically, the company was able to talk about good news for its catalog, merchant business. This quarter did not answer the question as to Marvell’s future, but it was some much needed good news.
For the past year or so, many of Marvell’s core products – especially in storage – saw sharp declines due to a variety of factors (i.e. post-pandemic inventory cycles, competition, product cycle timing and changes in customer buying patterns). As a result, the Street had started to look at the company solely for its custom ASIC business. As we discussed back in December, custom silicon is not a great market for a company like Marvell. They have a lot to offer here, but customers can be fickle and that can spell unending headaches for a company. We recently heard an investor discuss the company for 20 minutes without once mentioning any of the catalog parts.
Fortunately, the cycle has turned. The company’s older product lines seem to have turned the corner. Also, their optical products, acquired via Inphi, are doing very well with an upgrade cycle underway. Marvell paid a lot for Inphi, but that deal seems to have turned out well as the unit has been executing its roadmap strongly.
For what it’s worth, the custom business seems to be doing well, with a number of design wins. On the call, management mentioned that many of these design wins are “multi-generational”, meaning they have design roadmaps with customers over multiple years, greatly reducing what we see as the chief risk of the custom business.
That all being said, Marvell remains in a tight position. They are executing well, as has been their pattern for years. But they are stuck between a fearsome competitor in Broadcom and the growing desire from their major hyperscale customers to design ever more of their own silicon. For instance, we noticed that Marvell did not present at Hot Chips last week, and we were not the only ones who noticed. At that event, Broadcom had some very powerful products to show off that could conceivably cut to the heart of Marvell’s networking products. Of course, that Broadcom part is still a few years out from being commercial, let alone competitive, nonetheless the risk is real and Broadcom is executing just as well as Marvell.
When it comes right down to it, we still think Marvell risks finding itself sub-scale. As such, we cannot rule out the possibility that in a few years they end us a target for one of their larger peers. If we had to pick we see a really good fit between Marvell and AMD. With the acquisition of ZT Systems, AMD has demonstrated that they want to provide a more complete solution as it competes with Nvidia. Today, AMD lacks a networking business which Marvell has. Marvell alone would not make AMD fully competitive with Nvidia’s NVLink and EthernetX, but it would be a step in the right direction. We have no idea if such a deal will ever happen, but the logic seems fairly clear.
Leave a Reply