MRVL Earnings "Measuring Stick:" Ways NVDA & GOOGL Add to Growth Story

**SPEAKER_1** (0:00)
So, joining us now for more on all of the expectations, we are joined by Stephen Sopko, Semiconductor and Deep Tech Analyst at Hyperframe Research. Stephen, thanks for joining the show today. Marvell has, of course, become one of the biggest beneficiaries of the AI infrastructure spend. But you argue the stories say less about the chips themselves, more about the moving data around these massive AI clusters. So why then is right now InterConnect becoming so critical in this next phase of AI?

**Stephen Sopko** (0:28)
The most important thing to think about is that these clusters are increasingly based on racks of computers. We talk a lot about how the unit of compute used to be the chip. Now it's the entire rack of 72 chips. Making those racks talk within themselves and talk to each other is the biggest, I think, limiting factor other than things like power and infrastructure that we see in terms of growing AI.
And making those chips talk to each other as efficiently as possible with the least amount of heat and the least amount of energy is really one of those gating factors on how fast AI can grow.

**SPEAKER_3** (1:09)
Stephen, how important is it that we just not, I don't know how long ago it was at this point, had Jensen Wong on stage pointing over at this company as the, I guess he said what, the next trillion-dollar company given the landscape that has developed around Nvidia, not only as sort of the key, you know, GPU, Compute, Data Center name, but now we have this giant funding, you know, arm as a result of Nvidia as well. It just seems like right now Jensen and team, not only are they the kings, they're the king makers at the same time.

**Stephen Sopko** (1:42)
That is certainly one way to look at it. One of the things I've talked about for a while is how Nvidia is looking forward at anything that could inhibit the growth of AI.
Meaning that in order for Nvidia to hit its targets, everything around its compute needs to be scaling at the same rate. The huge announcement yesterday between Nvidia and AWS is a great example of that. So in that announcement, we all focus on the 2 million GPUs, which frankly is a huge number and definitely worth focusing on. But buried in that, there's also a lot of information about AWS and Nvidia partnering on that interconnect, on that communication within the racks. So one of the things when I look at a company like Marvell, I look at a company that enables ship sets from multiple vendors, whether it's Nvidia or AMD or Amazon or whoever else, to all talk to each other so that you don't have to buy your entire data center from one company in order for it to work. You can actually source from multiple suppliers and look to a company like Marvell to help those multiple different types of technologies talk to each other.

**SPEAKER_1** (2:51)
Okay. So as far as what we've seen in just the recent overall investment in obviously the AI infrastructure spend, you've described this optical interconnect as the stake and the custom silicone as the sizzle. So now as we head into earnings, what metrics or commentary would tell you that that optical demand remains really still this primary growth driver today?

**Stephen Sopko** (3:11)
That is going to be one of the big questions that I'm looking at during the call, is how does management talk about that optical internet interconnect, particularly in light of the announcement from Nvidia and AWS yesterday. I don't think that announcement is bad for Marvell in any way whatsoever. I think it does give us an insight to where the engineering effort is going. And since Marvell is uniquely positioned to benefit from both the Nvidia stack and the increasingly open stack that's available as an alternative, Marvell can straddle both sides. And so I think that's going to be a big part of the company's growth.
A lot of us want to look forward at Marvell growing through custom silicon, totally different conversation. But I think this interconnect piece is the story for this quarter. And obviously, three quarters of the company's revenues in the data center, that's going to be the story for the next few quarters.

**SPEAKER_3** (4:05)
You know, kind of the layman might lump a lot of these names into one bucket, call it semiconductors, and include a bunch of names. But as you start to peel back layers and kind of bucket them more specifically, is the closest comp Broadcom when you're trying to find a direct competitor for Marvell?

**Stephen Sopko** (4:23)
That is the challenge. I mean, Broadcom is a much, much bigger company, but you can look directionally and see Marvell looking at Broadcom success, looking at the success of other companies that have done custom ASIC design and say to themselves, okay, Interconnect is a fantastic business. The future growth could really be for us in those areas.