AMD More than NVDA Competitor: AI Factories & META Offer Long-Term Bull Case

SPEAKER_1 (0:00)

That's Stephen Sopko, Practice Lead, Semiconductors, and D-Tech at Hyperframe Research. Thank you so much for your time. Always lovely to check in with you.

So it just crossed this significant level. Now the fourth US chip maker to actually achieve this status after Nvidia, Broadcom, and Micron. What does it mean?

Stephen Sopko (0:19)

That's a great question. I think what it means when you look at these four companies, you're seeing where the need is in the AI market. So Nvidia, obviously.

AMD is really being driven up by two dimensions. One is CPUs. We can talk about that at length. But I think the other is AMD really being seen as that second source for AI factories. It's a lot more than just chips. It's an entire system that companies like HPE and Lenovo are building right now, and they're going to be going out to customers at the end of this year, beginning of next. So I think what you're seeing is this recognition that AMD is really that second source for companies and frontier AI companies that are building out those AI factories.

SPEAKER_1 (1:06)

How much of the move in AMD, obviously that we've seen this week, I mean, firing today because we've got a lot of semiconductor chip stocks on the back foot, has been about Muse given that Meta is such a big customer.

Stephen Sopko (1:19)

I think Muse is a huge driver. I mean, anybody who's used agentic AI where the AI actually does things for you instead of just talking to you, I've been using it at work for almost a year and it becomes something that is as natural as turning on the light. I mean, it becomes very, very much a part of your day-to-day life. This idea of Muse being pushed out and people seeing AI as being something that can actually accomplish things, I think is going to drive a lot of usage. That usage really is going to put pressure on CPUs, because CPUs are the conductor of the orchestra. They're the chip that tells all the other chips what to do. While the GPUs, the AI chips are just running very, very, very fast, they need CPUs to control them and tell them what to do. Cisco has a terrific term when they talk about this. They say humans click, but agents swarm. You're going to see one person with 100 agents out there doing all kinds of things, going back through old emails, reaching out to old friends.

And I think that is going to drive a lot of the need for CPUs going forward.

SPEAKER_1 (2:27)

I want to get to that in a second, but I do want to pick up on the fact that you use Muse. I'm just wondering, I was talking to my previous guest about the fact that it's kind of caused another episode of the AI disruption fears as we saw in the financials yesterday. I'm just wondering in your everyday life as you use this technology, Stephen, what has it replaced?

Stephen Sopko (2:46)

Well, in the case of Amazon Quick, which is the tool that I've been using at work ever since it came out, it doesn't necessarily replace things. It just makes things possible to do that you wouldn't do before. Case in point, it's not making airplane reservations for me. What it's doing is it's planning out my two-week trip. And in the pre-Agentic AI era, I probably wouldn't have bothered looking at 15 companies and three cities and all these different possible things I could do.

And frankly, I wouldn't have an assistant who would do that either. Their time is better spent on other things. So what I like to call is it's the, it makes things possible that weren't worth the time or effort to do before. But suddenly, once they start happening, it raises the bar on what a good trip looks like, what a good deck looks like, what a good email to your aunt looks like.

SPEAKER_1 (3:40)

Yeah, it certainly is an assistant in sort of everyday life. And back to obviously the success of this technology, what do you think it will mean for CPU demand down the road? Because we've already seen the initial reaction as to the potential, what it could mean for this subsector.

Stephen Sopko (3:59)

Very true, I mean, Intel is talking about that they can only fill a certain percentage of their orders, particularly around CPUs. Because it's one of those things we've all seen coming for a while, Intel, AMD and all the companies that license from ARM. And I think the critical thing now is the software is starting to catch up with what the chips are capable of doing. And we saw that in the most recent sort of evaluations of current generation chips, where you're comparing them. In some cases, the software is now fully using the chips. And that's good because companies like Nvidia have moved to a new generation every year. Whereas it used to be a new generation of AI technology every couple of years. And all of that is running when we look at those trillion dollar club members, you look at AMD. A lot of what they announced through the evaluations the last couple of weeks is they're using a type of communications between computers that really relies a lot on Broadcom. And so you look at those two stocks, they're very much, look at those two companies, sorry. You look at those two companies, look at Broadcom because Broadcom enables a lot of the communication between those chips. Broadcom also designs a lot of the custom chips that the big hyperscalers use.