ADI & ON Hidden Strengths in Tech Trade: Dave Nicholson on Analog AI's Importance

SPEAKER_1 (0:01)

All right, let's go deep here on some of the tech stories, including going deep on analog chips in our tech spotlight. Dave Nicholson, Chief Research Officer at Futurum is with me. So let's talk about that. Analog AI, what does that mean? What kind of trade are we talking about these days? How has it changed?

Dave Nicholson (0:19)

Hey Nicole, well, it means two different things, and I think that's really important to understand.

When we think about the chips that Nvidia and others are selling, GPUs, those are digital methods of processing AI workloads. There are analog ways to do that. That's what analog AI is about, but that's more of a future research thing that's underway. So make sure people don't get confused. What we're talking about are the devices that are in and around all of the data center infrastructure in addition to things like CPUs and GPUs, things like optical cabling, connectivity, and very importantly, power supplies and the way to step up and step down power within a data center.

SPEAKER_1 (1:09)

So at this point now, how fast is this part of the story growing in comparison to the digital side?

Dave Nicholson (1:16)

So the truly in-play analog part that we're talking about is marching right along in step with digital. So for example, NVIDIA announces an 800-volt architecture that's coming and these companies are partners in that. So there's no difference there.

In terms of true analog AI, right now, those things are science projects. But I would say within, we're talking more like a two-year timeframe that this can massively disrupt the digital AI side of the world. However, do you think Jensen Wong is going to sit by and watch someone develop analog AI and then not acquire it? Chances are very good that these analog AI contenders will end up getting snapped up by the existing digital AI vendors.

SPEAKER_1 (2:10)

So you were talking about ADI and something like analog devices and how that's already working. How is that? Why is that?

Dave Nicholson (2:20)

Because there isn't a ton of competition in this space and it's something that's absolutely critical. We've gone from 10 kilowatts or 15 kilowatts of power going into a rack, to now a fairly standard AI data center infrastructure is going to have racks, think of things like a big black refrigerator, that are consuming hundreds of kilowatts, getting close to a megawatt of power on a single floor tile.

Sucking in all of that electricity and then distributing it out to hundreds and hundreds of components within the rack, is it's not something to be taken lightly. Companies like ADI for Analog Devices, they specialize in this and they are riding the wave of AI, and their margins seem to indicate it. They're getting software like margins on hardware that they sell.

SPEAKER_1 (3:17)

I think about the pricing power and the inventory, how much the inventory they have, are the customers restocking? Isn't that part of the channel checks too?

Dave Nicholson (3:26)

Yeah, absolutely. ADI is pretty clear on levels of inventory that their customers are burning through, that are based on prior commitments already. During that period of time, before inventory has been depleted, they've gone through two price hikes that don't seem to be affecting customer behavior. So maybe it's not appropriate to call it a backlog, but it's definitely known validated demand that will be landing right in the middle of their price increase goodness as far as they're concerned.

SPEAKER_1 (4:01)

Yeah, and not in the same category, but Nebius has had price hikes, NVIDIA has done price hikes, totally different from what you're saying. But it seems that there is demand and scarcity across the board, as you're talking about the Analog Devices price hikes. Now, what about name like OnSemi? What does it need to really tell the market these days?

Dave Nicholson (4:23)

So I think ADI and Analog Devices and OnSemi are good contrast. Because if OnSemi can get to a point where they are more fully utilizing the resources that they have at hand, they will achieve the same types of margins that ADI is getting today. Their challenge is that they've got fabs that are at 60-70 percent utilization right now. As that ramps up, they're going to be more in the same arena. So very similar business models, very similar ability to ride the coattails and the winds of AI.

But they are more of a longer term play. Right now, ADI is priced for perfection. OnSemi assumes that you believe that they can get to the same point that ADI is.

SPEAKER_1 (5:18)

And I think about the different categories such as autos and also you had industrials in there. I think that was the other segment. How are we doing in those segments?