MU Massive Earnings & Guidance Beat Weighed Down by Supply, Power Constraints

0:00Welcome back to Market on Clothes. I'm Sam Viders on the floor of the New York Stock Exchange. Let's get talking about tech. These earnings from micron, Ben Baharan CEO and Principal Analyst at Creative Strategies. Hey Ben, thanks so much for joining us. Thoughts on micron? Interesting. Reaction we've got. I mean strong earnings. We saw record revenue here, but it looks like the market has a few questions around the constraints that they've flagged. What do you think? Yeah, I mean, this is something relative to the constraints of the entire industry. You know, when you just look at some of the things that Nvidia has announced this week, the addition of multiple new racks of compute all taking up memory, it feels like again, right? The upside to every memory name is really just going to be on their ability to meet the demands in those constraints. So we fully expect that they will be demand or supply constraint for the foreseeable future across, again, not just Nvidia, but all of the hyperscalers. And I think there's some truth to the narrative of there is concern about just the whole,

1:00is there a digestion period coming in a couple of years? Can capex maintain? Is that worth the overall capex that the memory makers need to make to fill capacity to meet those demands in what is again, been a typically cyclical type of an industry? We think a lot changed, but again, right? All of these names are really just constrained on their upside by how much capacity they have. And I think some of that does get priced into their forward earnings. Is that the issue here? Do you suspect that's what sort of stall the momentum here for tech broadly speaking is some of the capacity constraints? I mean, because they have been well telegraphed, but I mean, you know, the momentum just hasn't managed to get going for some of these names this year despite giving so much visibility, seeing so much demands. I mean, you know, Jensen Huang at GTC this week talking about that one trillion through to 2027. And then obviously with an earnings picture like this, which looks very good, but still some worries it seems. Yeah, I think the constraints yes worries. And

2:02I think again, right, a lot of this has been priced into those constraints or at least what the assumptions of the constraints that the supply chain can go and meet demand with, which is sort of a variable. I think that's the work everybody is doing in their due diligence to figure out what's a, what's a reasonable number amidst the constraints and then how much does that get priced into what we think that revenue can be if they can ease some of those, right? So I think because that's a known story and because the last work really of the last couple of years has been understanding those constraints, including on the power side, right? The ability to build and sell racks into those new data centers for greenfield growth. It seems fairly well understood that now the question is how much upside beyond what those constraints are priced in and I think that still remains a question. Efficiencies can be given on some product roadmap as well as if some manufacturing gets pulled in, but I think that's a known story and you're seeing that not across micron. I mean, you're seeing that within video as well and a host of other names, it's that same structural ability to meet that demand with what

3:03I think is relatively understood or priced in regarding those constraints. Yeah, I'm just sticking with micron because these lines continue to come through here. I mean, it's saying non-demand significantly in excess of supply here. We know that micron and some of the other memory plays. I mean, the South Koreans as well are trying to bring on extra capacity, but that's still going to take time. How are you thinking about that delicate balancing act where by the time that they're able to meet some of the demand that that demand starts to come off? Yeah, I mean, right now we don't think that that comes off really even in the timelines that they meet. If you just look at the revenue numbers right in terms of CapEx, even just looking at what Nvidia right guided to and the trillion that you talked about, CapEx has to drastically continue to increase. That's not just a one-time thing and I mean, CapEx by the hyperscalers. So I think when you factor in all of that into a long form of demand, like this constraint, particularly around memory, we think we'll last for some time because memory is not and the players in memory aren't,

4:06you know, they don't just go and build the same number of foundaries that really someone like TSMC can do or that they work with with their manufacturing partners. So it doesn't feel like they're even ramping up, right, to the same demand. And again, because they've been conservative, because memory has been cyclical, they don't want to over invest, right, in fabs to deliver memory. So it feels like this in particular remains constrained for some time. On top of the fact that everybody's scaling their accelerators, building new products for accelerators, even some of the dedicated CPU racks that you see coming out now have a ton of memory, largely so come in them. So this remains, I think, demand constraint or supply constraint for really longer than most people are estimating. I thought I was really interesting micron coming out this week talking about, you know, obviously bringing on new capacity and manufacturing in Taiwan, which is a market where US policy has been about trying to de-risk from and we've even seen TSMC setting up

5:07shop in Arizona. Do you have any thoughts on that? Yeah, I mean, I think it's one we don't really make a lot of memory manufacturing here in the United States anymore. If you just look at where the skill sets are really where that supply chain is, that's typically been in Asia and I think that's part of it. And so I think it would be a little bit more tricky right, logic being made here, both between not just Samsung, but also TSMC and then obviously Intel. A different set of base makes it a little bit easier to bring that on even though I do think TSMC is still having trouble to get engineers and workers into those facilities to bring up the fab. So it's a whole part of the supply chain talent. I think there's a lot of things that just make it easier. Again, on a timeline that needs to be met, which is within the next couple of years of which they're scaling the ecosystem around Asia is going to make it easier for them to hit those milestones than if they did it on US soil. Okay, makes sense from a geographical standpoint. Obviously, Nvidia GTC, the big story of the week with those figures that I threw out there. It's been interesting listening to Jensen, what he's been saying about actually receiving H200

6:11orders speaking of Asia from China as well. Also talking about how open core he thinks is the next chat GPT. We saw a nice rally in some of those AI tigers over in Hong Kong overnight off the back of that. I'm just wondering what else has really stood out to you this week about some of these announcements. What we should be paying attention to? Yeah, I think the main one really is just the way that Nvidia is disaggregating their infrastructure away from what was just, you know, the last few years been primarily around GPU racks or accelerator racks. We're not talking about CPU racks, accelerator racks, networking switch racks, all things that need to be networked together over copper or over over fiber to function as a compute pod. So the fact that it's not one product, it's now disaggregating or bundling into three, four, five products or more. Just again, drastically increases the demand or the demand need for all compute that we're bringing across the board and all of the underlying, again, memory, storage, networking, copper stuff, optical, you know, back end switches. So it's just as competitors follow

7:15swoop as we think they will with the way that Nvidia is building out their product family. It's just going to again demand more and more constraints on the supply and the supply chain industry because their whole vantage point is what does these these pods together or this bundle bring in terms of TCO, which is monetizing those tokens, which is what, you know, we need the hyperscalers to do is to make money on their back end infrastructure. And this is the way that Nvidia's building that TCO, which again, just adds to a good portion of their tam really across every vector with the exception of just a couple tam buckets. So we think there's a lot of upside, but again, we're going to remain pretty heavily constrained just via the supply chain. Yeah, absolutely. You know, that perhaps is why we've seen a degree of hesitation in this market, even though it's been a pretty favorable set of developments and headlines out of GTC, Nvidia ending, eight tens of percent lower today. Ben really appreciate it. Thanks so much for your thoughts for us and reacting to these micron earnings as well. Ben Bahar and their creative

8:16strategies that come