The Information's TITV
00:01:27 · Akash Pasricha, Patrick Moorhead
[00:01:27] S0: Very spooky indeed. So, look, I mean, executives laid out all the reasons why the outlook was not what investors may not have been hoping for. I guess the main question I have for you is, is this all out of Intel's hands? I mean, is there no responsibility here on execution? What's the story? [00:01:50] S1: Yeah. So I'd like to hold definitive judgment until we see what AMD does, because they are the closest direct competitor of Intel. But it is a fact that there are substrate shortages out there, And it is a fact that the Windows PC market is not doing well. I mean, Microsoft, they were down, what, 20... They... You know, down 22%, and they guided down mid thirties on inventory. So it is an industry issue. Now, Intel does have some design uniqueness in that its parts are done on different manufacturing nodes and are kind of spread out in various places. And I do think that TSMC is going to do AMD a lot of favors when it comes to substrates and wafers and packaging, just to see if they can go take some market share away from Intel. You know, the tragic accident out of Fab twenty eight in Israel, you know, that has got to be looked at as as a one time horrible event. [00:03:09] S0: Is there... So I I guess you mentioned the balance here between execution and external environments. If you look at the company at large, so executives said last night that they probably won't generate free cash flow in 2027. Are you okay with that?
00:03:09 · Akash Pasricha, Patrick Moorhead
[00:03:09] S0: Is there... So I I guess you mentioned the balance here between execution and external environments. If you look at the company at large, so executives said last night that they probably won't generate free cash flow in 2027. Are you okay with that? [00:03:28] S1: So I'm okay with it as long as there are good reasons and CapEx going into the Foundry business, which is an absolute requirement. I am okay with that. And then if I look at, you know, the turnaround in certain businesses, I mean, Q4 data center was up 15%. If I look at the profit improvement on the data center side, it dipped 41% versus 23 a year ago. They are generating good cash flow, and the biggest drain right now are the investments that are going into Foundry that's losing 28¢ per revenue dollar. But it's very similar to the negative FCF that I think we're seeing with the hyperscalers right now, where they are building upfront for their capacity. And the reason why AMD doesn't have that is because they don't have their own foundry, and they are subletting that to TSMC. [00:04:40] S0: Right. Although I have to tell you, we've had this conversation on the show previously when TeraFab was in the headlines and we were I talking actually made the point that similar to the way that SpaceX is building capacity, you could say that the chip companies are too, but the difference, of course, is the political pushback in building data centers versus chip factories. Foundry business seems to be a lot better looked upon, and so at least doesn't have that headwind against it.
00:04:40 · Akash Pasricha, Patrick Moorhead
[00:04:40] S0: Right. Although I have to tell you, we've had this conversation on the show previously when TeraFab was in the headlines and we were I talking actually made the point that similar to the way that SpaceX is building capacity, you could say that the chip companies are too, but the difference, of course, is the political pushback in building data centers versus chip factories. Foundry business seems to be a lot better looked upon, and so at least doesn't have that headwind against it. [00:05:16] S1: Yeah. No. That... That's exactly right. I think pretty much everybody is a resounding, yes, we need to do that. And also, I I I think that that shipmakers have literally fifty years of experience in some cases partnering with their communities upfront, setting expectations. Some fabs get tax abatements, some don't. But in general, there there is less pushback, which is kind of ironic given the amount of chemicals that a fab that a fab uses, and it doesn't, suck an inordinate amount, of of power, but it but it does pull it. But I'm unaware of anybody's electricity costs going up because a foundry was put in versus, let's say, a data center. [00:06:09] S0: So when do you see the foundry business turning a profit finally?
00:06:09 · Akash Pasricha, Patrick Moorhead
[00:06:09] S0: So when do you see the foundry business turning a profit finally? [00:06:14] S1: Yeah, I think what we're looking at is '29 or '30, and those expectations were set out there pretty clearly. Part of this is just the time it takes to not only build the capability, and that's not only building the foundry, but also the transistor technology. And then you have to get far enough along so you attract customers, right? Lip Bu Tan, Intel's CEO, did talk about they would make some announcements in the next two quarters on foundry customers, which I think is very important. [00:06:55] S0: But waiting to hear who
00:06:57 · Patrick Moorhead
[00:06:57] S1: knows I how to send know. I mean, it's likely Google and AWS on the packaging side. It's on the wafer side that is the biggest unknown, but I do believe that it's gonna be NVIDIA and Apple, out of the gates, because they need a second source, and, they also see the need to manufacture more The United States. And a lot of that CapEx that's going in right now that's making them cash flow negative is investments that is going into 14A node and 18A and AP nodes to build out the capability. The semiconductor business is a cruel business in that if you find yourself with the wrong architecture, you're literally gonna have a hard time fighting back. It'd take you four years, okay? You lose the recipe in a foundry, it could take you even longer. And then when you try to connect the two together, which Intel is doing between design and foundry, it is and continues to be a challenge. So I look at milestones. Was Intel able to crank out 18A? Check. Yields are getting better every quarter. Is the feedback on their early designs called PDK kits for, 14A. It looks like everything is out on time, and what I'm hearing in the ecosystem is those are higher quality. And in 2027, you have this convergence of Intel designs where their PCs and their newest servers will be on very similar manufacturing nodes. What that allows them to do is move wafers back and forth between client and server that they can't do today. They currently have no servers on 18A, yet their premier client is on 18A, and the servers are either on Intel seven, 10 nanometer, which yeah, a few desktops are still on, but it's very hard to shift manufacturing when one part of the market becomes hot. But out of the other side of my mouth, I will say, those parts that did converge on the same process, Intel was able to move a lot of capacity on the server side. So I do want to give them some credit for that.
00:09:31 · Akash Pasricha, Patrick Moorhead
[00:09:31] S0: Right. Hey, Patrick, so I know you said you're obviously looking ahead to AMD earnings. But, you know, based on the commentary that we got and the results we got from Intel, what are you expecting from the other chip companies? Broadly speaking, how does this affect what you think they might report? [00:09:52] S1: I think data center is gonna be full steam ahead. I think TSMC will will cross every bridge and break every rule to be able to to to give AMD what it needs on on the server side. And then on the PC side, we will likely see a lower number, from from Intel, right? Because they're all selling to the same people. They're all they're all pulling back, on that. And and therefore, think, I think net net, they will have a better guide. The one thing I wanna say about the PC market though, is that demand is not gone. Let's say you had a five year old PC and you looked at memory prices, and you're about to pull the trigger on a new one, and you're like, I'm gonna wait. Like, you can't... PCs don't live seven, eight, nine, ten years. I mean, some do, but, but not a lot. [00:10:50] S0: Right. But just very quickly, you talked about AMD, but what about all the other chip companies that inevitably do rely on the same memory in many cases?