← Business Breakdowns![Intel: Cyclical Recovery or Secular Demise? - [Business Breakdowns, EP.149] — Business Breakdowns cover](https://megaphone.imgix.net/podcasts/a181d4ba-caf1-11ee-ac63-038c251c9fc7/image/23f104.png?ixlib=rails-4.3.1&max-w=3000&max-h=3000&fit=crop&auto=format,compress)
Business Breakdowns
Intel: Cyclical Recovery or Secular Demise? - [Business Breakdowns, EP.149]
01:06:37
Loading summary
Transcript63 lines
- [00:01]
Matt Russell
Highlighting today's sponsor if you're looking for a simple yet sophisticated investing experience, check out public.com whether it's stocks, options, fixed income or even crypto, it's all there. You can even earn an industry leading 5.1% APY on your cash and Public is trusted by millions of investors. NerdWallet just gave public 4.6 stars and calls it an impressive experience. Can sign up in two minutes at public.com, all of your investing in one place.
- [00:37]
Colossus
This is Business Breakdowns Business Breakdowns is a series of conversations with investors and operators diving deep into a single business. For each business, we explore its history, its business model, its competitive advantages, and what makes it tick. We believe every business has lessons and secrets, secrets that investors and operators can learn from, and we are here to bring them to you. To find more episodes of breakdowns, check out joincolas.com all opinions expressed by hosts and podcast guests are solely their own opinions. Hosts, podcast guests, their employers or affiliates may maintain positions in the securities discussed in this podcast. This podcast is for informational purposes only and should not be relied upon as a basis for investment decisions.
- [01:26]
Matt Russell
This is Matt Russell and today we are breaking down intel to cover Intel, I was joined by Todd Alston, CIO of Parnassus Investments. Todd started covering semiconductors in the mid-90s and has since lived through eight cycles in the sector. So we made sure to cover the characteristics of how to separate secular changes from cyclical changes, which is all the more important considering Intel's history. We could have easily spent an hour covering that background of Intel. The backstory. But let's set the table with an overly simplistic storyline right here. In the late 80s, a newly appointed CEO, Andy Grove, made the pivot to exit memory chips and focus on logic chips. The result was intel stickers on what felt like every laptop and desktop I ever touched. In the 90s and 2000s they were the leading edge chip designer. But things shifted. Intel missed out on the mobile market, and then intel missed out on EUV technology, and now they're left playing catch up. They've even started to bleed market share in those personal computing devices, which obviously have lost share as we spend more time on our phones and more time thinking about all the AI chips and what that's going to power in the future. So to summarize, intel has been falling from its iconic status. In our conversation, Todd helps outline what went wrong, what is changing today, and what we can watch or monitor as this progresses into the future. When you look at Capex budgets from Google, Microsoft, Meta. It's almost impossible to overstate the importance of that semiconductor ecosystem to everything around us. So please enjoy this breakdown of Intel. All right, Todd, thank you for joining us here on Business Breakdowns. I'm excited to get to intel, one of the most important strategic assets in the US sitting in a fascinating industry and you're quite the perfect guest and I thought we could start there. I just wanted to touch on your experience in this space which spans decades. What has it been like to be looking at semiconductors from the early 90s? What are some takeaways when you think about this space and maybe just a little bit about your background and how much time you spent looking at some of these names would be helpful to kick us off.
- [03:47]
Todd Alston
Yeah, Matt, it's so awesome to be here today. It's almost like being a kid in a candy store talking about intel and semis. So for me it's a neat personal journey that I started at Parnassus when I was 22 years old. I'm 51, I'm still at Parnassus. And I was so fortunate, Matt, that 1995, I'm 22 years old, our founder, we were a high active share, high conviction contrarian stock picker and I was tasked with going down to Silicon Valley starting around August of 1995 to look at semi stocks because they were cyclical. But there was a great long term secular trend in semis and I think that's just been an incredible journey that started in the summer of 1995 with a memory cycle. We had built too many memory fabs and back in those days there was a lot more players that were building large fabs. Unlike today where there's only a few major manufacturers. Back then there was tons of memory manufacturers logic, all sorts of different players. So it was an incredible time. And so for me to live through now eight cycles as a practitioner, it was a great journey. And I would just say this is an industry that's cyclical, that has secular dynamic innovation and when you overlay those two things, it's incredible opportunity that in short, when you have cyclicality around a secular trend, it's an incredible place to invest when you can get that cyclical recovery and then a secular RE rate because bottoms people always over extrapolate negatives. We'll get into why semis are cyclical, but it's that secular RE rate that happens with cyclical recovery that is the essence of return. Getting those things right can be quite an endeavor and I'm blessed to have Been through eight cycles.
- [05:35]
Matt Russell
Yes. Maybe we can talk a little bit about that from an industry perspective. I came from a world where something like the trucking sector had a pretty traditional cyclical, without any secular re rate market and it was going to trade somewhere around 15 times on a normalized basis. But that was it. Every cycle was more or less the same. As you mentioned here in semis it looks very different. So can you bring us maybe up to speed on today's cycle? What that's like, how that's maybe different from whether you want to pick one previous cycle or multiple previous cycles. Anything that differentiates what it looks like today versus the past. When it comes to the cyclical and secular trends going on in semiconductors, that's.
- [06:16]
Todd Alston
A great place to start because. So you go back in time, in brief, I lived through PC cycles, the dot com bubble which was an epic demand overshoot. And then going through the last decade and a half, we had the mobile 3G, 4G, 5G, you had mobile cycles and then obviously cloud compute, data center build outs and now the AI boom. What's unique about this cycle is that AI seems to be such a mega cycle. In my career we've had these major cycles like I mentioned, but this one has such an amplitude to it because the amount of invest and the amount of exponential market demand growth is so large. And then the balance sheets of the investors, be it Google, Microsoft, Amazon, Meta, there's incredibly large balance sheets that are self funding. Unlike say the dot com bubble where you were raising a lot of debt, you had unproven business models and so that caused a super spike into a shorter cycle. I think this cycle, Matt, you've got almost unlimited balance sheets going after an unlimited exponential market with companies that can put cash flow, then scarcity around fab capacity. It really sets up for quite a cycle ahead that appears to be more durable with more upside, more lasting than some of the cycles we've seen in the past.
- [07:34]
Matt Russell
And let's introduce intel here. It's a name that's been synonymous with semiconductors for a long time. I think the reputation, at least from an outsider's perspective, has waned or had its ups and downs. At the very least we'll get into a little bit of the history in terms of where they got to where they are today. We're to going where do they sit today? Just in terms of positioning, however you would frame them. And maybe if your perspective is going to be different from investors, it would also be interesting to hear what the consensus view of intel is Today, yeah.
- [08:05]
Todd Alston
So I think starting with the consensus view, I think it's incredibly negative. I mean we can get into the history that they missed mobile. They fell behind TSMC, especially in that 10 nanometer node, to 7 nanometers. They were late to adopt EUV. AMD's gaining share in servers and PCs and what they're left with is declining monopolies, lower return on capital and a real battle going forward. The consensus is almost like jumping over a two foot bar. I don't think it's very high. So what's the intrigue? And I think where this call can get really interesting is when Pat Gelsinger joined the company. Pat was basically 19 years old or a teenager when he joined intel with an internship. He was an Andy Grove protege. He was Intel's first CTO and, and really has the genetics of the old intel, which is about process, technology, leadership, pushing the laws of physics and delivering incredible value. So right now there's a chance that this could be, in Pat's words, one of the great American turnarounds in corporate American history. And where that would really go, Matt, is a couple things. First of all, intel has tripled down on getting back on the leadership cadence of manufacturing. They fell behind tsmc. So point number one is they're going to go through five nodes in four years, which is really aggressive architecture adoption. Where they just finished Intel 7, they're ramping up Intel 4, which is their first EUV launch, Intel 3, which is hitting the first half of 2024, that's really the second wave of their EUV. And you're going to hear terms Intel 20A, Intel 18A, which is really four waves in one that Pat's trying to pull off from a design architecture, which in short is not only the EUV launch, but it's also this term called gate all around technology, which is different ways of having three dimensional channels and more transistor gates. It's new ways to power devices from the bottom of the device up, which is power via, which is called backside power. And then new ways of packaging these chips, chip on wafer on silicon, which is a really new way. So would you think four waves, five nodes in four years? That's the first part of what Pat's trying to do. The second thing, what Pat's doing is splitting the company in two. One side's going to be the design designs for CPUs that go into PCs and servers. That's the design side. Then he's splitting the foundry side, which is really the manufacturing. We just talked about. So that's going to be key, not only to build some of Intel's parts but also to build maybe a foundry of the world. And that's where the upside could be really enormous as they look to be an alternative to say Taiwan and Korea to manufacture. So that's something where 2024, 2025 with all the geopolitics, if they can regain device leadership and then become more customer centric. You don't do all this without a whole EDA network and all the designs get ported in with customers. You can have the technology to build it, but then you have to have an ecosystem around it. Pat's doing all these at once. So it's a little bit of build the airplane and fly it at the same time. But I think he actually has the ingredients that potentially really pull this off. And as we started this, it's a two foot bar. So if he makes it, there's a chance that could intel be one of the magnificent seven in three to five years. I'm not here to say that's a slam dunk, but that would be the roadmap of possibility.
- [11:24]
Matt Russell
Let's just zero in on the two foot bar a little bit to understand why what went wrong before it had to be corrected. I think you mentioned Andy Grove, well known CEO, great books from the 90s. What happened to the DNA of Intel or what were some of the missteps that resulted in that missing of mobile, missing the EUV technology in the early stages. What were the things that went wrong and resulted in this need for correction?
- [11:51]
Todd Alston
And he was a pure legend CEO and the glory days that I think market cap of intel was around 4 billion when he joined and it peaked at over 200 billion. So I think that speaks for itself. Those numbers are directionally many, many multiples. So he really set the stage through their x86 Pentium and just really building a monopoly on the PC market that ended up going into the server market with their CPUs and of course process technology leadership. So they had all the Ace cards in their hand and when you looked at the 2005 timeframe, they really had tremendous leadership and ultimately locked down even Apple to build the chips for the Apple laptops and et cetera. So they were riding on top of the world and really when the company started to hit that real rough patch is around that 10 nanometer timeframe in, let's just call it roughly 10 years ago where you started to see them, they missed mobile on some real opportunities where they had chances to build some more parts for Apple that they could have done at some lower profit margins. But I think they were too PC centric. They weren't willing to dilute the cash cow, the gravy train. So they missed mobile. They missed an opportunity to get volume parts with Apple on iPhones at the same time. Then it came down to process technology. And unlike TSMC, it was early to adopt EUV and that 10 nanometer going to 7. They went into things like double patterning and immersion, different lithography that they tried to get more out of older technologies and didn't really adopt euv. So slowly. What started to happen was two things. Number one is they fell behind on process technology at 10 nanometer going to 7. But that had a compounding effect because AMD was then porting designs to TSMC with a more advanced process technology. At the same time, Intel's a double bet on we're going to design it here and build it here. So if you make a small error on designing it in terms of leveraging your architecture at the same time, you can't build it as effectively. You have a cost disadvantage. And so that's really the essence of what happened at 10 nanometer going to 7. So they essentially had close to, you don't want to call it 100% but 90% share in servers and PCs. Now that's dropped to 80% share in PCs and say 70 in servers. And that is going to erode here a bit more into 2024 until some new products come out. Then at the same time they missed obviously with Nvidia on the AI part, which ironically was GPUs were something pat was working on years ago. We can get into that more in detail when he left intel in 09. But they really missed out not only on mobile, then missed out on some of the AI wave. So too PC centric, they missed EUV. They fumbled 10 nanometer and then they were just poor capital allocators. They spent 15 billion buying Altera, 15 billion on Mobileye. They spent 8 or 9 billion on Mac a few years ago. So they were just churning in capital in some of the wrong places. So all that led up to where we are today, which is declining moat gross margins down 17% from the peak R&D spending up 7 and a half percent of sales, operating margins going from 30% to 7. So you can see the wreckage now is running through the P and L.
- [14:59]
Matt Russell
When you start to institute rebuilding or rebranding or taking back what you once had, the effort that goes into that, I think you laid out a lot of what they're trying to do in terms of strategy and there's a lot going on there. I'm just curious, as an outsider or someone who's even inside the industry, what do you think about as the milestones to be measuring? And maybe we could just use something as simple as chip adoption, whether it's in mobile, PC or with something related to AI? Are these big announcements that get made where contracts are won? Are there other interim steps that happen just in terms of you tracking something? In terms of on the way down knowing that things were going poorly and then on the way back up knowing that things are moving in the right direction? What are the signposts that you could be looking at just to understand what's happening there?
- [15:49]
Todd Alston
Here's what I'm looking for and there's a lot. But let me just boil it down to a couple. Number one, here is five nodes, four years. How are we doing on that with the announcement? So we talk about intel four, intel three. You're going to hear products like Granite Rapids and Sierra forest Getting out 2nd gen EUV products and getting them out, launching and trying to stem that share loss to amd. That's point number one, really important. Point number two is going to be their Intel Foundry solutions and the business model there. They've talked about down payments of potential large customers or customers that are, hey, we want intel to build a product on their 18A architecture. So look for either more customer deposits or customer announcements for Intel Foundry solutions. Point number three is getting back in the game with Nvidia and we can get into the H100, H200 and AMD's products. But really you're going to hear about Gaudi and their GPU and really it's going to be Gaudi 3 GPU launching later this year. And you're starting to see some spec outs like Databricks, who's an interesting software company's put out some interesting specs on how Gaudi 2 is showing some progress on AI and let's see how that works out. It's one thing to talk about specs, it's nothing to deploy it. We got to see big orders. So let's see what happens with Gaudy 3. Beyond that. We're going to have a really interesting question as we ramp training models. You definitely train a model. You have to use a GPU like 100 million, 200 million parameter model. You got to use a GPU. But what you're going to see at some point is there's going to be a lot of inferencing done, which is once you train a model, you got to use it and a lot of people use it. Some of that inferencing math is going to be done with CPUs and maybe a 5th gen Xeon processor can do that. And in the short term, budgets have gotten crowded out. People are spending, I mean meta talking about spending a lot on GPUs for Nvidia. What happens if all of a sudden budgets loosen up a bit and there's not only a traditional server PC rebound this year, which is something we think can happen, but just inferencing on the edge using a Xeon 5th generation CPU, that could be interesting. So keep an eye on that. And then finally just keep an eye on what Pat's doing. On the EDA side, he added a board member named Liputan. He was the former CEO of Cadence Design Systems. EDA is electronic design automation. It is like the Microsoft of designing chips. And synopsys and Cadence are the two big EDA players. But building an ecosystem is so important. To put it in a foundry, you need to have it batched through an ecosystem with all the software compatibility. And Intel's got to do a good job on that. So look for EDA announcements where they're forging better relationships on that. So that's what we need to see to really have the stock and the business work.
- [18:27]
Matt Russell
We did a fun business breakdown on Cadence Designs and it's so interesting to see how the ecosystem has its niche players that are so important to the value chain. If I take what you referenced there and just think about the profit pools that exist inside the business today. I imagine that CPUs is still a big piece of this. But what do those profit pools look like and do you have a rough sense of where they can be, let's say five, 10 years down the line just to understand the importance of each of those categories. Obviously the outcomes are unknown, but in your own mind, what the potential is there and what's going to be the biggest driver within the business?
- [19:08]
Todd Alston
Obviously disclaimer forward looking statements. That means humility. That's the first thing in bold font. But here's why we're talking about this today. And the stock's 47, $48. High 40s. When you think about their business, they were doing 78 billion in revenue at the peak operating margin above 30%. And the business model cratered as we talked about sales hitting roughly $54 billion this year. Operating margin 7.5%. Where does this go from here? So the first benchmark we have is they're taking out billions of dollars of cost, stabilizing market share. And our first parameter is getting back to a 25% operating margin. So our goal here by 2026 is for intel to get back to 25% operating margin, roughly $4 in earnings power. You put an 1819 multiple on that stocks mid-70s or so. And so the drivers there would be stabilizing server and PC market share, taking billions of dollars of cost out of the model and stabilizing the business. That's point number two. Long term I'm going to have two parts of what the dream scenario is here. The first thing is within three years, could they get their GPU business? We Talked about Gaudi, Gaudi 2, Gaudi 3, these are GPUs to say 2 to 3 billion dollars a year in revenue. Right now they talk about having a visibility into a billion dollars plus of revenue, getting those invoicing people, getting those GPUs and closets, invoicing them and ramping that to say 3 billion. At the same time, Intel Foundry Solutions, could that be 3 billion in three years? We've talked about down payments and deposits. So 18A ramps in 2025 getting that to 3 billion. Now you've layered on 6 billion. And then you start to think about what could these businesses be by the end of the decade? Could intel foundry solutions be 10 billion plus? I mean that's kind of where you're really into it. And when they look at that, what they're thinking about is 15 to 20% drop down free cash flow yield on that revenue. And then ultimately the long term goal with David Zinsser is 60% gross margin, 40% operating margin. So you start looking at getting revenues past 75 billion. Could they knock on wood, be on their way to 100 billion by the end of the decade? These are just hypotheticals. But your hypotheticals would be a double digit billion foundry business, a GPU business. I mean, you see the numbers Nvidia is doing already. I mean gaining 3 billion of GPU revenue is like a drop in the bucket. So you can start to make numbers bigger. If the AI accelerator market. Lisa sue talked about 400 billion in 2027. I'm not here to say that's the number. Could it be half? 200 billion? Yeah. Could it be more than that? Maybe. But you start to look at the size of these markets. If intel has any success at all in GPUs, could that be 10 billion plus? Foundry 10 billion plus. And then long term, as you build this up, could PCs grow low mid single digits and the basic server, we can get into the puts and takes of ARM and CPUs and servers, but could that grow a bit? You can start to build up pretty good profit picture there on that. That's well worth $200 billion market cap when Nvidia is at 1.5 trillion. I'm giving you a road map here. We can get into more detail where you want to take this conversation, but those can be pretty good upside numbers from Intel's $200 billion investment. Feels like a ground floor valuation. If this all works.
- [22:28]
Matt Russell
It sets the table really nicely in terms of the discussion. When you reference that historically the peak 78 billion of top line at a 30% op margin, I'm assuming that was mostly CPU business. When you start to think about that 40% op margin over time and the growth in GPU and Foundry, do those businesses have a drastically different margin profile where mix shift is going to be a big driver of that? It's a quite impressive number to hear. So I'm just curious, when you think about these three businesses or these three segments, are there drastically different margin profiles between the three?
- [23:03]
Todd Alston
Yeah, there can be. So the interesting thing is Foundry can be an incredibly profitable business. You can look at TSMC and I'm not here today to compare Intel Foundry Solutions at tsmc. So the issue is going to be a drag on Intel Foundry Solutions. And intel here in the upcoming weeks is going to give us some more detail. They're going to break the numbers out and you talk about the two foot bar. It's common knowledge that the numbers are probably going to look pretty ugly on Intel Foundry Solutions. Right now we're running depreciation over 10 billion a year. We know revenues are depressed, so we know we're off a low base. The foundry business will be diluted for quite some time. You got a huge fixed cost. When you think about putting $125 billion of capital in the ground the next five years, this is not a cheap endeavor. And that's either the opportunity or the risk or the exponential opportunity. So let's just break down the answers. I think yes, Intel Foundry Solutions, if it can cross 10 billion by end of the decade, it could have very attractive economics. Let's just call it 20% returns on capital potentially. I mean the company's talked about that at least in some analyst day slides. We'll see what they update. The GPU business could be incredibly profitable. I like to think of GPUs as Moore's Law squared, which Jensen talks about. Because in the same way Nvidia has their GPU with the CUDA operating system, there's a software component to these that gets really interesting. That's what Gaudi 3 is, which is an asset they bought from habana labs in 2019. So if they can build that Gaudi software stack around the GPU and then build that, there could be some really nice economics. But right now our revenues are incredibly small. So let's just have another conversation when revenues are at 3 billion and then finally I'm actually pretty bullish on their base business. Even though the monopoly days are behind us, I think with the five nodes in four years and if they can get Intel 3 to 20a to 18a, I'm actually pretty bullish. If these terms we talked about backside power, power vias, gate all around, those things can really get Intel's process leadership back. And if they can do that along with hey the chips acts that add some subsidy in there, that helps reduce the cost. And if they can stabilize share versus amd, I'm bullish of that business getting back to some pretty good margin. So net, I think you've got some drivers that could make this a pretty profitable business. But the issue is we got two major areas we're investing in and those are going to be diluted for a.
- [25:22]
Matt Russell
While with the GPU business. You mentioned Moore's Law Squared. I love that. When you think about the potential cyclicality in that market, does it look good different from what you have in the CPU business? And I'll give you again the cheap seats observation would be okay. If I have my MacBook and that's got the intel chip, that's market share. You go through the cycle, maybe there's a chip switch. But with GPUs, you could tell me that the customer's actually want to diversify their GPU set between Nvidia and Intel. But give me a little bit of an understanding of what that market would look like and if there's any differences, whether it's competition, share cyclicality, anything that would differ from the historical CPU business.
- [26:05]
Todd Alston
Yeah, so I think what's different in the historical CPU business will be this, the concentration mat. I think that's what's going to be so fascinating and we've seen very recently what meta spending on GPUs and just there's an incredible concentration winner takes all in that versus CPUs I think will be a bit more fragmented. So when I think about the cycle, we're obviously in a huge AI wave. Was it 45 billion last year in AI chips go into. If it's 400 billion in and by 2027, we'll see. But I think the interesting thing will be there's been such a pull in on training these models for AI, and I think 2024 is going to be a huge year. I think the first question we're going to have is in the middle of, say, 2025, we've had an enormous training rollout is how do they monetize these businesses? Because when you start building up, and I think New Street Research has done a good job on this, I was going to quote some of their numbers, but if you look at a $400 billion accelerator market by 2027, that's 1.1 trillion of data center spending. So clearly the big guys, Amazon, Google, Meta, et cetera, they're going to have to really have some serious business models to plow back to keep this going. So my sense is we've got another 18 months of a real big cycle where you're essentially jamming GPUs in a closet, plugging them in, training models and invoicing people. And I think that's going to be big. So the cycle within a cycle is you're now going to have a mobile cycle, a PC cycle, a traditional server cycle, which are different than this AI wave. And so we can unpack that a bit. But this AI wave is sucking the oxygen out of some of these budgets. So you could have a case where AI for the short term tops out mid-2025. Well, these other markets grow ironically, and there's digestion because we've now trained a bunch of models. Now we got to inference them on the edge with some other chips that aren't GPUs. The semicycle where I've said I've lived through eight, they might be different because we're going to have memory cycles, GPU cycles, logic cycles, and then PC cycles, which are overlapping in different ways than they have in the past.
- [28:03]
Matt Russell
Yeah, it's very interesting to think about all the variety there. When you mentioned that you expected it to be much more consolidated or concentrated in the GPU market versus fragmented in the CPU market. Was that specifically the buyer base, meaning meta, Microsoft being concentrated buyers, or were you also referencing the supplier category? I just want to make sure I understand that properly.
- [28:27]
Todd Alston
Yeah. So it's more the buyers than suppliers. So still, even in PCs, you really got AMD and Intel. And then obviously Apple has their own internal silicon for PCs, it's going to be concentrated that way for the actual manufacturing base, but it's more the demand base. When you think about the sheer demand that's coming from Google, Microsoft and some of these large AI players. It is the market right now, largely, which is different than a PC market. These massive players that are sucking down GPUs. It's just a different dynamic.
- [28:56]
Matt Russell
Yeah, it would be a fragmented market if those big players weren't spending that massive amount of money on perhaps.
- [29:01]
Todd Alston
Exactly, Matt. Yeah, so that's what we're working on. So we don't have all the answers for that, but that's something that's part of the equation.
- [29:07]
Matt Russell
Are they incentivized to diversify Between Nvidia and Intel, is there any case to be made that you want both servers or both chips from both companies powering your servers?
- [29:19]
Todd Alston
I think there will be. It's just the question is you have to have an alternative. And I think, you know, obviously Nvidia has got an incredible man. I think the next, say, two and a half years, it's going to be. Nvidia is in an incredible position with Cuda and the standards are setting AMD somewhat. You've seen some numbers that are coming out. Originally it was 2 billion. Then the numbers are going up a bit exponentially this year for what AMD is launching. So they're also in a good spot. And that's creating somewhat of a competitor, I think intel, with their Gaudi GPU 3. I think people want a solution. I think there's a certain, even a patriotic. I'm rooting for Intel's GPU and the Foundry, but they've got to get that competitive. And I think right now it just hasn't been there yet with the software stack. And I know people are experimenting with Gaudi too, and you read some specs that look pretty interesting, but then I'm like, where are the orders? Ultimately, it's good to read these blog posts, but that has to turn into a large invoice, and we haven't seen that yet. Look at some of these. Pricing at, you say 30,000 for a GPU. If Goudi is coming in at some lower prices, whether it's 5,000, 15,000, I'm not the one to say that, but I think they might try to come in a little lower in price to make up for some of their performance deficiencies. And can they be, shall we say, good enough? Because I think some of the interesting argument's going to be you're training these large models and what models need the best and what models need good enough. And I think intel could really play in that good enough market. And we just got to see how big that market unfolds. I think that's something we're going to want to keep an eye on.
- [30:49]
Matt Russell
And will some of those big players like Google, Microsoft, I'll use again Meta, have good enough models that they're building inside those businesses or, or these separate players in the industry? And I guess where I'm going with this is how much of Intel's potential growth in the GPU space will be shared customers with Nvidia versus owning customers themselves alone. How compatible are the two together and what's your outlook on that?
- [31:18]
Todd Alston
Like Matt, if they start owning customers, that's when you pop the champagne corks. It's about getting a small slice right now. I think it's about getting good enough. It's getting in the lab at some of these large players and we just have to see how this unfolds with them. Getting down the process manufacturing curve, building some tiles at TSMC, Gaudi 3 with some of the software compatibility, it's going to be all the above. And can we get a $3 billion slice out of $100 billion market and then taking it from there. So this is going to be quarter by quarter smaller until proven otherwise. So I'm not of the illusion we're necessarily even in that Bake off at the moment, but if we end up even getting in that room in that finals presentation, that's when it gets pretty exciting. But we'll have to see. This is something that 12, 18 months, we're going to know by the end of 2025. If this is going to happen or not would be my sense. It's a crawl before you walk before you run. And we're still crawling with it, to be honest, Matt. But I think there's a chance these guys at least crawl to walk. And that's the key. By 2025 and again getting to 3 billion and then beyond at the current valuation at 200 billion, you're just not paying for that right now. So that's where the investment gets really intriguing.
- [32:30]
Matt Russell
That perfectly answers the question. And I guess to flip the conversation back to CPUs, the legacy business that lost share over time, was that coming from smaller competitors? Basically taking those slivers of the pie themselves and eating it away from Intel. Where do they gain back share or grow that business in the future? Is it taking it back from those that gained ground on them or is there something else that's driving that segment, that base business back in the right direction?
- [33:02]
Todd Alston
Yeah, some unfortunately has been lost and probably isn't coming back. Like the Apple business. That was something that was in the equation through 2020. That's unfortunate. I feel like the next big driver for them is not only getting down these process nodes. We've talked about Meteor Lake coming out here in December and just getting better performance. But it's going to be, we'll see what the AI enabled PC. I think that's going to be a really big channeling what Pat Gelsinger is talking about. He's the most informed on this but really bringing out some of these next gen CPUs that very enabled for the AI PC. At the same time, maybe that market comes back. I mean the PC market was over 300 million units. It went back down here, let's just call it ballpark, 250 million units. We rolled off the pandemic and we're looking for consensus, maybe 8% growth this year. And can we get back to a $300 million unit market with the AI enabled PC? I think that's going to be really interesting to see how this evolves. If intel can really pull off some of these things they're working on, say backside power where they get better power efficiency, better performance per square inch of silicon, that would be battery life, performance, speed, latency, AI. All these things come together that I don't think it's one customer. They carve it all back. But the first thing is let's stabilize the share in the second half of the year 2025 with the AI enabled PC getting down these process technologies, backside power, power vias gate all around. If we can get our processors in a better position with amd, I think it's stabilizing and then there could be some chance for share gain in that 2025 standpoint, knock on wood. But it's not just one customer. It's going to be classic account by account and proof points. I think the key thing here for the stock is can they stabilize the share then we can start underwriting to the base business and that's the first stage we're looking for.
- [34:52]
Matt Russell
How long has that decline gone on for and have there been any signposts that you are seeing a bottoming out or stabilization?
- [35:00]
Todd Alston
Yeah. So really when you start looking up 20 going back to even 16 is when things started, it's hard to pinpoint a date, but just starting to fall behind on these things and it's been A real slide. And I think last year was just really tough. I mean you start looking at the share trends, last year was super tough for them and losing share. And I think you look at the first half of this year, it's still going to be a little bit tough, but getting Intel 3 out is just going to be really, really big for them. And again Intel 3 is really the second gen EUV in the silicon, in the fab, out to customers, getting specs. So really that's what we have to see. And so first half of this year I think they'll see some cyclicality that's actually more positive for Intel. They start descending at a slower rate and I think that's the first half of 2024 and then we'll see in second half of 2024 into 2025. So this has been going on for a while. Momentum is tough to turn around. And I think when you've had customers now betting on process technology and technology Windows, these are not easy. That's what makes semis so interesting versus software. You can send out patch things, get problems all the time on software. You have a mask and euv. This isn't something you go in and a couple guys in bunny suits where propeller heads fix it. It takes six months a year. These are setbacks and that's why we're having this conversation. But my confidence is increasing that Pat can pull this off. He's getting talent, he's getting better people on the board and we're just seeing things that are making us more optimistic about what's going on at the fabs.
- [36:30]
Matt Russell
And thinking about the CPU cycles, just more so the demand equation you reference there potentially in the future having AI powered CPU or laptops, computers, whatever systems that you might be using. What is the general trend line in terms of demand? Is it some type of GDP plus, is it fairly obvious where the growth is coming from on the demand side? And then the second part of the question is is there a strong case for a large replacement cycle once the server side of the equation comes in? Do I need to have a more high powered device to be interacting with servers, doing anything AI related?
- [37:08]
Todd Alston
So again we're delving the future so humility here but I would say if you look at intel, we're going to see a catch up this year. I mean if PC unit volumes are up 8 this year, we think they'll be up again in 2025. And that's really. Let's start with the pandemic distortions. I mean we saw enormous growth we went way over 300 million and PC shipping, there's a lot of PCs and we were down at say 250 billion last year. So we rolled off pretty hard. I think you're going to see a couple years of a catch up coming from the pandemic. Refresh some AI capability, maybe we get north of 300 million again on PCs. So that's a nice little recovery. After that I think your expectations need to be low single digits for unit volume. Can they get a little bit of pricing possible? So I think in the most bullish case could their Tam grow 5 to 7%? Maybe 7. You know that's possible. I would say let's not get into double digit territory on PCs at all. But I think mid single digits with some volume and price is possible especially if the AI PC works. So for us we had their business in 2023 bottoming out at 29 billion for client, roughly call it ballpark here. And in recovering 34 billion by 2026 we're seeing some nice recovery there. Could be bigger than that, but that's just conservative. Obviously data center is going to help drive that too. It's putting CPUs back in the equation. The CPU business I think we had somewhere 15, 7, call it revenues in 2023, 23.4 billion by 2026. So definitely some more growth on the data center side with CPUs and then obviously layering in that 25% operating margin there you start to see some juicier profits coming back. I think the CPU side could be interesting with a little higher growth if they can get back to seeing inferencing with CPUs. So we'll see how much of these ultimate workloads clearly are going to train models with GPUs and Intel hopefully does some of that with their Gaudi line. But I think it's the cpu, it's going to the edge. AI to the edge. And can that market get back to some decent mid high single digit growth Again I want to be cautious here. We'll have to see how this works out because right now there's a lot of distortions when budgets are getting pulled away that I only have a certain budget, I gotta buy gpu. So I'm crowding out other things. So there's a lot of distortion that first of all I think some of that can come back this year for CPUs. But long term could we see their markets grow 7,8%? That would set up well for Intel.
- [39:32]
Matt Russell
Variety of cycles is impacting the variety of Cycles, which is quite interesting, moving things around and thinking about the demand side. You've mentioned Gowdy a few times and the software component. I do want to make sure I understand that. And again, we'll assume it might not be much, but it sounds like there's a really interesting development with that when it comes to some type of software added into the equation. So is there anything unique from an economic perspective? And can you just share exactly what's going on there in terms of what it allows you to monitor or what type of dashboard you're seeing on the other side of that software?
- [40:08]
Todd Alston
Well, really what it comes down to is training a model. And I think all of us are becoming practitioners in this. And being a traditional semi, this isn't just DRAM and NAND chips and logic, and this is a different type of a cycle. There's a different overlay here for sure. So I just want to tread where we all are working on our expertise. But my how I'm framing this and how we're framing it internally is clearly there's Gamor's Law of Training. It's too expensive to train models right now. So being super simple, anything silicon providers can do with a GPU with software, to get on a Moore's Law of we're going to have more complex models and we got to reduce the cost dramatically to train them so they can be more evangelized. I think whether it's CUDA with Nvidia with AMD or Gowdy with intel, we got to get down to Moore's Law of training. And I think that's what Intel's trying to do on how they're combining their GPUs and packaging with the software and obviously accelerating that computing to get down a cost curve.
- [41:06]
Matt Russell
You touched on intel being a bit of a patriotic play. There is a lot here when it comes to geopolitical importance, and it's a unique situation when it comes to the supplier base and competition in the space. So can you just share a little bit of how you think about the geopolitical dynamics, whether that's a positive or a neutral dynamic for intel and anything that's happened that's a key milestone in recent years that would be supportive for Intel.
- [41:34]
Todd Alston
Geopolitics are super intense here, which I feel positive for intel. And only if you looked at intel as being an insurance policy. It's almost like you call up Allstate or Geico and the Gecko comes out. It's a little bit of an insurance policy on that. I take that really seriously and it's unfortunate that's the case. But we all know Taiwan's a great country, incredibly industrious people. But there's obviously risks there that we all know and don't need to get into at the moment. The same way you know, Korea and incredibly industrious people, there's a lot of proliferation risks there. So clearly stating the obvious, there's a lot of risks that every day you wake up and you don't know how the world's going to look and it's unfortunate we have to have this conversation, but that's how it is. So if you look at intel, they're looking to be the foundry to the world. Now that doesn't mean only chips being built in Silicon Valley, Arizona, Columbus, Ohio to all these different areas to New York and different areas that we've seen it also is Germany. They're looking at building large foundries, looking at investing a lot of capital ironically in Israel, which is another investment they're making, which is actually in another problematic area. But really for intel this is a play where the Chips act is a huge $53 billion. Intel has I think it's 8 to 12 billion in potential Chips Acts funding that could come their way. They're looking to be the foundry to the world. And I think if you talk to Pat, he would say we want to be that supplier and make more money than anybody else in terms of building some of these chips. I know it's a high lofty TSMC is going to be there but say excluding them make a lot of money because seeing the pandemic, Matt, there's a whole bunch of things there, signposts that say supply chains are fragile. And so I think this is an anti fragile investment and they're going to build chips on multiple continents. And part of the moat around this is they've got to get good people. You can build a fab, but you got to get the people. Matt I think we're speaking to, Pat obviously said publicly they've got to spend $50 million just training workers because you can't just put on a bunny suit and start managing a LAM research etching tool. I can't do it. You can't do it. We can talk these terms. So it's people training costs and quite frankly with the Chips Act, intel can't plow $125 billion of capital at a 30% cost disadvantage to Taiwan. There's no business there. So part of this is getting us competitive, getting up the cost curve, getting up the Tech curve and I think Intel's got a chance to pull this off, but it's incredibly important and I think ultimately if you're going to talk to those large incumbent players without naming names that they will be customers. But let's just whether it's Apple, Nvidia, Google, Meta, just different silicon. It's got to be built somewhere. And if I was running those, if I could do this relatively competitively, I sure would want another source. So we'll have see how this plays out. But I think Pat's quadrupling down to do this and if there's one CEO in the world that could do it, it's probably Pat. He's willing to put 125 billion on the table to do it. So we'll see.
- [44:24]
Matt Russell
And can you just provide the one on one on the chips act? Rough description of exactly what that meant.
- [44:30]
Todd Alston
It's really simplistically build something made in the USA where you're getting a 25% subsidy and the government as of the last writings is we're willing to $53 billion. So you gross up that by four or five times. I'm giving you directional math, but it's really a subsidy to help promote fabs built in the USA and do that to help them be more cost competitive. And there's other parts of it, it's pouring concrete and putting in very complex equipment, $250 million EUV machine, but it's also worker training jobs made in the usa. And so there's a lot to it that's in there to try to promote the industry.
- [45:05]
Matt Russell
We've obviously looked at everything from within Intel's perspective. When you think about competition, Nvidia is on an incredible run right now in terms of having success. But when you look around and look at the peer group competitors, are there things that you fear coming out of those competitors besides maybe invoice announcements, are there things that you monitor that they're doing? Things that are outside of Intel's control but things that you're monitoring closely?
- [45:33]
Todd Alston
We do. And that's one of the neat things here at Parnassus that we have three large cap funds. I manage our core equity product and CIO of the firm. So we also own Nvidia and so meet with them and have a significant position in that company. And we own intel, so we've in some ways hedged our bets on it and we remain, we've owned Nvidia since 2018 and remain positive on it. We also own AMD in our large cap growth fund so we own those assets. And there's a lot of scenarios, Matt, that all three of these stocks win. If you talk about a 500 billion doll semi industry going to well over a trillion by the end of the decade, there's going to be a lot of silicon and we need a lot of people to build it. So I think that's going to be positive. Now for intel, really the bet around to us is it's process technology leadership because that's where they're loading the fixed cost. So really it's about. I can't overstate this enough. It's Intel 3, Intel 20A, Intel 18A, and getting that. Because as far as intel as an investment, if they pull that off, maintain share in CPUs for server PCs and get even drops of foundry GPUs were okay. But it's like if something slips on the process technology, when I said 125 billion in five years, if that doesn't work, we have a big problem. Whether Nvidia or AMD announced H2 1, other great products, there's a lot of downside risk there. So that's really the point. And of course I'm looking at Nvidia and amd, and I'm also looking at the overall investment. It's one thing to look at these companies, but what are the capital spending numbers coming out of Meta, Google, Microsoft, Amazon? I want to know that the market's growing because as long as that's growing in investment and all of a sudden we didn't wake up nine months from now and say, gosh, we trained a bunch of models and we're not making any money. That's ironically the biggest problem. How are we monetizing? And some of that's even to come from companies like Intuit who are doing QuickBooks and different ways to monetize with AI assist. So it's John Deere with GPUs for Precision Agriculture. So we want to see this broaden out, that AI becomes bigger, not just in the big players, but it broadens out for other players that are investing in it.
- [47:36]
Matt Russell
And just to take it a step further and zoom out a bit more on the same ecosystem, how would you take those names which you mentioned, owning both Nvidia and Intel, and compare them to the niche players in the ecosystem, like a Cadence or an Integris, do you think from a perspective of the market growing and what you were just referencing there, would you have any differentiation between those two categories of businesses where the niche players versus the big behemoths in the market, I would say all.
- [48:06]
Todd Alston
Those companies are very well advantaged. Now the question is, do we have an AI peak in investment and is at the peak in 2025 is the debate. Whereas for Cadence and Synopsys you're just invoicing people, raising prices, bringing out new products. There's not as much of a cycle there. So that's one of the key things on those companies. You're not playing as much of a semicycle. On Cadence, Synopsys, they're going to grow pretty much regardless. Intel is going to be the tip of the spear on the CPU, PC server cycle, Nvidia on the AI cycle and AMD. So I think the opportunity is highest on Intel 1, Nvidia and AMD 2, and then cadence synopsys 3 on underwriting. But I think if you wanted to buy them, joke internally, go to Maui and drink beers for seven years and you're probably going to do okay. There's going to be some cycles in between. So that's how I would think about it.
- [48:50]
Matt Russell
You've laid out the story, the opportunity, both from I think the actual business perspective and then the low expectations perspective. Two unique things combining together. Is there anything else that we haven't talked about that you would point to just in terms of being an opportunity for intel or part of the bull thesis for Intel?
- [49:12]
Todd Alston
Yeah, thanks. There's so much here to go in Granular. Let me just say some things in detail is that first of all they do own a big chunk of Mobileye. Mobileye is an ADAS and they own roughly over 80% of that. On paper, it's a 25, $30 billion cap company that I know internally they're really bullish on and we'll see what happens to autonomous driving and growth in China and the whole EV area. But that's a call option. I don't think you're really paying a lot for that. It's embedded in the valuation, but we'll see. That could be potentially really interesting. Could that be worth multiples possible? Not here to say that it is, but that's a possibility. They also have their programmable solutions business, which is the old Altera. They paid roughly 15 billion for that several years ago. It's nested. These are FPGAs and programmable logic Devices, fuel programmable gate arrays. Pretty good business that's been somewhat mismanaged over the last several years that they're looking to bring in some outside capital and that could be potentially spun out. And FPGAs are great business could that be a double digit billions opportunity to create value possible. They also have a really interesting mask writing business. So when you think about EUVs and ASML, there's also drawing the masks. And they brought in some outside capital. I believe it was at a $4 billion valuation, they sold 20%. That could be potentially enormous. That that's going to be very valuable because when you're taking 193 nanometers of light, shrinking it to 13.5 nanometers and writing masks, they actually have some pretty interesting internal technology. They could spin that out that I don't want to put multiples on that. That doesn't pass the laugh test. But there's some scenarios that that's a very valuable business, that it's worth a lot more than 4 billion. That actually is meaningful to Intel. So you have an internal mask writing business that could be spun out. You have programmable solutions business that's going to be potentially monetized. You got mobileye on top of that. And so those are other things that we didn't even talk about that they could be value drivers. We'll have to wait and see. And then finally, just as vehicles go electric and EVs, there's going to be a lot of tremendous silicon opportunities in cars. Some of that will be mobilized, some might be in traditional Intel. So we'll see. But there's other lottery tickets that make this bar go from 5ft to 4ft to 3ft to 2ft.
- [51:26]
Matt Russell
My own commentary to add here would be we often talk about those things from time to time on this show. And potential lottery tickets that feel like extreme stretches. I think the case that you just mentioned, you provided things that would be way closer to being in the money. Or in the case of mobile, it's already in the money. It's just a matter of the upside potential. So certainly a lot of options, whether we call them lottery tickets, whatever it is, different assets that could provide serious value in the future. And that's really interesting to come this late into a conversation.
- [51:56]
Todd Alston
It is. I'll say one thing just for our process and for your listeners making it super simple. This is really hard to do this stuff and how many people in the world can do it? And it's really hard to do. And what's the value of solving the problem? So all these are hard to do. Not that many people doing it. Value of solving it high. And then you look at the incumbency and I think that's where you can start to put some tangible paths that there could be significant drivers in the last three to four minutes of this conversation that I would encourage people to dig in on those areas because those could surprise.
- [52:29]
Matt Russell
Yeah, I mean we covered asml and when I think back to that business and what it was for the long time and this money was going into it, they were doing hard things, but it manifested into something that I don't think a lot of people could have imagined at the time that it was taking place. And your description of that mass business and what they're doing there and the long list of very complicated things that they're doing there might be solutions that we're not even aware of in the future. Flipping the script a little bit into the risk category, execution just being an obvious one. And I think you've referenced that a few times. But are there other things that stand out or things that you're watching closely that you would fear keep you up at night when it comes to Intel?
- [53:08]
Todd Alston
Yeah, I would say a couple things. First of all is intel, we talked about their history and that could be a whole other podcast which we won't do at the moment. All the great things that this company's been. But they really lost being customer centric and I think with their whole foundry solutions they've got to really get that EDA ecosystem and become more customer centric. That's number one.
- [53:29]
Matt Russell
Can you elaborate on that a little bit? It's one of those things you hear a lot of businesses talk about. But what does that look like for the foundry business?
- [53:36]
Todd Alston
Yeah, so Intel's used to building their own designs and so when you have go fabless, let's just hypothetically say all of a sudden large company X is going to build chips and they want to use Intel. It takes a lot to port those designs and all the EDA standards and software to a very idiosyncratic design process into a fab. There's a lot of ground that covers between taking a design to taping out silicon in a fab. And I think that whole EDA design automation network is complicated. It takes a lot of very, very granular and tangible plugging in systems that you have to have an incredibly customer centric mindset. I think that's where Lip Bhutan joining the board, who is the former CEO and chairman of Cadence. Recently I was back in, I think November stepped up and bought some more stock. That's some validation that someone who knows how this works is on the board helping to shepherd this a bit. And he's gotten a bit more involved in some of the filings about some more RSUs and helping bid on Foundry. That's what it takes. It's really hard to do this. If it was easy, you could just build a fab and port designs in. So it's porting very specific designs that are bending the laws of physics where the ramp taping out, the cost, the defects, the yield, it is super hard. So that's just the risk that we're taking. And then I would also say just okay, a couple things like AMD just continues to crush it, they can't stabilize sharing CPUs and servers. And then finally we have a hard landing on the economy. We didn't talk about this, but macro does matter for semis. If we have a hard landing on the economy and you don't get a corporate refresh for PCs and some of the Mag 7 players who are investing a lot just pump the brakes a bit on investing. That could be tough. I mean the current semicycle looks pretty good. A lot of these upturns and we could get into that later, but a lot of the upturns in the last decade have been at least two and a half years. Things started to, at least for the stocks, bottom out October 2022. We're coming out of a cycle. It feels constructive for semis, 2024 into 2025. But at some point we overbuild. We could have an economic event, interest rates, inflation, geopolitics, wars. So all those things, the backdrop is constructive now in the cycle for a while here, but that could be a risk too. So again, I want to be thoughtful, pragmatic, don't live in fear. I think there's eight or 10 things that could work positive. There's three or four that could break negative. But overall the execution will be the thing to focus on.
- [56:01]
Matt Russell
Are your expectations that this cycle would be from a duration standpoint any different than previous cycles? I think you've alluded to it a little bit in terms of why it looks a bit different, particularly in the GPU space. But are you thinking about this when you reference that two and a half year period any differently than how you've thought about the past?
- [56:21]
Todd Alston
Yeah, it feels to me at this moment and I say this is where the humility comes back in. It feels longer. Matt, from the standpoint, say the stocks all bottomed in roughly October of 2022 and that's where the market did as well. So when you think about recovery, we had the pandemic distortions, so I think we're rolling those off and burning inventories. The automotive area still has some inventories and weaker demand. Memory looks like it's going to come out of its down cycle here as we speak. And then logic, GPU, CPU, server, PC seems set for a good recovery 2024 and then with the AI spending on GPUs I think the ecosystem of compute, they can't do it alone. There has to be compute on the edge and memory, NAND, DRAM, PCs, et cetera. So my sense is this could be a longer cycle than normal. History would tell me we've got another year of a really constructive backdrop for semis, at least for the stocks because obviously the stocks will peak before the cycle peaks. Could this cycle be longer? Possible. I'm not willing to pound the table and say it's a five year cycle but I'm willing to commit that at least in humility. The next 12 months looks really good and then we'll see what happens into 2025 with training inferencing the economy and the cycle. But it could be longer this time.
- [57:39]
Matt Russell
One I want to say that I speak to everyone, appreciate that you actually putting some numbers or timeframes around these things. We won't hold you to it but it's just helpful to hear and get the perspective. The other question I would have on cycles is and I know they go hand in hand but to the extent that a drop off in demand is the main driver and maybe tying it specifically to macro, so we think about 09 other recessionary periods. Do those cycles look a lot different than maybe those just driven by natural oversupply coming into the market resulting in a shift in things? Is there a big difference between the two cycles or the drivers of what would cause a correction in the cycle?
- [58:17]
Todd Alston
Yeah, I think this time around the AI investment is so large and those companies are so well financed that in the short term it's going to make the industry less cyclical. We're going to have to see how these models get monetized if it's truly less cyclical long term. But I think you could see potentially even a couple quarters of negative GDP growth this year and semis are doing fantastic and you'd be like how is that? Well, you have. The pandemic is distorting things too. You're coming out of a trough for PCs and CPUs and servers at the same time you have AI investments. So we could be sitting here catching up over coffee and there's a mild recession in Q3 but all the semi companies are beating estimates except maybe a couple automotive suppliers. So that's what makes us unique. So I think that's where we're at now. And we're going to have to see how this plays out by 2025, 2026, that when the cyclicality comes back in. And that's why I've learned from eight cycles and being 51 years old, you have to be dynamic, pragmatic, you gotta learn. If you don't have a growth mindset and learning all these cycles are a little different. And trying to kind of a youthful thought that each cycle is going to be different is really important. And this one's going to be different than the past. And I think that's where the deep pockets of those big investors make this cycle different than what we've seen the previous seven cycles. I've lived through.
- [59:33]
Matt Russell
That makes a lot of sense. And my natural follow up question would be in that 25 to 26 year period where you have the massive capital outlay they've built out, whatever it is that they're building out at this point is the expectation that there will then be a large drop off. Where would the run rate spend be from those players? I can only imagine it's the million dollar question. But I'm just thinking about whether this is foundational level spend where there's a massive amount of infrastructure being built that can then come down over time, or if this is going to be at least some new normal in terms of very, very high spend in this category.
- [60:10]
Todd Alston
Yeah, that's where actually some of our friends at New street have done some really good work. And just to quote some just mind boggling numbers, when you say $400 billion AI accelerator market 2027, that's 1.1 I think trillion in data center spent. Let's just say that happens. That's a super bullish estimate. I think they rightfully say that's bullish. They did a sanity check piece which I would recommend people to look at after this conversation. But I think we're going to have to see what the business models are. Are there multi hundred billion dollar markets in digital ads and AI assist co pilots and medical research, life sciences like I mentioned, even John Deere, Precision agriculture, intuit doing AI for QuickBooks where like hey, I operate 10 car washes and I can use AI to drive my business model to reduce the amount of towels we're using at the car wash and drop down 100 grand. There's going to have to be a lot of business models that kick in. So I think at some point it's highly likely we're Going to have a peak cycle where we have this gold rush investment. I feel really good about 2024, I feel pretty good about 2025. But somewhere out there is going to be that moment where the money's in the ground and where's the pay dirt on this and there's going to be some sort of a cycle within a cycle where I think as a long term investor, I'll frame it this way. If you're going to buy say Nvidia intel and not look at them for seven years, there's highly likely that 50 to 60% drawdown at some point. Could it be 18, 24, 30? There's a risk pocket in there and I think that's the question we're going to ask. So my expectation is we're going to experience that and then unfortunately some people will get shaken out in that. And then when we actually have all these models really trained and inferenced and out to 2030, if you exit then you're going to miss that next cycle. So there could be some risk management in 12 months. It feels like a bubble. That's another conversation. Don't feel like we're there yet, but there's going to be a drawdown at some point. Highly likely. And that's probably in that 2025, 2026 timeframe. At some point in there. We have to be thoughtful on that. We'll have to see. Don't know how to answer that completely, but that's how I'm framing it.
- [62:13]
Matt Russell
I love the nuance there and it's what I was hoping when I looked at your history in the cyclical space and just that next level understanding of how these things can ebb and flow. And even if you have a long term outlook, you could understand of what happens in shorter term periods. So this entire conversation has been excellent. Diving into intel. Also on the space. We close out all of the conversations with just a question about lessons that could be taken away from the business and maybe applied elsewhere. So when you think about intel, your experience, looking at the name, your history with the name, what would be the standout lesson from intel that you would think about in terms of looking for pattern recognition or matching across other investment opportunities?
- [62:55]
Todd Alston
Yeah, two things. I'll talk about intel and quickly the cycle on Intel. The first thing is that semi turnarounds take a long time and that's the opportunity and the risk. And so when Pat joined the company In February of 2021, he said this is a five year turnaround and that's just not what Wall street wants to hear. And so I think the key thing there was these things take longer than expected. And even as we wait for intel four, we wait for 18A 20A Gaudi three, you're going to have to be patient. And I've learned in semi turnarounds, especially on the device manufacturing side, you gotta be patient. But over time, usually patients will be rewarded if you stay focused on the signposts that we talked about. So semi turnarounds take a long time. And number two is in semi cycles, we all think we missed it. And it's classic to say, hey, the stock went up 20, 30%, I missed it. You probably didn't miss it in semis because these cycles can go on for several years and some of these stocks can go up a lot. And I think we all think we missed it and you didn't. Now there's obviously some point in the game, you're late in the game, but don't think you missed it just because you missed the first 20 to 30% of a semi trade in a cycle. Be patient with semi turnarounds. And finally, temperament is everything in being a cycle investor that these over extrapolate. When everybody in the earth is positive, it's time to get negative. When everybody hates it, it's time to get positive. And you have to have contrarian genes in this business. You also have to have a good temperament. You can have the highest iq, you could do the most work, know the most specs. If you can't handle the stress of it, that's what's tough. So stay chill, have a hobby outside of investing, go for a run, read books, do something outside of it, but just temperament, because investors will over extrapolate semis. And that's where the opportunity is great to be contrarian. And that's where I think intel is a contrarian investment in front of an exponential trend. That again, as we started the conversation, cyclical recovery, secular re. And that's what your mentality's got to be. And also around peaks, which I don't think we're at, but that's the key thing to stay focused on. So lots of learnings. And it's also a whole lot of fun because you get to learn about the future, future. And these things are just hard to do. When you talk about putting 60 miles of wires on something the size of your fingernail, it's just cool to talk about. So part of it's like, enjoy the ride. And that's what makes it fun too. So it's been fun talking about it, Matt.
- [65:14]
Matt Russell
It has, yeah. Equally from my side and I think from the audience side as well. Thank you very much, Todd. I appreciate you joining us and sharing the knowledge.
- [65:22]
Colossus
Great.
- [65:22]
Todd Alston
Look forward to seeing how this all plays out.
- [65:25]
Colossus
To find more episodes of Breakdowns ranging from Costco to Visa to Moderna, or to sign up for our weekly check out joincolossus.com that's J O I N C O L O-S S U-S dot.
- [65:41]
Matt Russell
Com.