From theCube Studios in Palo Alto in Boston, bringing you data-driven insights from theCube and ETR, this is "Breaking Analysis" with Dave Vellante. As an American, you can't help but root for Intel's CEO, Pat Gelsinger, to succeed. His vision to bring semiconductor manufacturing leadership back to the United States is, it's more than just a quaint, nationalistic sentiment. Rather it's a strategic imperative for the country, the country's military, its global competitiveness and access to future technological innovations in the AI era. But a strategy is dependent upon the success of Intel, both as a designer and a leading manufacturer of advanced chips. As such, this choice puts Intel in a multi-front war with highly capable leaders in both markets including names like AMD, Nvidia, AWS, Google, Microsoft, Apple, Tesla, and other chip designers, even now perhaps, OpenAI. As well, they compete with established manufacturers like Taiwan Semiconductor and Samsung.

Moreover, Intel's business model has been disrupted by Arm, which has created a volume standard powered by the iPhone and mobile technologies. And finally, China Inc looms as a long-term competitor, which further underscores the imperative. But the trillion dollar questions are, what are the odds that Intel's strategy succeeds? And are there more viable alternative strategies for both Intel and the United States? Hello, and welcome to this week's the key research, insights powered by ETR. In this "Breaking Analysis", we try to address these uncertainties and to do so, we welcome Ben Bajarin, who is the CEO and Principal Analyst at Creative Strategies. Ben, good to see you again. Thanks for coming on. Yeah, it's my pleasure. Thanks for having me on. All right, before we get into the discussion, I just want to take a quick look at the broad enterprise technology market. This ETR survey data, it comes from around, a little over 1,700 IT decision makers, it plots net score or spending momentum of a platform on the vertical axis and its presence in the data, which is kind of a proxy for market penetration on the horizontal axis. And this market accounts for about 1.8 trillion of worldwide information technology or your worldwide IT spend. If you think of the whole ICT market and you bring in telco, it's probably up around four trillion. So this is only a subset of the market. But nonetheless, the graphic shows that AI has become the sector with the most spending velocity since the announcement of ChatGPT.

But the real point is AI is not going to be a distinct sector, we think, rather it's going to be embedded everywhere, in every sector, from IT hardware and software, AI PCs, mobile devices, consumer electronics, autos, virtually every product is going to have AI and that AI will be powered by silicon chips. So Ben, let's take a look at the state of the market and the key trends that are relevant here. The market has changed dramatically over the past two decades. You got firms like AMD and IBM got out of the foundry business. Arm, we talked about capitalized on the smartphone revolution, completely disrupted the established business models. When PC volumes peaked in 2011, Intel's monopoly began to decline. TSM became the dominant manufacturer. Firms like Samsung rose up along the way with many others including a spate of novel chip designers like the ones that we mentioned before. So Ben, how would you describe the changes that have taken place in the market over the last decade or so? Yeah, I think you point to, you know a couple of those key transitions. Obviously, anybody who's focused on this industry and looked at it for a long time recalls the day where Intel was really the leader in product design and foundry manufacturing process technology. Obviously, when we look back and we say, what happened to Intel over the course of the last decade? We look at things where there're elongating cycles of Moore's Law. They didn't keep up with the same pace of Moore's Law. They moved to, it wasn't tick tock, it was tick tock tock, tock, tock, tock. All of that driven by economics and foundry and their ability to keep their foundries full. In that timeframe, TSMC started becoming more predictable, more conservative as a leading edge foundry, meaning that they weren't trying to be too aggressive in their no jumps. And because of that, they were able to keep relatively straightforward, no jump cycles and process development cycles, which became the best option for a lot of companies.

Obviously, also during this time, Intel was not a really a third party manufacturer, which I'm sure we'll discuss but you know, they tried foundry but they didn't really try foundry like they're trying now. So they really weren't a great second source. And you had scale, right? You had AMD, you had Nvidia, you had Apple, you had Qualcomm, you had MediaTek, you had all of these companies shipping millions upon millions of chips every year needing somewhere to manufacture that, needing to do so on a quality leading edge process and that became TSMC. And through the course of that process, TSMC became the leader in process technology, which was once accorded to Intel, and now they are still that. Now so, the dynamic of this battle still comes back to who has the best process technology so that the environment, all of these T-chip designers who want to make their products competitive on the best process will come to you and that is, by and large right now, TSMC. Yeah, thank you for that. You know, it's kind of actually embarrassing when you look back with hindsight, how COVID woke us all up to the digital reality and the notion that we rely on semiconductor technology for virtually everything. We were kind of asleep at the wheel there. And folks, they often talk about the global semiconductor supply chain, it includes companies in many geographies but as you know, Ben, it's really not a global supply chain. It's a supply chain that's highlighted by some very tenuous choke points. So this graphic depicts just some of those. And we've started in Taiwan, everyone knows, we just talked about the TSM and the razor's edge that they're on with the Taiwan Strait and China. And Apple and its customers rely on TSM for those advanced chips. And as I say, China looms large.

ASML, a Dutch company, is the sole supplier of EUV lithography machines which are required for making the most advanced chips. These machines, they cost upwards of $100 million and they have like 400,000 parts, it's mind boggling. And there are no alternatives to ASML for EUV machines. As you know, folks, Silicon Valley is the leader in EDA and Japan is where you get most silicon wafers and other like thin film technologies from a pretty small number of specialized firms. And even China, which is currently enduring the No Chips For You Act, I call it, is where critical rare earth elements are found and really highly concentrated deposits which make that country the leader in global supply. So these elements are critical in many phases of chip manufacturing. So Ben, how would you describe the current state of supply in semis? How do you think about these choke points and the challenges that they present to chip supply? I think it's one of the things that keeps everybody in this industry, especially those who make semiconductors, up at night because you do realize that there really isn't a central location where you can just have full autonomy and control of your silicon. You know, as much as we like to talk about this idea of nationalizing semiconductor manufacturing, no country will ever have 100% of the needs as given the graphic that you just said. It's a globally and tightly intertwined chain that I think people don't fully appreciate. Which is why collaboration and being able to work together to have a smooth supply chain is necessary. And when that doesn't happen, things get really bad. We've seen this over the course of the last few years. So I think it's one of those things where you've really got to pick how you manage your supply chain, whether it comes to leading edge or trailing edge, where you can try to minimize any risks 'cause there's always going to be risk, right?

We're not going to see alternatives to ASML like you talked about. We may have foundry opportunities in terms of diversity but there's still going to be a range of foundries that are located in different countries and the political tensions around those countries. So it's not going to change, so sort of my point, these choke points will remain, there might be a little bit more balance, again, as we get Intel competitive and Samsung as they continue to evolve in terms of foundry. But those core elements; the wafer supply, testing like you said, EUV machines, we're not going to make those somewhere else. So it still really is a global operation and any tensions in any of those environments can throw things off drastically. So something you said I want to land on for a second, so you said it's very unlikely that there's going to be like a national supply chain, and so that applies obviously for the United States, but also applies for China. A lot of people are very much concerned about China taking over TSM, et cetera. But again, if I understand you correctly, your premise would be that it's unlikely that even China would be able to create a national semiconductor supply chain because it's going to rely on EDA software from Silicon Valley, other components from Japan, et cetera. Is that a fair assertion? Correct, yeah, I think if you just look at everything that China's trying to do, where even today, you know they're using SMIC, which is sort of their largest manufacturer trying to do things in the leading edge, not able to get over the EUV hump, they're trying to invent processes themselves, it's not going well. They still obviously need EUV machines and then there's political strengths on top of them that's making that challenging. Where do they get wafer supply? It's a very mature ecosystem, right? If you think about it, right? The semiconductor industry is one of the oldest, just in terms of technology, if you look back to the invention of the transistor and whatnot, and it is a very mature global supply chain and it is intertwined for a reason, based on where you get materials, who has those core expertise. And this is one of the reasons why it'll even be hard for any company to acquire or any country to acquire those assets, China included. So I don't have any belief that that changes, it appears that as much money as China wants to throw at this problem, it's not just a money, it is also a talent and it is an equipment problem and doesn't look like really any company, any country is going to get over that hump. But we don't see anything changing, looking like that's going to change. And it's enormous market, I think, you know I've seen estimates of $0.5 trillion growing to a trillion dollars, by the end of the decade you may have better numbers than that. Even Lisa Su said, the AI chip market is 400 billion. So I don't even know, that may be additive to the numbers that I just shared. But I want to pivot to Intel's strategy generally and its foundry strategy specifically. These are just a few of the relevant promises Intel has made, its integrated device manufacturing 2.0 strategy is really designed to make manufacturing more competitive and efficient, both for internal and external customers. So they promised to save 10 billion by the time it exits 2025. The company used to have gross margins in the 60% plus range as well below that today in the mid-40s and it's promising to get back there eventually with a very aggressive 40% long-term operating margin goal.

I mean, those are Oracle-like operating margins. Intel is building out, and they don't really make much hardware, Intel is building out at least new six new fabs that I could count. One in Ohio has been delayed but its goal of having 50% of global manufacturing outside of Asia is what that is all about. And its objective is to return to process leadership and be the second largest foundry behind TSM by the end of the decade. So Ben, Pat Gelsinger, he called Intel's first foray into foundry, he said it was a hobby, which I think was probably accurate. Why do you feel or do you feel this time around it's going to be different? So I think it's important to look at like what happened the first time around that made this difficult? Intel said they would make chips for other people but yet they still drastically prioritized Intel product technology, they didn't share that technology, they didn't share the leading edge with other people. And so you look back at just some of the customer stories of why that didn't go anywhere and it was sort of obvious, right? Like Pat said, it was a hobby, it was a very, very weak attempt. It was one of those things like, "Yeah, we'll say we'll do it but at the end of the day, "we're really not committed to a foundry effort." This is very different from a lot of different ways, not only that there is true division in the company in terms of what Intel product has access to and sort of the walls that are put off between Intel product technology and foundry, which is an important reason, right?

Because if I am a competitor, let's say I'm AMD, I have no idea if AMD's ever going to use Intel foundry but let's just use this for hypothetical. I would want to be certain that my IP, my technology, my strategy, my roadmap is not being shared with Intel, right? 'Cause this is an Intel Intel fab. So there is walls that have to be developed that prior to this, really nobody had any confidence was, it sounds like customers are certainly feeling that there is a process that Intel has put in place to keep kind of intel products isolated from others, is there? And then on top of that, there's services part of this business, right? TSMC is at the end of the day, not just a manufacturer but they are a services company, right? Their customer service is very good. They work with their customers, they're viewed as a partner, they go well beyond what's needed in order to deliver for those customers.

So there's a services angle to it. So when we looked at Intel sort of doing this, those are the things we looked at, right? How separate is this entity? Is this truly a company that looks like it's functioning on its own, even though Intel's essentially the funder of it, which is kind of the way I think about it. They're giving them capital, CapEx that they need, that they're not getting from either the revenue or from customers like TSMC does. And so, they're basically floating foundry in a broader level. But at the end of the day, are they just treated as a customer? And now that it looks like those walls are there and they are getting customers, you know a rumor came out this week that NVIDIA is going to use them for advanced packaging and some capacity. We know MediaTek has done a deal with them. I think they now have five major customers for Intel Foundry.

So it sounds like those customers are viewing this as a legitimate foundry effort. The wins that they're getting, some of the business that they're getting for capacity from TSMC makes it feel like this is very, very different. Again, they're not out of the woods, there's a lot of execution that has to come in in terms of their four nodes in five years, meaning that they need to execute on 20A and 18A and whatever's beyond that, predictably for customers. And really it's that customer service point that I think has drastically changed, at least that's what we're hearing from people who are getting in and seeing the Intel toys for the first time. And broadly, feedback has been pretty good in terms of impressiveness with the technology that's there. And that's again, I think this first time that truly Intel product is being treated like a customer instead of the priority that they were with Intel foundry this go around, and that's absolutely essential for this to work this time around as well. You mentioned four nodes in five years, earlier, you alluded to the fact that TSM was much more cautious about pushing that technology. Why do you feel that Intel will be able to achieve that? I mean, Pat has been very vocal about, "Hey, people said we couldn't do it." TSM took the more conservative approach and that paid dividends. What's different about Intel and will it be able to tap those same benefits? I think part of the problem too that Intel had was they made some very bold density targets in their transition while at the same time, they hadn't fully embraced EUV, right? I think when you look at one of the major points that they backtracked on was the way that they embraced EUV for foundries. It was very mature, they tried something else that didn't work. And as a part of what they tried, they also tried to be very aggressive in their density goals. None of that worked, right? Those are the two probably fundamental reasons that cost them process leadership. They've now embraced EUV and they're now bringing these technologies in a more predictable fashion, backing off some of those density targets, even though it looks like they're relatively on par with TSMC. So those two things kind of changed this, right? I mean, a lot of this was, again, TSMC'S playbook, which they're using.

There is a difference though coming with 20A and 18A that I think is really interesting because Intel is trying to do two things at the same time. And while it sounds big and difficult, it does sound like they're executing these two things, which is they're bringing backside power, which is a new innovation to manufacturing. Others will have it as well, TSMC, Samsung, et cetera. But they're bringing backside power to a node at the same time they're moving to a new transistor technology in RibbonFET, which is essentially nanosheets. So these two things, nanosheets, which is the broad industry term which everyone will use, and backside power, which is the same, are coming at the same time for Intel with products called RibbonFET, which is the transistor design and PowerVia. Initially, like this was very concerning to me because they're trying to make a leap here and their track record hasn't exactly been stellar in some of these things.

And so there was worry that this would cause some delays, right? That's kind of what everybody was looking for. All the investors that we talked to, customers in the market they're like, "We're just waiting for them to say it's delayed, right? "2OA is delayed, 18A is delayed, whatever." And because of we haven't seen indications of that effect, it sounds like they're taping out customers or sampling on both of these things. The signs are good that both of those transitions are going to happen at the same time which is really impressive because TSMC is not jumping to both of those at the exact same time to the degree that Intel is. And it looks like Intel has pulled that off which does put them in a competitive position. I'm not going to say leading because we can debate what leading process technology looks like all day but it puts them in a position to be in the conversation with a very quality process and two new innovations in transistor design and backside power that it does sound like a very, very attractive to customers.

So it does look like they're on track, I think that's positive. Obviously, a delay would be really bad at this point, but all of Gelsinger commentary and what you're hearing from customers doesn't sound like that's going to happen. That they are going to meet 20A and 18A timelines, which to me feels positive. Ben, I want to make sure I understand this. Thank you for that, by the way. So my understanding is that TSMC started with EUV back as far as I think 2017, so they've got, let's call it a six year lead. But you're saying that RibbonFET will potentially allow Intel to maybe not leapfrog, but catch up, close the gap. And if I understand it correctly, Intel is leading in that regard whereas TSMC is perhaps being more conservative. Do I have that right? Yeah, and in two points, right? So again, Intel calls this RibbonFET, this is nanosheet transistor design, which both Samsung and TSMC will have. Intel's getting there more aggressively first at scale. This is for whatever process it technically equates to, which is roughly like two nanometer, maybe it's just sub two nanometer and backside power, right? So backside power which Intel calls PowerVia, it's the new power delivery mechanism to transistors. Again, it's a really big leap for the first time that we're going to see this type of design. So we've got a big leap in transistor design, right? If you go back to Intel's tick tock days, right? A tock was a kind of big transistor shift forward or redesign. This is the first one in probably a decade in terms of significance while simultaneously adding a new power delivery mechanism in PowerVia, again, it felt really ambitious but they are right there in terms of delivering at the leading edge and I've talked to a lot of folks in the industry who actually think PowerVia, so their backside power technology, which is very different than the way that Samsung and TSMC is going to implement it, is actually extremely innovative. And there's a lot of people believing that that will reward them some very good benefits competitively from a process standpoint compared to others. And do you feel that, whatever, six or seven year delay in adopting EUV is such that the RibbonFET and the backside power approach that they're taking will allow them to close the gap, even though it sounds like, well, TSM and Samsung are maybe taking a different approach, they are going to be using similar technologies probably in a more compressed timeframe. But do you feel like based on what you know, that Intel's approach, despite that six or seven year lag, will allow them to maybe not leapfrog, but close that gap? Yes, I do. And I think that their commentary as well as what you hear from the industry and even the fact that TSMC'S chairman, I think two quarters ago on an earnings call even addressed this saying that they're not worried, they think they'll still have process leadership. Again, that's a huge debate like what does process leadership mean? The bottom line is they are going to be right there at a node around two nanometers with a brand new transistor design in RibbonFET and an entirely new way to deliver power to the transistor via backside power. So whether they're in the lead or not, that's sort of irrelevant to me, I don't think this whole like, I need to be in the lead to win customers plan, I think it needs to be a quality process technology that's competitive and a viable alternative. That's more of what I focused on. And from everything we're hearing and we're seeing, it sounds like they will be there for sure with 18A in the 2025 timeframe, probably scaling in 2026. So I would say, so let's kind of try to define a little bit at least frame what we mean by leading. And I agree with you, you can get into these silly debates about what is what but you said quality, it's competitive, it's a viable alternative, I would add, it's got to be profitable and it's got to be done at volume. Would you agree with that? If they can get to those characteristics, can we agree that that would be considered leadership? Yeah, I agree. And again, your volume point comes back to they need a customer rather than Intel, right? They're not going to become the second largest foundry by revenue, which is the point that Pat has made, right? He's not necessarily saying second foundry by volume but by revenue. So that's an ASP call. So again, making high margin chips like NVIDIA's server chips, for example. Advanced packaging in chiplets, very, very high margin designs. You don't need the same scale, you're going to get higher dollar amounts, content dollars or dollars in production but they need more than one customer. They need two, they need three wafer scale customers, which is the term that we use. Somebody who consumes a lot of wafers in order to make sure that that foundry that they're spending now 30, 40, sometimes people are talking about $50 billion of foundry, which back in the day when we were all doing this in the 2000s, a foundry was $10 billion.

Right now it's 30, 40, 50, that's a tremendous amount of money that you need to capitalize and make money on in terms of how you fill that in capacity. So it's all those things, right? Wafer scale, high ASP dollars, advanced packaging, which is the chiplet technology comes into that. But they need a series of those wafer scale customers to come into play. And they need to be competitive, right? The process needs to be competitive and it needs to be something that a company believes they can build on and compete with other players who aren't building on that, right? Who might be building on TSMC, for example. So if all of that happens, it's a very positive story for Intel. 'Cause at the end of the day, right? If Intel pulls this off, Intel Foundry will be more valuable in terms of revenue than Intel product, right? That's the end goal here. Like that has to happen as a part of this, economically.

But all of those things have to happen; wafer scale customers, quality process and technology, and again, customer service and predictability, right? They cannot have delays. TSMC doesn't have delays, they can't have delays. So all of that execution has to play in their favor for this foundry turnaround, which is tied to the intel turnaround to take place over the next 10 years. So that's a good lead in to what I want to do next, which is a take a financial snapshot of selected chip designers and manufacturers compared to Intel. So this really underscores the multi-front war that we talked about earlier. And just to caution, these are rough estimates, it's not precise, it's not made to be an investment vehicle. But what we try to do in these things, and we pull data from trailing 12 month revenue, we do revenue multiples off of that, the point is really to try to get a better understanding of the business model and the framework and the quality of the business. So that does tell that story. So you look at Intel, it's revenue growth that's been stagnant, although it is promising improvement. Its gross margins, as I said earlier, have been decimated along with its operating margin. Its balance sheet is really not large enough to fund all those fabs, so it's got to do so with free cash flow out of its product business but the free cash flow margin is really tanked as well.

And so all of that has crushed the company's valuation. Intel was worth around 0.5 trillion in 2000, along with Cisco, was one of the most valuable companies in the world. It's now trading at a revenue multiple that is in the low single digits. So if you start in the light blue, there are two manufacturers we're showing; TSM and GlobalFoundries. TSM is the leader, as you can see by its financials relevant to GFS. Samsung's another large player but it's a giant conglomerate, so we can't really compare it and break it out. But TSM is of comparable size to Intel, a little bit bigger, and it's worth almost four times Intel. GlobalFoundries was formed by taking AMD's manufacturing operation and later it, basically, got given the IBM microelectronics division. It had a deal with IBM to manufacture 10 and I think seven nanometer technologies.

But it had to back out, my understanding is it was just too hard and too capital intensive. And so IBM sued the company and that was kind of ugly but GlobalFoundries at least was able to survive. Ben, how do you see Intel Foundry services stacking up? Today, we talked about that quite a bit, relative to these two and Samsung, they have their site set on 10 billion, you know maybe going to do a couple billion in 2024 but my question really, to add on what the conversation earlier is, around the learning curve, you know Andy Jassy has this comment that there's no compression algorithm for experience, the learning curve for Intel is obviously steep. So how do you see them effectively competing in foundry? Add to what you said earlier, and I'm specifically interested in that sort of compression algorithm and the learning curve and your knowledge of what's required to really get on that curve in this space. Yeah, so I still think TSMC is probably the best comparable to use in an analysis. I mean, Global Foundry is there but for the most part, as you pointed out, they compete at the trailing edge, right? So they're at stuff north of 10, 14 nanometer and beyond, right? So that's microcontrollers, that's a range of products but they're not competing at the leading edge, right? Everybody who makes products today, like you said, Apple, Tesla, Nvidia, AMD, Qualcomm, MediaTek, et cetera, even the custom stuff going on at Amazon and Microsoft, they want to be on the leading edge, right? They may not get to the leading edge because there isn't capacity and somebody named Apple generally absorbs the vast majority of that first generation leading edge capacity but they all want to be there, right? So the competition is really for the leading edge.

And interestingly, when we looked at this from Intel for the past two years, our whole thought process was well, they're just going to focus really on the leading edge. And then recently they just did a deal where they're actually leveraging a 12 nanometer process technology with UMC and they're doing that as a collaboration for a custom process, which is actually a really big deal, and they're going to use their depreciated assets as a part of that, which is a very TSMC thing to do is to take fabs that are already depreciated and you're making money on, you're just printing money on, and you keep those full. Didn't think Intel was going to do this and now they did, right? So just an interesting creative development where they're leveraging an asset, they're taking this book out of TSMC and they're making money from depreciated assets which is very smart.

This is a couple years away from being into implementation but that deal is done, super, super interesting. But at the leading edge, right? This is the part to me that's the most interesting because there is no capacity like you talk to everybody out there and we are up against the absolute limits of supply. We are entirely supply constraint at the moment, at the leading edge. And for whatever reason, you know I've heard different things about Samsung. Samsung doesn't seem to be what many consider a viable alternative. We don't hear Nvidia, we don't hear Qualcomm, we don't hear MediaTek. We don't hear these companies looking to Samsung at the leading edge. So when I say leading edge I mean something three nanometer, which is what we're at today and beyond, right? So moving to two nanometer, 1.5, wherever that goes.

They're really not considering Samsung. Like I don't know if that's a process thing, I don't know if it's a commitment thing, capacity, whatever the issue is, you hear more about Intel being front and center, having customers come in the door, evaluating their manufacturing process, evaluating their advanced packaging, and you're seeing them now get deals, right? There's a lot of customers, like I said, I think they have five at this point, we don't know who they are. We just know that they're big, they say they're wafer scale. Like I said, Nvidia has been rumored but we don't have any of this terribly confirmed at the moment. But the reality is they're getting customers for the leading edge. Samsung is not. So I say that because again, when you unravel this back to the fact that we are supply constrained, the entire industry needs a second source.

And if for whatever reason that's not going to be Samsung, Intel's in a really good position to take some of that and have some of that leading edge capacity that Nvidia, AMD, Qualcomm, et cetera, cannot get access to with TSMC. So I think when you put it sort of in that broader perspective, it paints where they're at with this opportunity at the leading edge but I throw in this 12 nanometer deal just to say it does look like they want IFS, right? Intel Foundry is willing to be creative and leverage those depreciated assets and foundries to also make money to offset that capital burden as well. Which I didn't think was going to be the case probably a year or two ago. Interesting. I mean, you're right. I mean, hey, if Samsung doesn't aggressively compete, that opens the door for Intel as a viable second source. All right, I want to come back to that same chart and look at the green. These are a few of Intel's chip design rivals. They got Nvidia, AMD and Arm, which really doesn't, I mean, they license their architecture, which is why Arm has 100% nearly gross margin. But all three of these firms have, to me anyway, more attractive financial profiles than Intel currently has and far, far better multiples. Of course, Arm just popped this week when it blew away its numbers and raised, but the other thing is they don't have to fund factories. So Pat Gelsinger, Ben, has said it now costs $30 billion to build an advanced chip making facility. Can it throw off enough cash from its chip design division in client, in data center, et cetera, to fund its foundry aspirations?

You just mentioned sort of a clever move that they're making with depreciated assets. You know the question is, is that enough? And at the same time, can they compete with the companies that we just showed in the blue, the chip designers? Yeah, so I think it's a good, it's a good way to again, understand what is essentially a separation of Intel, right? I get this debate a lot, honestly from folks on the street, right? That Intel should just split the company. The challenge there is where would they get the money for foundry, right? Someone has to give them that money. The government is contributing, they have opened up and said, "Look, if anybody wants to co-invest in these foundries, "we'll run them, we'll use our assets, "we just need more capital." Because this is a high fixed cost business. That's the single thing drawing down their margins is how expensive it is to keep funding these fabs, right? And they need to be full, right? They can't just be capacities and Intel just did four nodes in five years, they need to amortize those smartly, but at the same time, they need to keep those factories full with volume.

So there's the Intel product, which I really less worried about, right? 'Cause at the end of the day, Intel can make product to TSMC, right? They do make some products at TSMC, they make some tiles for Meteor Lake right now as a part of their chip load design at TSMC. For the next roadmap, Lunar Lake and others, they're making more tiles at TSMC, in fact, more than Meteor Lake. So they are actually a customer and in some cases they're becoming a large customer. So at the end of the day, Intel product can keep making their, that doesn't change our capacity problem, we'll still have a capacity issue with TSMC if Intel Foundry doesn't work where now we just got all these people really fighting for the leading edge, that would be a hot mess. But the reality is, from a foundry standpoint, right? It's, not like, the turnaround is just not soon like, this is the thing I keep trying to tell everybody like, the economics of this is a 10 year journey.

Because if we're just going to start at maybe some degree 20A, but really 18A, like 18A is the one that it sounds like most customers are interested and beyond, so what's beyond 18A. That means that we are looking at the economics of them having to monetize those, make their money from that point forward versus that point backwards. So that high fixed cost business still remains intact, Intel product can be there, Intel product can compete. I'm not too terribly worried about their designs, it really comes down to is the process technology good enough to get enough people so that they keep those foundries full going forward? And those pay for themselves. I think it's going to be hard to monetize these last few despite what they're doing with UMC. They just don't do that the way TSMC does. So their margins essentially will still be on the rocks for the next few years.

But that's to me, the sign that includes this, right? If margins improve because of both product, but also because of the revenue increase in foundry, then that would be the sign that they're not just sinking all of this money into additional CapEx to fund foundries that aren't turning a profit. And that's where I think the financial models for foundry as itself become very clear. Intel's going to release this at the end of the month, to a degree, they're going to release this, they're going to talk more about this at the end of the month with their IFS Day where they get very, very deep in the weeds around IFS strategy and what that's going to look like as a business, the financial model for IFS, so that will help. But it's really a go forward from here, can margins improve? Can they monetize those foundries? Most of what the last few factories have been probably won't see that same level of monetization, which is again, why this is a much longer total financial and value turnaround than a two to three year window. Yeah, I mean, the numbers are astounding me. Okay, so they're going to maybe squeeze $10 billion of efficiencies out of foundry and internal efficiencies, that pays for a third of a leading edge factory. You saw Sam Altman just yesterday raising between five and seven trillion for semiconductor manufacturing. I mean, it seems like those are the types of numbers that we're going to need. But I want to revisit Wright's Law, we've talked a lot about volume, we all know Moore's Law, but the lesser known Wright's Law becomes important in this discussion. So Theodore Wright, he was an aeronautical engineer, famous for formulating this thing called Wright's Law, otherwise known as the experience curve effect. It basically says for any manufacturer or any product, as the cumulative number of units produced doubles, the cost per unit goes down by some constant percentage, let's say 15%, just pulling a number out but that's probably pretty reasonable.

Combine that with Moore's Law and you get a duplicate of effect. And this is highly applicable to chip manufacturing 'cause each new generation of technology initially costs more to build per unit than a previous generation so when you start manufacturing a new process node, you lose money. So Intel's monopoly was one when it beat risk. And our premise has always been that it did so because it became the PC standard and PCs were the volume king until early last decade around 2011, they peaked, notwithstanding a bump up in COVID. And today, Arm-based designs have a volume advantage, probably 10X the wafer volume. So Ben, given that Arm has that wafer volume advantage relative to x86, share your scenario as to how to Wright's Law plays in this game of semiconductor manufacturing. And how does Intel, whose volumes are basically flat, you've kind of alluded to this but go right to the heart of it, how does it get to a point where it's not bleeding money trying to catch up to TSM? Yeah, the big one just goes back to the point I made about needing wafer scale customers, right? They need people who will consume thousands and thousands upon thousands of wafers per month in terms of production. As you pointed out, that used to be Intel when it was really the PC industry and Intel had 99-ish percent of volume and chip sales at high end compute. It was easy for them to monetize that. They were their first best customer, to use an Amazon term, at high margins and getting leading edge quality products, it was very easy for them to keep monetizing and move to each fab because they knew they could fill it with Intel chip capacity. That's the big sort of run here, right? Intel produces something shy of 300 million chips a year, obviously, like I said, ASP is rising, so that helps.

But you absolutely need wafer scale. And so that's where I think, you know they've been very clear, in fact, and Arm has announced this several times on the last few earnings, they have pointed out that they have a deal with Intel. Intel is in a position to make Arm chips for anybody that wants it, right? You got Broadcom, you got Marvell, you got Tesla, I seriously doubt Apple will ever do it but maybe if the government forces some degree of chip design for local products or government to be made in the U.S., maybe they have no choice but they're very, very deep with TSMC. You know, Nvidia, like I said, another one that's coming there. They need those high margin, those wafer scales and we know they have a deal with MediaTek on IoT, maybe that could come to Arm chips, maybe that could come to Qualcomm, right? And I think Cristiano Amon has said they're not opposed and they've been talking to Intel and exploring it and they're going to pick the best process, right?

That's the thing. So that's why it comes back to having a true competitive process, a predictable roadmap and a real customer services like mentality which is the structure of IFS. And if they do that, they're in a position to get those wafer scale customers as, again, those wafer scale customers can't get access from TSMC and I think Qualcomm, MediaTek are good ones. MediaTek is a Taiwanese company, and I know they're very, very entwined with with TSMC but Qualcomm to me is a good example, right? Qualcomm has not always used TSMC, they've used Samsung before. They compete with Apple, I think you'd probably argue that they're one of Apple's primary SOC competitors for the things you talked about; smartphones, obviously, in moving into automotive, they want to compete in PCs.

They don't have access to anywhere near the leading edge that Apple does because of apple's involving with TSMC. So if anybody, in my opinion, could completely change this game and would be very interested if Intel cruise, it's going to be Qualcomm and that is a wafer scale customer and then some, and I think if they could get that business, that would drastically change foundry. And at the same time, meaning that Qualcomm and team have vetted that Intel process technology is competitive or even better than TSMC'S leading edge, that even puts Qualcomm in a much better position to compete with Apple in ways that they can't right now because they're not going to be on three nanometer, they're not going to be on two nanometer to the same degree that Apple will for maybe two years behind, just in terms of how this runs out in customer priority for TSMC. So they could use that desperately and if that all plays out, they're in a very, very much stronger position to compete and they're a wafer scale customer for Intel.

So they're the one, based on sort of your point, right? Which is where do they get scale? They're the big one. They're a wafer scale customer that could be primed to work with Intel and even better compete in the market than they can today because of their limitations to the leading edge with TSMC. You know, to your earlier point, because they've separated church from state, and to the extent that they can get that quality and the competitiveness become a viable alternative, get their cost down, because of what you were saying about we're maxed out on capacity, there's no reason why a Qualcomm, if those other conditions are met, or I would think even, I don't know, would Apple potentially to hedge its bets, use Intel as a second source if it can get there, why wouldn't it? Yeah, it would and again, we don't know what's going to happen with Taiwan. I mean, that is a risk, right? If for whatever reason we were cut off from Taiwan, from chip access, Intel is really the only option in my opinion because even Samsung who has some scale here, it's again, not the leading edge and again, also not really a U.S. company. So there is still some risk but really Intel would be really your only option, right? So we kind of need them to get there. Again, it sounds like they'll get there from a capital standpoint but even, you know, you mentioned what Sam Altman is trying to do, which I think is really interesting also a super complex problem. He's going to be at Intel's IFS Day at the end of this month, so we'll see what he says. But Intel has been very vocal. They said, "Listen, if anyone wants to help fund "these things, we will run them, "we will use our manufacturer expertise, "we will help put customer products in these." And if that's a semi-custom design, like they have all of the means to do this.

They just need those customers. And again, the hinge point had been, they were really not a competitive foundry and their process was behind. As all of that shapes up, things look very, very interesting with all these dynamics we're talking about politically, the fact that you can't get capacity from TSMC, that there are companies out there like OpenAI, even like Microsoft or others, who would love to be at the leading edge and can't, right? They just can't because Apple is there, AMD is there, NVIDIA is there, at TSMC. You're just not going to get leading edge wafers, they're desperate for these things. So if it's even competitive, and that's the the thing like, eh, I keep saying, it doesn't even have to beat, it just needs to be competitive. There are customers who desperately need that today. Yeah, and to your point about the risks of TSM and China, I mean, I found it ironic that Warren Buffett sold his TSM stock but he kept his Apple stock, that seems like the risks are potentially similar. Okay, we got to wrap but before we do, let's handicap some of the scenarios I put on a LinkedIn poll this week asking ahead of our talk here at Ben, which the following outcomes was most likely. Intel leads the market, however we define leadership, I just let the respondent define leadership and I think we did a pretty good job here of roughly framing it. That's number one. The second one was Intel has to do a joint venture or sell the business. Or three, Intel ends up a distant third. 3/4 of the whopping, my 50 respondents, said the outcome wouldn't be leadership, which one could infer is Jack Welch's number one or number two position.

But Ben, how would you respond to this poll? Would you say leadership as we just defined it, would you say it basically failed, they got to shut it down or do a JV or how would you handicap those three? Alex, bring 'em back up just so we can have Ben review 'em one more time. Market leadership, JV or sell the business or distant third after TSM and Samsung, what would you say? Yeah, so I think again, market leadership as defined by what, right? I don't know if they need to be technologically leaders, they're not going to be revenue leaders. I mean, I don't see a scenario where they're bigger than TSMC, right? TSMC is going to pass 100 million in revenue coming up this year, most likely. Intel's not, right? How long would it take for Intel foundry to be that big? Again, probably never, unless something drastic happens. So in terms of revenue and or size, scale, probably not, a comp with TSMC. The JV side one, you know, again, they've explored this, like I said, they are open to capital coming in and funding foundries with them to some degree. UMC is this as well like they won't call it a JV but they've essentially done that. You could imagine that that could be a plan B scenario. With again, the caveat that the government cannot allow Intel Foundry to fail.

So I don't know how that works or what that means for our country or our taxes or our debt, but they cannot let this fail, right? It's just too important to the national security from a manufacturing standpoint. So somehow, foundry lives in some capacity. I do though, I do believe, right? And this conviction just again comes from a lots of things I've heard in the supply chain about Samsung's challenges. I do think that they will do what Pat said, which has become the second largest foundry by revenue. I don't think they'll pass TSMC but I do think if you look at Intel's foundry revenue roadmap over the next 10 years, that to me is an achievable goal and it's a reasonable goal. If they passed TSMC for some reason, great, I don't see that happening but great. But having the goal of second foundry by revenue, which again could mean $40 billion of foundry, 50-ish, give or take, that seems reasonable between an Intel product and wafer scale customers at the leading edge.

So I think that's where we'll end up. I don't know how long it's going to take but I think that's the most likely scenario. With, again, like you said, worst case scenario, JVs and the government has to do something to make sure that boundary doesn't fail because it's too important. Yeah, so based on that answer, I would put you in the number one camp because I would consider, you know again, Jack Welch, you're number one or number two, that's leadership. So I think that would firmly place your answer in the number one category. Hopefully, you're right. You know, last topic I want to hit on is those manufacturing facilities that we showed earlier, I think I counted six. So that means well north of $100 billion to fund that. So that's why, I mean, really, Sam Altman's number caught my attention. They're getting you money from the U.S. government and the EU but you know, let's call that, what? 20 billion. You know so they got to pay the other whatever, 100+ plus billion that it's going to need. And so I just want to, I want to close there and just come back. I know I've asked this question before but where does that cash come from? You know, that additional cash, it's presumably got to come from their free cash flow, right? Does that play into your scenario? Yeah, I mean, look, they can't fund it all by themselves. I mean, that's the reality, right? That would drastically hit their gross margins in a way that no one wants to see those gross margins decline. We know they're getting chip X money, we know another round of that's going to come. We know that they're getting prepayments from customers, whoever those are. Again, I have no idea the sizes of that. They're going to need that collaborative effort of revenue in what they call, you know Dave Zisner, their CEO, has called this their smart capital strategy. They're going to need to keep using that, right? Government abilities to redistribute cash. They just can't do it on their own. So knowing that, it really comes down to who are those customers that will prepay? 'Cause that's what they're taking. Who are those customers that'll continue to help fund?

Who are those customers that, like UMC, will put capital, co-capital, into some of these environments and then, right? What can they get from government agencies? All of that's going to be necessary until this can really work itself out by itself, which again, I have to imagine that they have a scenario where we do all of this right, we look at the cashflow, we look at the customers at wafer scale, they should make money, right? Intel wants to grow revenue, they want to be extremely profitable. Again, foundry is a key part of that. So they have to have models that work when all of this goes right and those are the things where I think they'll look at those strategic partnerships. And again, government capital 'cause like I said, the government's not going to let them fail, we'll continue to give them a good portion of any money that goes into it. And again, something creative, like you said, like what Sam is doing becomes an interesting option. And you'll see other sovereign funds realize this drastic capacity issue that we have, look for ways to solve it and realize that Intel's a pretty good bet in a good position to help solve some of those things going forward. But strategic capital is still necessary. And the other thought I've had is that, you know the U.S. government, which is attacking the likes of Amazon, not so much Microsoft, but you know, hey, they may be next, and Facebook and certainly Apple, maybe the government gets those companies in a headlock and says, "Look, we'll ease off a little bit "but you got to throw some money "into the chip manufacturing way." Ben, we got to go, thank you so much. You got a great depth of knowledge and really appreciate you coming on the program. Anytime, I'm really glad to do it. It's good seeing you again, Dave. Yeah, good seeing you. Okay, that's it for now. I want to thank Alex Myerson and Ken Schiffman on production and they run our podcast as well. Kristen Martin and Cheryl Knight helped get the word out on social media and our newsletters. And Rob Hoof is our editor in chief over at siliconangle.com. Remember, all these episodes are available as podcasts, just search Breaking Analysis Podcast wherever you listen. I publish each week on the cuberesearch.com and siliconangle.com. You want to get in touch? Email me, dave.vellante@siliconangle.com or DM me at @Dvellante, comment on our LinkedIn post. And definitely check out etr.ai, they got great survey data in the enterprise tech business. This is Dave Vellante for the Cube Research Insights powered by ETR. Thanks for watching everybody, we'll see you next time on "Breaking Analysis". (bright upbeat music)