Intel Corporation (INTC) Earnings Call Transcript & Summary

Okay. We'll go ahead and get started. Tom Lantzsch, Senior VP, GM of Intel's IoT Group, I want to thank everybody for joining our discussion today. I'm Rick Schafer, Oppenheimer's lead semiconductor analyst. And I am joined by Tom, Senior VP and GM of IOTG. Tom has got, I think, 30-plus years' experience in the tech industry. Prior to joining Intel in 2017, he held a number of leadership positions in several Fortune 500 companies. Most recently, I think 10 years or so, right, Tom, at ARM as EVP Strategy. Before we begin, I'm going to just quickly read the standard disclosure, and then we'll get going. Today's presentations contain forward-looking statements. All statements made are not historical facts are subject to a number of risks and uncertainties, and actual results may differ materially. Please refer to each of the participants' companies -- participating company's most recent respective earnings releases, 10-Ks and Qs for more information on the risk factors that could cause actual results to differ from those expressed or implied in the forward-looking statements in each company's presentation. So I thought before we begin, after I read that, I wanted to thank Tom for joining us. And I thought maybe to kick things off, you could give us some insight into your lengthy tech career, as you say, 30-plus years. What led you to Intel? And any other opening remarks that you'd like to make, Tom?

Yes. Great. Thanks, Rick. And thanks, everybody, for joining us. Yes, it's a great question. I had a great run. It's now almost 40 years, I have to say. So the 30 years is approaching close to 40. I think it's 38. So I guess, we have to round up now as I turn 60 this year. But yes. One of the things after we had a great run at ARM, had a great run at all those companies, and actually, when this opportunity presented itself to me, it interested me for the following reason. ARM, a fantastic company, but there's -- with the business model of ARM, you could only do so much to really move the needle. I mean we sold processor cores largely to companies that then made chips that then came into systems. And it was somewhat frustrating if you saw new trends and your ability to actually take advantage of it and innovate because as you didn't do a product, you really couldn't really change -- move the needle and predict things. So at that time, I sort of saw sort of a different world in this IoT space, which was, I think, the conventional wisdom at that point in time was that you're going to have all these small connected things. They're going to get connected up into a gateway and the processing was going to happen in the cloud. And that certainly is still true today for lots of different aspects of this IoT space. But I didn't think that was really where the interesting action was going to be. And I really believe that workloads were going to get drive down closer to where this data was getting created and that a lot of the technologies that were going to be historically driven by cloud was going to come down closer to enable these applications to happen closer to where this data was created. And there was going to be sort of this new category of products that now people are talking about, but 3.5 years ago, no one was talking about, which are these edge devices that were going to get increasingly more intelligent. And so the combination of Intel's assets and the big compute capabilities that we had and the ability -- the ecosystem that Intel had around it, I thought they were they are and we are uniquely positioned to take advantage of that trend. So I don't know that I would have come -- if they would have called me to run the PC business, I don't know that I'd have been that interested. I don't even know if they would have told me to run the cloud business, so I'd been that interested. But I really feel we're in a really interesting disruptive place. So that's how I got here. Before you ask a question, I'm going to put a question or put the question -- one question out on the table because I think our audience will probably want to know, this seems to be the question of the day for Intel, which is the 7-nanometer question. So let's talk a little bit about that. Obviously, Bob, in the earnings announcement talked a little bit about the delays on 7-nanometer. I want to talk about that in the implications of our products because ultimately, that's what customers buy. They don't buy lithography nodes. I'll start by saying this, I mean, we're pretty disappointed in our internal performance on that. So I'm not going to shy away from that. However, I think we've got plans in place, both in the short term, to put out into the market some pretty impressive products for the next couple of years based upon our 10-nanometer nodes and improvements in that. And equally as important, I think our new -- what we're taking advantage of and using our new packaging technology and disaggregation technology enables us to have a lot more flexibility to deal with these kinds of changes. And so you've heard about that a little bit on Ponte Vecchio, which was the first lead [ custom ] product we were going to drive 7-nanometer on. We're actually making some changes on how we're going to deal with that product and actually take advantage of internal and external manufacturing to enable that product. So we were able to react to that. And it is a pushout of our CPU products by probably 6 months, so again, disappointment. For my business, less of a problem, to be honest. We're not -- I mean, although we take advantage of leadership products that come out of the other product groups, we're less dependent on that on our business value. So again, I just wanted to put that out on the table: concerning but not critical. And I think you're still going to see some really compelling products coming out of Intel in the next couple of years that customers and competitors will take note of.

Well, thanks, Tom, for tackling that elephant in the room. Just a quick follow-up to that, I mean, do you feel like your business lends itself to external production should management [indiscernible]?

Yes. So if you look at our business, maybe we can talk a little bit about my business and where we're going, and then I'll answer that question because we are -- I think we may surprise people. So let me tell you a little bit about our business. So yes, based on what I talked about earlier and sort of the strategy, and so we saw this what we called workload consolidation in these applications and driving that closer to where the data is. There's -- to drive that strategy, which we've been on now, we started it in September of 2017, so we're like 3 years into the strategy. And to enable that strategy, there were sort of some other pillars. Part of those pillars was what was going to be the killer application that was going to start this momentum. I mean frankly, when we go address our customers and their end-market problems, and really, what differs my business maybe from my other partners is that we're less about selling a chip, and we're more about selling a business value to end verticals. And we do that with ecosystem partners. But when we looked at what -- and those end customers, and it doesn't matter, cut across any vertical or retail space or factory, could be an airport, could be public institution, whatever. The last thing they want is another box, frankly. It's like they're not asking for another widget because they have to manage them and secure it and all that. So while we're trying to figure out is what's the application that's going to start to drive this convergence of really getting customers to begin to change to think about how they want to re-architect these solutions, and we thought it was video systems, vision systems, computer vision-specific systems. And this business, it had a history of doing -- selling products into that space into the computer vision space. And so that was one part of the strategy. The application driver was computer vision technology. And that cuts across every vertical from public sectors to retail stores to factories to an inspection to you name it, it's going to -- it's happening in a COVID world as people enter buildings. We've got companies now doing things like thermal scans of employees as they walk in the door and making sure they have a mask on. These are all things that are driven by camera-based systems. So that was the killer app. To enable that killer app, we needed to modify our product strategy to do that. And that included both IA products as well as non-IA products. And by that, we had inference technology. And it's really about inference at the edge to go do that. And so Intel had acquired, prior to my arrival, a company called Movidius that really focused on low-power, high-performance inference technology, and we are taking that and leveraging that into that space. And so all of those products we build outside, all of them today. And we'll be announcing a couple of these. I mean we've got 2 generations of it already in silicon in customers' hands. We just aren't publicly announcing them right now, but those are coming out the door. Interestingly enough, they're also ARM-based. So we're not even wedded into a situation where it's got to be IA-based. So we're focused on solving customer problems. So the answer to your question, yes, and we're already engaged on that. The other sub pieces of that strategy was creating a developer community to enable this AI inference technology at the edge of the network are these computer vision technologies. And we put a very aggressive plan in place to develop a brand-new capability. It's called OpenVINO, is the development tool that we announced 2 years ago in June. And really, the problem that, that was solving, and we continue to expand on that, was that when you look at AI inferencing at the edge of networks where data is created, that most of the models and most of the activity is done by the -- in the cloud, which is great because you can have a ton of data and you can have infinite compute. So these models are created there. They're trained there. They're created there. But when you go to deploy them, unfortunately, you don't have infinite compute, and you don't have infinite power, and you don't have infinite capabilities. And so the problem we're solving for those customers is taking models that were developed in this world of almost infinite supply capability and being able to put it into a form factor that they could deploy at the edge of the network. And really, that's what OpenVINO does. It's network-agnostic and really focuses on that problem. And it's hardware-agnostic. It'll take a model and help our customers to find is it best to run it out of CPU, a GPU, an FPGA or Movidius neural network acceleration technology. You write it once, it will obfuscate it and drive it down to the right hardware profile and help you determine if it's performance you're after or power you're after or price you're after, it helps you go through that myriad of choice. So that's really where that developer community was. And then the last part of the strategy was, which is also very unique about this business, our business is -- whereas my colleague's business in the computer like DCG and CCG, the laptop guys or the cloud guys. 10 customers will represent 80% of their business. In my world, 10 customers represent 30% of my business. To get to 80% of my business is closer to 100 customers. So we are very fragmented in our customer base. And our verticals are very fragmented in their ecosystems. And so although we drive this horizontal capability, which is applicable to all of these verticals, how we go to market is very vertically centered with our ecosystem partners. And so we wrap around solutions with these ecosystem partners to enable that to go to market. And we've got a program called market-ready solutions that we develop with these ecosystem partners. These are tried and true capabilities based upon business value with our partners. And we have like 450 active solutions that we've developed with these partners cutting across all these verticals. And year-to-date -- or since we developed this program over the last 2 years, I think we had 15,000 deployments. So it gets a lot of scale. It's a lot of scale. And we do so in these various vertically centered. So we're very different in our -- how we go to market, how we position our value proposition than any of the other businesses in Intel. So a little background. The other thing I'll say for the investor community is, which is also maybe a surprise in this IoT space, the ASPs on my products are over $100. Again, not what you would typically think about in this business, but very Intel-like, I guess, I would say.

Yes.

I rambled a lot.

No, that was great. There's a lot to unpack there. You certainly hit some points that I'd love to flesh out a little bit more. Okay.

Sure.

One high-level question is just you listed a lot of potential end markets or focused end markets for it. If you were ranking your top -- I mean, for your top 10 -- I'm sure you did rank top 10. The top 3 or top 4 kind of really verticals that you're kind of seeing as the lowest-hanging fruit maybe or the best opportunities?

You got to put these into dimension on time, right? And you have to put these now in the dimension in a COVID world versus non-COVID world, right? On the phone with you guys, this -- at this time given my quarter last quarter, I mean, we were on a roll for a long time, and all of a sudden, COVID really hit us last quarter. So I'm going to answer, look at it strategically. I'll answer the question versus tactically. Tactically, there'll be a bunch of stuff going on in this COVID world. I mean we're -- there -- the design win activity short term to get people back into the office and keep people safe and using this technology is on a -- is really hot right now, I'd just -- I would say. I gave you examples of everything from thermal imaging of people to mask wearing to whatever keeping people safe. I think in the short term, we're going to get a blip. That wasn't really in our strategy 6 months ago, but it's happening. And there's nothing really we have to do differently. It's just a different deployment. Long term, we think that there's a long -- the longer term -- the industrial sector is ripe. I mean industrial sector, utility sectors, these are ripe for major change for lots of reasons. Manufacturers are now realized and especially driven by this COVID thing that the ability to manage these facilities and have flexibility on how they do this is going to be increasingly more and more important, their ability to get access to things when they can't travel is going to get increasingly more important. The energy grid has gone through a massive transformation over time. It's -- we were used to a world where we had a fixed thing called a generation station that created x amount of power, and you're pretty much managing. The demand was what the demand was. Now in a world of renewables where supply is highly variable, very variable; wind, very variable; light; sun very variable. So the equation on how the grid will be managed is going to -- is fundamentally changing in massive ways. It's not set up to do that. And so that -- this technology will be directly applicable to the AI everywhere, everywhere. And it doesn't matter what the vertical is. It doesn't matter what the application is. AI-based systems are -- and driving towards autonomous type of approaches on things. Pervasive is -- and again, it cuts across a lot of different verticals. I can't -- we're in -- we've done a program talking about workload consolidation. We've done a program with a major fast food restaurant. I'll just define it as a fast food restaurant. They probably don't like to be called that, but I'm going to call it that because I don't know what else to call it right now. But they had 34 different applications inside that restaurant. That's each of their stores, everything from their cash registers to the digital signs to running inventory systems to camera systems to everything. We've been able to take what was 34 different unique applications and basically consolidate that into a workload and run it on 2 Xeon servers, including everything like cameras that will test the quality of the food, I mean, that will determine how long the food has been sitting in a basket like a French fry and determine using AI technology if it's appropriate for their customers or not, or how many cars pulled into the parking lot, or doing algorithms like how many cars are driving down the street, how many cars are pulling in, what's the count, all stuff using camera technology. So it's pretty pervasive, and all of these are booming in many ways.

And Tom, you hinted at it. I'll just try to get that to flesh out a little more. But on COVID, I think you guys have grown consistently in your group for 5 years running. Obviously, you mentioned last quarter a COVID-inspired stumble. But like getting back on the horse, getting back to growth, can you talk about some -- maybe just a little bit more of the impact from COVID, just...

Yes. Look, I think I'd categorize it this way. We -- again, we're selling products largely going into businesses trying to -- it's a capital play in many cases for these businesses. I think in the short term, I guess, this is a blessing of being part of Intel in the short term that those businesses were pressed with buying capital for their employees to work from home or do things like we're doing right now, driving this remote access and video system. So their capital dollars were getting deployed to those kinds of activities. It impacted my business, for sure. I mean there's no question about it. That's going to change. I mean this is -- when is it going to change? Your guys' crystal balls are as good as mine. Obviously, we're watching it closely like everybody. But that's going to change. I can say that our design win targets, which is really what we try to track to predict our future, basically, you can anticipate what my goal was going into this year based upon our historical performance, and I didn't see it changing prior to COVID. [ Argonne ] design win goals this year that we're working with our sales team, we're ahead of plan for this year. So activity is robust with regards to customers and companies taking a look at how they have to advance their technology. So long term, economics aside, GDP aside, capital budgets aside, I'm pretty convinced this thing, when it snaps back, it will be back on the track. Inventory levels in general in the industry in our space are pretty tight. I mean customers started taking pretty aggressive action pretty early. So it's not as though the channels have a bunch of inventory. So when this happens, it could happen pretty quickly. But again, you guys follow this stuff as much as I do, and I'll let your crystal ball determine that.

So I'm curious because 5G is starting to happen. IoT just seemed to be sort of peas and carrots, seem to go hand in hand. So I mean, could 5G be part of that inflection or part of that bounce back that you discussed, as a service provider start to turn up service and we start to see IoT adoption?

Yes. It'll help. Look, everything that gets connect -- I look at it this way. And again, one of the things we do with our ecosystem, given the fact I'm not really in the small things kind of business and microcontroller kind of stuff, so we try to collaborate with everybody in that space that we can because the more things we get connected, the more things that I think are going to need to have compute at this first hop, so I'm motivated by that. 5G offers some pretty -- the 2 things 5G offers is interesting in my -- most interesting in my space is low latency, faster -- lower latency and its guaranteed quality of service. Because the systems that we're talking about that I'm working on mostly, I wouldn't say they're mission-critical for these companies, but they're pretty important. If you're running a factory, it can't go down. I mean there are aspects of this. So it's less about breadth of -- number of connected things in my world as it is the quality of the connection and the amount of data and the stuff that you can do with it quickly, the latency of it. But that being said, we're doing a lot of work, and we have a joint -- we actually have joint work between us and my colleague, Dan Rodriguez, who runs the networking -- NPG Group, Networking Processing Group. He really is the 5G sort of infrastructure guy. And we're a -- we have joint programs where we're taking edge inference technology, my technology and 5G technology and putting it in a box effectively. And showing customers how they can take advantage of bringing these 2 capabilities together, again in a way that is manageable and at the right cost point. So it will be a catalyst. It's a little early to see exactly how that catalyst is going to play out, but certainly, it's going to help.

And I just got to ask, I mean, in what way, if any, as your guys' decision to exit kind of back off from 5G modem? Has that...

Yes. No. None. Very little. We're partnering with modem providers to go do that. I can almost argue, in some cases, it helps us. I mean, again, if your goal is connect as many things as possible and we're one of those suppliers, then everybody else looks at you with a little bit of questioning when you go talk to them and being transparent. Now to me, I don't care. Anybody's modem is a good modem, right? Connect. Connect. More data sooner, I'm happy, right? Secure it, drive standardization across the security on that thing, great. So as my business is involved, it actually probably helped.

Okay. Okay. It makes sense. And you've touched on it a couple of times, so I'll just ask a question on AI. Does inference at the edge mean aggregation at the edge? Because you've mentioned latency, and there's always going to be a certain degree of that out there. I know 5G helps that along. I'm just curious how -- basically how we do that?

Yes. Well, look, I just saw a report. And we've had this belief, but now other people are talking about, they're talking about today, most of the inference in AI systems is done in the cloud. But I just saw there was -- I think ABI Research did a report that in 2025, inference at the edge will be bigger than inference in the cloud. And we believe that. We're not analysts and chase those numbers down, but we do believe that. And I'll give you an example why. I'll just try to give you some like real examples. Let's take Mobileye, which isn't in my product category, but I think it's a pretty good example. For every kilometer that is -- a car drives in a Mobileye system, I think there's an EyeQ3 system maybe, it collects over 4 gigabytes, 4 gigabytes worth of video content is coming into that camera. We actually inference that to take action immediately in the car. It can't do it fast enough. You couldn't send that to the cloud and control an accident, so first of all, impractical. But of that information, only 10 kilobytes, 10 kilobytes do we actually keep and send to the cloud. 4 gigabytes, 10 kilobytes. It's not a hell of a lot of difference than when I put a high-speed camera on a manufacturing line to do quality inspection on a product. The amount of content -- I think the answer is a factory. One HD camera will produce like 2,000 terabytes a day. Now you're not going to send all that to the cloud. You don't need to send all that to the cloud, right? So -- and now you talk about you're going to do an automated store like a frictionless store. I mean there's just no -- you couldn't afford it. Even if it was fast enough, you couldn't afford to send all that data. So it's just going to drive all this action down to the edge from economics, from maybe response time, it's just impractical. And so that's what's going to be the key catalyst that drives us. And so we're looking at a strategy and to take advantage of this, that I actually do some camera chips. So we're working on this tiered strategy that this application theoretically could start at the camera, could go to the edge or could go in a network device in Dan Rodriguez's business and end up in a cabinet at a telco operator and be able to move the application to where it's best suited, that's most economical or is the most safe place. The other aspect I'd say is that if you believe a world that AI is prevalent, which I think most of us probably do, and you believe a world that a lot of that AI activity is going to happen closer to the data for the reasons that I said, which we certainly do, securing those algorithms is going to be really important. Because that is going to be -- everybody thinks about the data being valuable. Actually, for our customers or partners, it's going to be the algorithms that are valuable. That's going to be the intellectual property of my end customers is what they do with this data to improve their algorithms because that's going to drive their competitive advantage. So again, the other area we got a lot of emphasis is how do you secure these things because this is going to be a big thing. And again, unlike the cloud where, frankly, interesting cyber-attacks really happen when you get physical access to things, I mean, like touch it. And as we know, if you want to try to get into a cloud service data provider's data centers, you're probably going to go through massive amounts of security and all sorts of physical security, like physical badge readers, retinal scans, you name it, guards. These edge devices are sort of hanging out in space. I mean they're in -- they could be in a cabinet. They could be on a pole. They could be in a factory floor. They're more exposed. So that's another big activity we've got is how do we secure this stuff because this is going to be really important in the future as the industry matures.

That makes sense. And I guess, if I just pivot slightly to products, and I know I want to get -- talk about IoT chips kind of categorized into the 3 main buckets, right: processors, sensors and connectivity chips. I mean can you spend a second talking about the importance of each? And maybe you're not as big on the connectivity side, but I -- but correct me. But I just -- if you could walk us through some of the...

Yes. Processor, so -- processing and processing technology, just like we're seeing in lots of other places, is becoming more heterogeneous. So it's not just going to be CPU-centric. It's -- the world to run these workloads is going to be increasingly more heterogeneous if you need to fit it into a form factor. There's probably no better example of that than the phones we carry every day but apply that to my world as direct correlation. Connectivity is -- it drives us. I mean the more connected things, obviously, the better for us. So we try to do things anything we can to enable that. Pushing standards as much as we can, obviously. But we're not really biased to one versus the other. We have no bias one way or the other. We partner with people across the WiFi space as well as the modem space to whatever to enable that and help our customers do it. And again, it's more in the context of providing it in a secure way and help them drive that security activity. And sensors, yes, we really don't touch much of that at all, other than unique sensors drive unique models for AI. And equally, over time, what we will see, if you want to get to the really -- the more complex merger of multiple sensor technology to do AI models, the richer it will be. I mean it's just like us as people. It's not just our eyes that do stuff. It's our eyes and our ears together is more valuable than just our eyes or our ears. So you're going to see this merger. And so we'll follow sensor technology from that perspective, but it's not that we do much in that space.

And if you look at your processor tech, obviously, you've got a deep portfolio there. Can you give any examples or just discuss in general kind of what advantages that's brought to Intel? I mean you've got Atom, you've got Core, Xeon. You know what I mean? Like you're bringing all that together. I don't know how much each of those play or where -- or you know what I mean, if you could just flesh that out.

Yes. So our goal -- again, ultimate -- I don't know if it's our ultimate goal, but here's sort of the advantage that we have and that we take advantage of. If you believe in this workload consolidation story, which is really what we're talking about, which, again, in many ways, mimics what's gone on in the cloud where you've been able to virtualize workloads and use containerization and orchestration to move stuff around, that stuff's coming down, which ultimately will fit into probably a Xeon-like kind of product. That's probably at the end of the rainbow, that's where it gets to. But to get customers on that journey, being able to start with a function that's just on an Atom product where they'll do 1 or 2 of these and then migrate them up to Core and then move them up a row to Xeon without changing their software investment is a huge advantage. It's also a big advantage that a lot of the markets that we service today still have a big legacy Microsoft dependency, honestly. It doesn't matter if it's a factory, a retail store. Microsoft has been a big player in these, and they have historical applications based upon that, so those are difficult to move off of. I mean they're just -- first of all, customers don't want to move off of them because the developer may not even be around anymore, in many cases, and yet, it's an important portion of their business. And so being able to consolidate those kinds of workloads is advantageous to us and offering across that product portfolio. For just the numbers, if you look at my business last year, let's just rough it out, about 20% of it's Xeon. 60% of it's based on Core. 20% of it's based on Atom. So the sweet spot for my business is really Core-based CPUs.

That's really helpful. I didn't know that. Maybe a quick question on competition, and then we can slip to some of the Q&A questions that have come in. And really, on competition, I think you're kind of uniquely -- in a unique position to answer this question. I'm just curious where you see the most competition and how you view ARM. And that's the part I'm curious, especially from you [ with ] that DNA.

Let's go to the last one. And again -- so in AI-based work, I think we -- in video and computer vision technology work, you can't -- NVIDIA is obviously a strong player, and we can't ignore them, and clearly have a big ecosystem of developers. And our work on OpenVINO is not quite what they do, but is -- I mean is chasing that developer community in many ways. And so we watch that. NVIDIA Jensen is -- I look at them at a product development standpoint. Their model is very impressive. I've known them for a long, long time, even during my ARM days. So I clearly see them as sort of setting the bar on that. And so we watch them closely. In parallel to that, which may surprise people historically, which is -- we've sort of been impacted by not only the COVID thing, but we've also as a business been impacted by the government entity list in China and specifically with some of the computer vision players, mainly Hikvision and Dahua, who have been big historical customers of mine. I watched HiSilicon a lot, the Huawei semiconductor ARM, not talked a lot about in the West, very capable development team, very capable. And I benchmarked our product performance against them. And we really went head to head with them when we were at those companies. Okay, the entity list has sort of changed things, but they're a force. I mean they're a real deal. AMD in the day in and day out, less so, less so. Again, I know Lisa from a long time ago. I've known her for a long time. I mean she's focused on -- very focused where she's focused right now. If I was her, I'd do the same thing. My space is obviously a lot more fragmented, a lot more small, so I'm clearly not probably in the place that she's going to fire the first shots. But if it's a CPU for a CPU, obviously, and that's all our customer is going to buy, is a jump ball situation. But those are the ones, the one, NVIDIA, the most, I think. ARM, I use ARM. Again, so I don't really -- I partner with ARM on security stuff. We were collaborating with them on security stuff to connect things. So ARM is an architecture that -- I don't really see it that way so much. The ARM partners that take advantage of it. Both HiSilicon and NVIDIA are big partners of ARM. So from that perspective, I'd call them out much more than ARM as an architecture.

And the potential sale of ARM, though? If it ended up...

Yes. I scratch my -- I mean, I was involved in selling the company to Masa. And I ran that company for a long -- not ran it, but I was -- Simon and I like sat next to each other for years. And trying to always figure out how we could ever get a premium through an acquisition like who would ever want to buy us and create this number? I can tell you, SoftBank was never on the list. And not a lot of people could be, frankly. The ARM business model was so based upon trust and bareness and neutrality. And it's very difficult for me to understand really who could take advantage of that and create value with that as an asset if you didn't have that framework. We were dealing with customers on very sensitive things. You've got to remember, so if you're developing an architecture today, it takes 2 years to do it. Then you give it to a chip person. It takes 2 years to take it and make it into a chip. So we were dealing with customers on stuff like 4 and 5 years out in the future. If the wrong person gets a hold of that and those customers feel threatened by sharing that information, I just don't know how that works very well. I think that's going to be a struggle. I'm perplexed on how any one company -- I mean, there's a few, but any one semiconductor company, for sure, could take a position into that. I find it -- I scratch my head a little bit. I know that there's been lots of rumors in the press on people. So I don't want to...

Yes. I think a lot of people are scratching their head on that very deal.

We could not figure out how to do that. So somebody's smarter than I am. Obviously, I left, so I'm gone. So maybe the next smart guy figured it out.

Okay. And I think we've got time for maybe one question. I'll read some from Q&A. And someone is asking, if you could explain the use of chiplets as a foundation of future circuit design, if you've got an opinion on that.

Yes, I think it's helpful. Again, when you really look at -- I'll just take my business, for instance, as a great example. When you really look at what I take advantage of from my colleagues, it could be a server chip or a client chip. The core CPU complex, call it the CPU, maybe the GPU complex, the memory complex on that, that clearly scales really, really well with photolithography. I mean that's where you get the big bang for the buck. Where you don't get the big bang for the buck is all the I/Os. Honestly, that is not a big bang for the buck and because analog just doesn't scale as well. And so -- and in my world, here would be a great example. We've got a product coming out this year that will demonstrate this. We're taking the base CPU die from, let's call it, a client device. And what we change is the I/O structure. Because my customers -- I'm talking to you on a MacBook Pro right now. It's got 4 USB-C I/Os on it. When I take -- my industrial customers do not want USB-C I/Os. That's not what they want. So the ability to mix and match certain things easily at a low cost and take advantage of lithography on certain things and take advantage of the mother ship, highly advantageous. And so historically, we've done a pretty good job of being -- and if you can put it all on the same chip, it's theoretically better, but a lot of times, that has time implications -- associated with time-to-market implications on it. So I'm not trying to ignore it. But there's -- if you can get the packaging technology right to enable it to be able to do this mix and match at the right cost, it's very interesting. It allows a lot of flexibility. It allows a lot of innovation that you were constrained by.

Got it. And I think that's a perfect comment. I'm watching the clock. So I'm actually taking you a couple of minutes over, Tom. So...

That's all right. I hope this was helpful to everybody. And thanks, obviously, for your time. I appreciate it.

Well, likewise. It was great seeing everybody. Thanks, Tom Lantzsch, and thanks Tony Balow over there for joining us today from Intel. And that's our call. So thank you very much.

We'll see you. Bye-bye.

Bye.