Applied Materials, Inc. (AMAT) Earnings Call Transcript & Summary

Good morning, and good afternoon, everyone, joining us on the East Coast. I'm Mike Sullivan, Head of Investor Relations at Applied Materials. I am pleased to welcome you to our eBeam Technology Briefing and New Product Introduction. Joining me is Keith Wells, who is Group Vice President and General Manager of our IPC business unit, which includes PDC, our Process Diagnostics and Control group, along with POC, our Process Optimization and Control group. Prior to joining Applied, Keith spent 27 years at KLA, where he led the development of advanced optical and eBeam wafer inspection tools. And he later served as General Manager of the Company's eBeam and Optical Inspection divisions. He also spent a couple of years as a Corporate VP at Lam Research, focusing on analytical techniques and optical sensors. A quick housekeeping note. At the end of today's presentation, you'll be able to find all of Keith's slides and animations on the Events page of our IR website at appliedmaterials.com. Now, Keith, let me hand the webcast over to you.

Thank you, Mike, for the great introduction. So I think what we're going to do, go right into our agenda for today. We want to start with a quick technology primer about optical and eBeam technologies to set the stage for some of the other things we'll discuss in today's presentation. Then, we'll launch into eBeam market segments and the applications within those segments. And now, today, for the first time, we're going to introduce the concept of cold field emission technology and how it's enabling Applied to better serve the eBeam market and those eBeam applications. We'll talk about some new eBeam product introductions today. And then finally, we'll turn it back over to Mike for Q&A. So let's start with kind of a technology primer about Optical versus eBeam inspection. A lot of what we're trying to do here is to produce sufficient resolution in order to find defects on wafers. The way that works for optical inspection is to try to use the shortest wavelength possible at the highest numeric aperture, which is really that angle you're seeing on the slide there. The larger the angle that is, the more resolution you can get out of the optical system. So we shine the light down on a wafer surface and we'll either collect reflected light, which is usually termed bright field inspection, or we'll collect scattered light, which is dark field inspection. And this works very well for surface defect detection. But as I progress in the presentation, what we will point out is where it doesn't quite work for some of the very buried defects or the high aspect ratio inspections. In eBeam, what we do with eBeam is we take a source, ideally a very bright source of electrons. We send it through, again, optics, but they're electron optics. And we focus the beam down onto the wafer surface. And the interaction with the surface creates 2 types of electrons. One, are secondary electrons, which gives us information about the surface of the wafer. And that's how SEMs have traditionally worked for many, many years, is only collecting secondary electrons. But today what we're seeing is there's a much more valuable application if we can take high energy electrons and inject them down into the surface of the semiconductor structure and collect what's called backscattered electrons. That gives us information about defectivity and metrology deep into the stack of the 3-dimensional devices that's now being created in the semiconductor industry. And this is creating an inflection we believe and a revolution in the use of eBeam technology. So let's go on and talk a little bit about resolution and the applications between optical and eBeam. So if you're looking for surface defects, optical is a pretty good way to go as long as you're looking for surface defects greater than about 30-nanometers. It has lower resolution, and the trade off with that lower resolution is higher throughputs and better coverage. But when you want to find something that's below 30-nanometers in size or you want to find something buried within a structure, it's becoming more and more predominant that you have to use eBeam. So eBeam we believe in the future will be adopted for sub- 20-nanometer defects. And as you'll see in today's presentation, especially for the buried defects, that there's no other way to detect. Ideally, customers would always pick the best cost of ownership opportunity. But when the best cost of ownership opportunity isn't there, they turn to the best technology to serve that application, and we believe that's eBeam in the future. Now let's talk a little bit about resolution. And we've given an example here, where we've taken a standard definition image and we've compared it to an ultra-high definition 4k image. This is analogous between optical and eBeam. Yes, you can clearly tell what the image is in the standard definition. But as you go to the high definition image, there's just much more detail, there's much more information. And that gives you actionable insight. The same thing is true for optical resolution. Optical today, it may be able to find a difference, it may be able to see that there's something there that maybe shouldn't be there, but it fundamentally can't tell you what it is. And so if you apply the resolution of eBeam, not only can you detect it, but you can usually tell exactly what it is. And in this example, you can now see that it's a protrusion, a pattern that shouldn't be there. So in a single inspection, you can both identify and classify what the defect is. Where in an optical inspection, you have to take 2 steps. You can find it optically, but then you have to go image it with eBeam in order to figure out what you found. So there is incredible value to our customers in inherent resolution and being able to debug their problems faster. And here's an example again for EUV patterning. EUV patterning is pushing the limits of what we can do today. And therefore, our customers are demanding the highest resolution solutions. So with optical, you may be able to find defects at these pictures, but probably limited to 30-nanometer defects, which are barely at the design rules that people are actually printing today with EUV. But with eBeam technology, you can find defects clearly below 20-nanometers, down to 10-nanometers and at the same time, you can actually resolve pattern. So not only could you find defects, you can also use metrology with eBeam and you can actually make measurements about those patterns, about those structures. So again, you're almost getting 2 for the price of 1. You're getting both defectivity capability as well as metrology capability in a single inspection. And we believe customers need this in order to rapidly debug their EUV problems. Now let's go back to something I talked about a little bit earlier in the presentation, about some of the limitations of optical. Optical has 2 fundamental limitations to it to find defects that are buried deep in the stack. One is this interaction angle, this big angle -- if you're taking this big angle and trying to place it down into the stack in order to find defects, the problem is, is the defect is at the bottom of these trenches or in the gate-all-around stack. And what that large angle does is it interacts with a lot of volume of the wafer that you don't really care about. You don't really want to interact with that volume of the wafer because that creates scatter, it creates other noise sources and it makes it difficult to find small defects in the presence of those noise sources. So what we're typically seeing as we go to deeper and deeper type applications, DRAM and 3D NAND, that you're really limited to 50-nanometer defects through optical techniques if you're looking deep into the pattern. That's really not the case with eBeam technology, because we don't have this concept of numeric aperture, NA, or this large angle. We have a pencil-like beam and we can direct that pencil-like beam directly down into these high aspect ratio of structures and image what's at the bottom of those structures without creating noise by interacting with the rest of the device. And so that gives us, I think, an inherent advantage in being able to find those defects in DRAM and 3D NAND structures. Now the other thing that's become apparent that we need within this type of technology is, as we put those electrons down into those very deep trenches and we generate secondary electrons, which are the surface secondary -- surface sensitive electrons, we can't get those back out. So what we need to do is we need to be able to put very high energy electrons down into these high aspect ratio structures, we have elastic interactions and those high energy electrons then travel back up the structures. And you can see the difference here in the 2 imaging techniques. If we're only using secondary electrons, we're only going to see information about the very top of the device. So this is useful if you want to measure, say, critical dimensions of the contacts in that image. But if you really want to understand what's going down, happening at the very bottom of these high aspect ratio structures, that's where the backscattered electrons are necessary. So by adding backscattered electron capabilities into our tools, we're not only being able to see the top of the device, but we're able to see very deep into device and give our customers information that they couldn't get without the backscattered imaging capability. And if there's defects down there, they're certainly not going to see them very effectively with optical imaging techniques. Now let's turn a little bit to our markets and why we're pretty excited about our opportunities in our markets. So in '21, the total WFE was in the mid $80 billion range and process control comprised about 13%. It's been about 13% to 15% for a very long time, but it's also a very healthy market to be in. So out of that 13%, our customers spent approximately $11 billion on process control. And then, out of that, approximately 20% was eBeam spent. And we break that down into 3 different types of markets within eBeam. One is the eBeam inspection market; 2, is the defect review market; and 3 is the CD-SEM market. They are all roughly about 30% of the market. When we look at the growth across the industries in these various markets, so between '19 and '21, WFE grew 68%, process control slightly outgrew the market at 72%, optical was about even with WFE. But what we saw in eBeam is growth of 95% across that same 3-year period, primarily driven by eBeam inspection and defect review. So we believe that today we're seeing signs where eBeam has outgrown the WFE and we expect there to be similar trends going forward as the applications demand more and more eBeam technology to properly solve our customers' problems. Now let's go into a little bit more detail about these markets. So I think the first market, which is defect review SEM market is pretty easy to understand. Customers use an optical inspector. They inspect the wafer. They may find millions of candidate defects, but they can't really image them and therefore they can't really classify them and tell whether they're important or not. And so what a defect review SEM will do is it will revisit sites to be able to see and distinguish critical defects from nuisance defects. And this allows our customers to concentrate on the root cause analysis of the defectivity and debug their line. They use the defect review SEM both in R&D and HVM routinely. The second area is eBeam inspection. Now eBeam inspection is reported out as one market. It's actually really 2 markets today. One is what we would call traditional eBeam inspection, which is looking for small defects, the under 20-nanometer defects and those buried defects. There's also an element of looking for electrical defects, where you're actually looking for shorts or opens, and you can also find those through eBeam inspection. The other part of the market is really a metrology element. As I said earlier, because eBeam has the resolution not only to find the defectivity, it has the resolution to resolve the patterns, you can do both at once. And so you can either do top level critical dimension measurements. Or what's more interesting and actually emerging is customers can look through multiple stacks. They can actually do 3D metrology and they can figure out what is the overlay between one patterning level relative to the other. They can measure the CDs. They can look at chips. They can do a lot of characterization. And they can get actual insights at relatively high speeds and do full wafer coverage. And so we see this eBeam metrology on wafer, on device as an emerging market that will also drive this eBeam inspection segment. And then, finally, there's traditional critical dimension SEM. And critical dimension SEM is sensitive to measuring the critical dimensions of the top surface, the surface that was just patterned. And that information is used to feedback to the litho cells and the steppers to do calibrations to understand if the stepper is behaving properly, if the stepper is operating within spec. This is also used in high-volume manufacturing. So we believe that these are very robust markets and we expect them to continue to grow as the industry continues to shrink. And so how has Applied done in this market over the last few years? So here's Applied compared to some of its competitors in the marketplace. So in '21, that was our first year that we exceeded $1 billion in revenue for Applied. We estimate -- or I should say the outside companies that track this industry estimate that we were approximately at 50% market share and that we've grown 13 points of market share over this period of time. And if you look at our revenue versus our competitors, we're 80% more than the nearest competitor. So our technology in this marketplace is proving to be very effective and proving to drive growth for us in eBeam. So let's turn to why that is. What is our technology and why is it offering us growth that's outpacing our competitors? So I want to talk a little bit about what cold field emission is. So traditionally, the way you've created very bright sources for electron inspection is to heat up tungsten filaments. That was many years ago. And then we got a little bit more sophisticated in how the tungsten tips were made and we applied very high temperatures to them and a technology was created back in the '90s called thermal field emission. And this is how almost every traditional SEM works today that's in a high-volume manufacturing environment in a semiconductor fab. Starting over 10 years ago, we wanted to see if you could take cold field emission and make it applicable to high-volume manufacturing. That was no easy feat. And we'll have a video later in the presentation that talks about how we did that. But why would you want to do that? Why do you want to take cold field emission out of a lab and put it into a high-volume manufacturing environment? Well, the basic ability to operate at these lower temperatures and to have a very precise crystalline structure of the tip means 2 things. One, you can get resolutions below 1-nanometer, which is significantly better than what you can achieve with thermal field emission. And 2, the source of brightness. The amount of electrons that you can get out of these sources can be anywhere from 3 to 20x greater than what you can get out of thermal field emission. And the importance about that is the amount of electrons you can get out of a source is directly proportional to how fast you can do inspections or metrology applications. So we're talking of the ability to do those types of applications 3 to 20x faster than what our competitor could do if they only have thermal field emission. So it's incredibly important from our perspective starting over 10 years ago that we -- if we wanted to dominate this market space, we would do so by bringing cold field emission to it by volume manufacturer. And so -- I love this slide because seeing is believing. So what we're seeing here is the resolution that you can get out of thermal field emission versus the resolution that you can get out of CFE, and it's dramatically different. So we can either get this higher resolution, or as you'll see later, if we want to operate at the same type of throughputs, we could -- if we want to operate the same resolution, excuse me, as thermal field emission, we can get much faster throughput. And the key to doing all this was to make it production worthy. And we will talk again about that a little bit later today as we talk about the new products that we put out into the marketplace. So I want to show you a little bit of a video that probably gives maybe a better explanation than I just gave you between -- the difference between TFE and CFE sources. [Presentation]

So I hope that gave you a great explanation of the differences between TFE and CFE. Here's another way to think about it. Because we have very, very high brightness sources in CFEs, we have a wider operating range, allowing us to go inherently faster than the TFE technology. And because we have the emission of the electrons from a very specific part of the tip, we also have the higher resolution. And so this flexibility gives our customers a lot of agility about how to use these tools once they have CFE sources. If the customer wanted the same resolution as a TFE, they, on average, could run between 5 and 10x faster. If they wanted the same throughput as a TFE tool, they could probably run at 1.5x greater resolution. So this gives our customers a lot of capability to go in and debug the very difficult problems that they're trying to debug today, such as the gate-all-around transistor. And in many cases, with this type of technology, avoid going to TEMs, which is a different imaging technique, which can take hours to turn around and give them information. So we think that this technology has very broad applications and is, of course, superior to the TFE technology. How do we see it being used in the marketplace? It certainly can be used very effectively to go back to older generations, such as planar CMOS or FinFET. But where we really see this being used today is in the gate-all-around area, especially for inspection capabilities, because we believe this is the only technology that really allows our customers to see very small buried defects in the gate-all-around structures. So let's go forward and take a look at some of these applications in a little bit more depth. So here, we have 3 different types of applications that is driving eBeam adoption. So one is, in the gate-all-around, one of the bigger problems is that there silicon germanium remains, residues remain, but those residues are buried in the structures. And if they have poly-gates on top of them, sometimes it's very difficult for the optical inspector to penetrate down and see those. When they can, they can usually see defects, say, in the 30-nanometer range with optical inspection. And customers do use optical inspection for the gate-all-around application. But once you get to the very smallest residues, then the only way you're really going to see them and be able to debug them is with eBeam technology. And since these defects are at the very heart of the transistor, it's incredibly important to drive to a 0 defect level. And so you're going to need the resolution and capabilities of eBeam to be able to find these type of detects in the buried stack. Another application that's driving eBeam adoption is simply the resolution of EUV. So what I mean by that is as our patterns are getting smaller and smaller and customers need to measure those patterns with very high accuracy or they need to inspect those patterns with very high accuracy, they can really only turn to eBeam in the future, because the defectivity here that starts to matter is in that 10-nanometer range and the only thing that really gives you the capture that's necessary to know that you have 10-nanometer defects on your wafer is eBeam technology. And then, finally, we've talked a little bit about high aspect ratio. So when we go to DRAM structures, we're talking about usually several hundred nanometers. And this will only get more difficult as we go to 3D DRAM in the future. And if you've got a very small bridge at the bottom of those structures, it's extremely difficult to get light down in there and have interactions that allow you to see things in the 20-nanometer range. The only thing that's really going to do that is the application of eBeam technology. So now I'd like to show you a little bit of video about how we ended up making CFE technology HVM capable. [Presentation]

Thank you, Ralph. So today we want to talk about a new product introduction, which we call the SEMVision G10. We've been a leader in defect review SEMs for many, many years, having 75% or greater market share in this space. And so naturally, it made sense to take the CFE technology out of the lab and deploy it to our SEMVision product line. And what we've been able to do with this is demonstrate sub-nanometer resolution for logic and memory. We've done that by finding the -- demonstrating to our customers that with superior resolution, at least 1.5x previous generation tools, we can find the smallest defects. And with the addition of backscattered electron collection, which we talked about a little bit earlier, we can image down into deep trenches, finding defects in those areas that we couldn't possibly before. And we're doing that with basically also higher resolution on EUV structures at lower dose so that we don't damage it. And all this is done 3x faster. So the cost of ownership of this product is impressive versus both our previous generation tools and also our competition. And so we believe today without this type of technologies, our customers wouldn't be benefiting from the ability to quickly debug some of their FinFET and memory processes. But even more importantly, they're using this to debug their very, very advanced nodes for the gate-all-around introduction and advanced logic. Now I want to talk a little bit about adoption. And here, we introduced the product in FY '20. And what we're seeing is amazing adoption. And I think this says 2 things about the product. One is, of course, that our customers are voting for this product. But two, the product is stable and high-volume manufacturing capable. I don't think you see this type of growth on a product line over this short period of time unless you can demonstrate to our customers that the risk in adopting it is quite low. And so the CFE technology I think we have demonstrated is quite stable, quite reliable and quite capable for high-volume manufacturing applications. And so we've now seen today since the product launch greater than $400 million in total cumulative revenue. And really, what are some of the reasons why this type of growth has occurred? So our customers, again, are trying to find defectivity in their gate-all-around structures. And you can do that in some cases with optical, as I said, with larger defects in these structures. And so when they use the traditional defect review SEM using a TFE technology, they often go to the sites that the optical inspector directs them to and they see nothing. Now that's a problem. Because if there really is a defect buried in that stack and they can't see it, they'll often make the decision to detune the optical inspector. They'll call this a false defect. And if they go back and they detune the optical inspector, then what they're really doing is shooting themselves in the foot. What's the best thing you could possibly do is that give them confidence that when they go to a site, if there is a defect there, you will be able to detect it and image it and identify it. And that's what we're seeing with the SEMVision G10. It's allowing customers to identify defects in these buried stacks and give them much more confidence that their optical inspectors are working properly and they can, therefore, tune their optical inspectors to get the most out of them. And this is essentially why we have market share in excess of 75% in this segment, because we're able to do this and we're able to do this in HVM environments. Other applications that have emerged with the G10 is really imaging of EUV resist. EUV resists are quite sensitive to eBeam and if you image them with either too high of doses, you can damage them. And so it's one of those things where you go in and you're doing an image in the area, but you're also changing the material. You'd rather not do that. What you'd want to do is to have the highest resolution possible and then inspect them with the lowest possible dose so that you're not altering the EUV resist. We feel that we've been able to bring that type of capability to our customers with the G10. Here's an example where if you'd reviewed the same layer with a TFE system, you can see that there's something there. You can see what you might call a protrusion there in the image. But what you don't really know is: Is that a killer electrical defect? Is that a short? Is that something that I really should be concerned about? Or is it simply a nuisance? By applying the G10 with its superior resolution and the lower dose, we're able to tell the customer that, that is indeed a short and that is a killer defect and they need to take action against it. We call that -- what we're doing for our customers is giving them clear actual insight for this type of application. And then, finally, an example is the DRAM example. So as the stacks have gone to several hundred nanometers, defects at the bottom of those stacks, residuals at the bottom of the stacks that are shorting 2 individual cells will create a defect. That's what we would call a killer defect. It's very important for our customers to be able to image that and determine that, that defect exists. Again, what we're seeing with the TFE technology and without the back scattered electron technology that customers can simply miss these types of defects. Again, they may be guided there sometimes by an optical inspector. If they go there, they see nothing, they detune the optical inspector. In the case of using the G10, they go there, they see the critical defect and they get the actual insight that they need to determine what their yield killing defects are in DRAM. And this, for them, greatly can accelerate their ability to debug their advanced DRAM processes. Now let's go on to another product, I think, that we wanted to talk about today. So we've spent now 2 years debugging the CFE technology and making it high-volume manufacturing capable. And over those 2 years, we've seen that we have the ability to image defects that other technologies couldn't. So it makes sense that now we should take this technology and turn it into direct inspection. And so that's what we've done with the PrimeVision 10 product, is we've taken the CFE technology and we put it into an eBeam inspector. And here, we're providing resolution on the order of a nanometer to find the smallest possible surface defects. We're also adding an highly efficient backscattered electron detection in order to find very small buried defects. And we're doing this at speeds up to 10x greater than what a TFE type technology can do. So we're talking about best-in-class inspection, finding defects that other TFE technologies can't find and doing it 10x faster than what they could do with the TFE technology. And here's an example. We have a few of these systems out in the marketplace today. We're exploring these types of applications with our customers. And what we're finding is exactly what I said earlier. While maybe optical inspection can find a few of these silicon germanium remains, the larger ones in the gate-all-around stacks. If you really want to find the very smallest ones, the 10-nanometer ones, those still matter. And the way you're going to have to do it now and in the future is to apply eBeam inspection to gate-all-around in order to debug this type of problem in your process line. So we're very confident that our customers are going to pull for this technology because it's the only type of technology that can enable them to find these types of defects in these critical gate-all-around stacks. And another example here in a DRAM structure. Again, do we have the resolution to really tell what this defect is? So here, we have a bit line application and it looks like a cell, a DRAM cell is shorting to the bit line. But we're not quite sure with a traditional TFE. And if we run this on the PrimeVision and we look at the exact same type of defect with the resolution, we're able to determine that indeed it's a short, it's a critical defect and it's something they have to take action around. So these are 2 proof points where we're seeing the CFE technology applied to EBI as being something that will be enabling to our customers. And let's talk about one other application, which is really the EUV patterning challenges. So all the industry cutting edge is moving to EUV adoption and they're all seeing the similar challenges. Here at the very, very small design rules, both local and across wafer, critical dimension uniformity is very important. Also what we're seeing at these very, very small line widths is something we call line edge roughness. The variation in the lines that are being printed can have a significant impact on electrical performance. And that, of course, is also increasing -- that variation is increasing as we've gone to smaller and smaller design rules, approaching 30% of the total line width that variation. And then finally, as we wanted the EUV steppers to have as fast as possible throughputs, we tend to want to use the lowest possible dose in order to pattern. But there's a statistical process we call stochastic pattern defects that creep in here. And as you turn down the dose, you have a much higher probability of inducing these types of defects. And so all of these type of problems are really being best served by eBeam technology because you need extreme resolution in order to see them, identify them and correct them. And that comes to an application that we've been looking at Applied that really shows the strength of Applied as a company in the semiconductor industry. And so the 2 things I want to talk about here is the ability to integrate process. Applied has the largest portfolio of process equipment in the industry. And when we look at trying to optimize EUV patterning, what we found is that in a single chamber, we can combine both the process of etch and deposition and do that through cycling, we can create better pattern fidelity for our customers. So this integration is unique to Applied and has turned out to be a very effective way to reduce some of those variations I just talked about in the EUV patterning process. So what comes out of this integrated process is what we hope to be improved patterns. But you really don't know unless you can image those patterns and get actual insight around what's really happening at these very, very small design rules. And what you see using TFF technology is sometimes it's confusing. Do you have low line edge roughness? Do you have a protrusion? Or do you have a short? Once taking this to CFE technology, it's much easier to tell what the line edge roughness is as well as whether or not that's not a protrusion. Once you image it with CFE, you find it's a bridge, it's a killer defect. Now this actionable insight is then used to feed back to the tuning process and allows our process engineers to come up with the exact right recipes on this integrated process in order to minimize these type of defectivity. So what we're seeing here is the unique strength of Applied Materials. Not only are we able to do process integration, but we're the only company in our space that has both process and high-resolution metrology and inspection capabilities. And what that's doing is shortening the time to market or the optimization of these advanced processes, both in our labs in Santa Clara and also at our customer sites in the field. And here's an example of what it really means to do an edge and depth. You can see that they're in the center of the image, running, whereby going through this cyclic process, we're able to make sure that the pattern fidelity and the smoothness of the pattern is maintained through this process. Okay. So in summary, we've had great momentum in eBeam over the last few years. When we look at our calendar year '21 revenue, it was over $1 billion, and the year-on-year growth for that revenue was 95%. When we look at what's happening in the eBeam services sector, we see a lot of pull from our customers to service these tools. They're very sophisticated. They're something where our customers believe we're the world's experts in maintaining these tools. And we're rewarded by having greater than 90% of our revenue on service being subscription based as well as the renewal rate being in the low 90% range. So it's a very, very good service business also coupled with this technology. Now today, we talked a little bit about the G10 and the success of the launch of the G10 over the last 2 years with greater than $400 million in cumulative revenue. We're seeing it becoming the tool of record for debug of all gate-all-around processes. And we're seeing it highly adopted in both FinFET and memory applications across the industry. And then, finally, from our learnings from the G10 over the last 2 years, we're introducing the same CFE technology into eBeam inspection and we believe that we will find unique defects with this technology that no one else can find. And we can do it at throughputs and rates that are 10x greater than what you could do with standard TFE technology. So I want to thank you for listening to me today. At this point, I'm going to turn it back over to Mike Sullivan for Q&A.

Great. So thank you, Keith, very much for your presentation. And now it is time for the Q&A portion of the meeting. Before we begin, I'd like you to know that the materials from today's meeting are being made available on our website, the Events page of our Investor Relations website at appliedmaterials.com. Also, when you sign off today, you're going to see a 1-minute survey. Please do give us your feedback to help us improve future events like this. Now there are 2 ways to ask a question in today's meeting. First, you can use the raise hand button on your screen to join the audio queue. When you're selected, you're going to see a little prompt asking you to unmute your microphone. Since we do have a little bit more time today than we would have on an earnings call, feel free to ask multiple questions if you have them. Also, if you would like to ask a question later, if you hear something that's interesting and you want to follow up, just re-queue, and we'll come back to you. Second, if you prefer to submit your questions in writing, please just type them into the Q&A box on your screen, and I'll read them out for you. Okay. So what we're going to do now is assemble the queue. And it appears that our first question is from C.J. Muse with Evercore ISI.

Great. Can you hear me?

We can.

Perfect. Wanted to really kind of hit on high level. The industry has been talking about eBeam for what feels like a decade plus replacing optical. Yet sitting here, looking at your numbers for '21, it's roughly 20% of the market. So as you think about some of the new structures that are counting both in fab logic, memory, where do you think that penetration for eBeam can go? And in particular, I'd love to hear your thoughts on the eBeam inspection side solely? And what you can do to really drive the higher productivity to enable the penetration there?

Great question. And I think it's -- we have to be always a little bit cautious about saying there's going to be some type of wholesale inflection between optical and eBeam inspection. As we said at the very beginning, they're complementary to one another. So I don't think that within the next several years, we'll see something where the market all skews over to eBeam and optical inspection goes to a few percent of the market. And I don't want to try to make that statement or give you that impression. But what we're seeing is these structures, these architectural changes in the marketplace are driving our customers to adopt more and more eBeam solutions. When you look at optical inspection, sometimes if the top layer that you want to inspect through is poly, you can't get light into the poly structures that is less than 380-nanometers. So you can't actually see down into that structure at your highest resolution for optical inspection. And so the only way our customers are able to see down into some of these very complex 3D structures is to use electron sources, high-brightness electron sources and this backscattered electron technology in order to see down in them. So I think the inflection will be gradual. But as they create architectures that demand more and more of that capability, I think we're going to see more eBeam adoption. Another example -- and I don't have a crystal ball. But when you start looking at how 3D DRAM is planned, I think that, that's going to be another architectural inflection where you're going to see greater eBeam adoption because of the limitations of optical inspection.

Great. And then, Keith, the second part of C.J.'s question was in inspection specifically then, do you think that we'll continue to have very large markets for optical and eBeam inspection? Or -- because I think there was a point in time when some companies were arguing that eBeam would take over. But we're not expecting that, right, in terms of market size?

Yes. Again Mike, I'm not going to say that I think that eBeam is going to take over from optical. But what I would say, when we talk to our customers today, they are asking us very pointedly, "What is your strategy for scaling eBeam when we get to the angstrom nodes, when we get to the A-20, A-14? Can we expect you to build and deploy high-volume manufacturing inspectors because we believe we're going to need them?" And I think if you ask that same question of our customers 5 years ago, they would have said that, that eBeam technology does just fine in R&D. That's where we think we're going to use it. And our plans for HVM adoption, especially in advanced logic are either 0 or very limited. So again, I don't know exactly what will happen out in the later years in '28, '29, '30. But we're just hearing our customers say that we believe we will need that for HVM. And I think that is a change in the sentiment in the industry.

Okay. And is it fair to say that we're still investing research and development in our own optical inspection product lines and see that as a growth opportunity for the company?

Absolutely. And I think that's a great point, Mike. It goes back to the idea that these are going to be complementary technologies even out in the '28, '29, '30 time frame that I mentioned. I don't know exactly what the market split will be out there, but I do know that there'll probably be a very large optical market still. But there will be probably a larger eBeam market than what we're seeing today. And so from the Applied perspective, we want to be in a position to actually take advantage of both those markets in those outer years.

Okay. Great. The next question is a writing question from one of our analysts on the East Coast. And what they're interested in is that we have a competitor who's talked about multi-beam eBeam technology. And so the question is, are these products multi-beam eBeam or not multi-beam eBeam? And if they're not, why not? What are the differences? And why are you not going down that route if that's the direction you're taking?

It's a great question. And I'll try to give as simplistic as answer as I can and hopefully not overcomplicate it, okay? So one of the things that's interesting about the multi-beam architectures -- well, let me back up and answer the first part of the question. The technologies I talked about today are not multi-beam. So I just want to make that very clear. They're all single beam. And there's 2 reasons why they're single beam today. One is the collection of backscattered electrons which allow you to see down into the structures and image those very, very small defects in those structures. There really is not a viable technology that allows you to collect backscattered electrons and scale to multi-beam today. So the people that are working on those multi-beam technologies today are really only providing surface solutions. So they're providing the ability to look at surface patterns or surface defectivity. They aren't allowing the customers to see deep down in the stacks. So we think the technology that we're talking about today has a unique place in the market by being able to enable the detection deep into the stacks. 2, when you look at what you can actually scale multi-beam to, it's still very much limited by the source brightness. And so many of these technologies today that you see the customers deploying are TFE sources. And so I'm not going to be here to talk about anything today about multi-beam or a multi-beam strategy. But if we were to do multi-beam, we would do it with the CFE source. We would have a much brighter source. And we believe our multi-beam technology would then be much, much faster than what our competitors are introducing today.

Okay. Excellent. So what I think we'll do is go back to the audio queue. And I believe we have a question from Harlan Sur, who's with JPMorgan.

Can you hear me?

Yes.

I've got 2 questions. My first question, the resolution improvement is very compelling with CFE. I would have thought, though, that the first use case would be for CD-SEM giving critical dimension control. And measurement is so important in next-generation technologies in addition to metrology applications like edge control or placement errors. So what's the road map for introducing CFE into your VeritySEM and PROVision platforms?

It's a great question. We actually talk about that all the time in our yearly strategic planning processes. The CD-SEM introduction into the product, we could do that. But unfortunately, it doesn't quite make sense because this technology actually drives the cost of the tools up fairly significantly. It's an expensive technology. And when you look at the CD-SEM applications, it's not really necessary. You can actually get the type of resolutions that you need to meet the CD-SEM market with TFE, but you just do it quite slowly. And our customers today aren't really pulling for a very, very fast CD-SEM technology that would justify the expense and the effort to put a CFE technology into the CD-SEM. So we'll keep monitoring that. And in the future, the market may inflect. It may make sense to put it in CD-SEMs. But today, we don't see it because the economics for it aren't there. On the PROVision product line, I think that what you will see over time is a migration of all our products to the CFE technology.

Perfect. Then my follow-up question, kind of more of a high-level question. So your innovations in eBeam have certainly driven the strong market share performance that you highlighted. You gave us the calendar '21 numbers? I know, obviously, calendar '22 is not done yet, but a good proxy for this is your fiscal '22 which ended in October. I believe the team said back during that earnings call that your PDC business grew close to 35% in fiscal '22. Within that, how fast did your eBeam business grow? And relative to your WFE outlook for next year, do you expect PDC as well as eBeam business to continue outperforming?

I'm going to probably ask Mike to answer that question.

Yes. Well, we -- so yes, I think you've got the prior statements correct. I think that we are going to wait for the third-party market researchers to come out and provide updates on the splits that Keith gave you for 2021. So we're going to be working with TechInsights, giving them our revenue for each of these 4 segments and letting them produce that. They will come out in April and they'll be using our information, where we typically use Q2, Q3, Q4 and Q1 of the next fiscal year, is our calendar year equivalent. That's the process we've used for a couple of years, and we'll do that again. I think fair to say, Keith, still strength in eBeam. But we -- one of the things that we've also talked about is that we do have some new optical technology as well. So we're hopeful that we're not only going to be a grower in eBeam. We want to be a grower in optical as well because that's a nice big market for us and we have some interesting products that are for some differentiated segments. I don't know, Keith, if there's anything else you want to say on that one.

No, I think that was well said. I think we should wait till the numbers come out. And I do think we're going to see growth in both segments.

Yes. Okay. All right. So our next question comes from one of our analysts who's on the East Coast. And the question is that we have several competitors. Keith earlier showed that we have about 4 companies that have some share in the eBeam segment. And the question is, how -- do any of the other customers have this new CFE technology? And if not right now, how soon will they be able to have it? So in other words, how much of a competitive advantage do you have on a time basis, if you can help us with that one?

Yes. I'm not going to actually comment about specific competitors. So I'll give you kind of a generic response. As Ralph said in the video we showed, this technology has been in laboratory since the '80s. So I would say those other competitors probably have that technology in the laboratory. And in fact, one of them you can actually place an order for the CFE technology or a laboratory type tool. But I think what we're trying to emphasize here is that no one has brought it out of the laboratory into an HVM environment. And that's because when customers talk about what you have to do for an HVM environment, you have to make the tools stable and repeatable. And then if you're going to, what we call, flash the tip over time -- and I think Ralph talked about that -- you've got to do that extremely quickly and take very little time to do that such that the tool can remain up in the 90% utilization ranges. And I wouldn't ever say no that they couldn't do it, but it took us 10 years to do it. And I think that we have a sustainable lead in this area for the next several years from what we see in the marketplace.

Okay. All right. So we're going to go back to the audio queue. And I believe we have a question from Timothy Arcuri, who's with UBS.

Yes, Mike. Can you hear me now?

Yes.

So Keith, I had a question. So you said the systems' revenue is $1 billion. And on the services side, you said that over 90% of that revenue is subscription. But you didn't tell us what the services attach rate is for an eBeam. So I guess what I'm asking is what's the size of your eBeam services revenue? And then I had a follow-up to that as well.

Yes. Mike, I think I'll turn that over to you because I don't know if we had individually disclosed that.

Yes, we haven't broken that out, Tim. I think it's a really good question. What we have been willing to do is to talk about how much of our services revenue, parts and services at the AGS segment level come from subscriptions versus non-subscriptions and what we call transactional. And so for the company overall, it's 60% subscription, 40% transactional. And what we disclosed here today, we did decide to peel the onion on what that looks like in this business. And so the subscription portion versus -- for the company over 60%, but for this business, eBeam, it's over 90%. So a very high amount of subscriptions. And Tim, if you're interested, maybe Keith could talk about why that is, because it's really higher than anything else in the company. And I don't know. I think that was the gist of your question, right, Tim, it's just the revenue from services from PDC?

It was, yes. And then, I guess, just you can talk about this, but my follow-up was just going to be that is eBeam really inspection in general, for that matter, any more or less amenable to services' attach rate than any of the films applications. And it sort of sounds like that's what Keith -- you were just going to answer. So that would address that.

No -- yes. And I'll just give you a little bit more color. The types of subsystems you see in these products are very, very sophisticated. And that level of sophistication, I think, really drives our customers to feel that it's best left to the experts. The adoption rate in the process side, I think, is due to the fact that the level of system sophistication at least on some of those tools is not as high as the metrology and inspection tools. And so that's what really creates the differential in the adoption rates of the subscription revenue.

We talked about the [indiscernible] rates. At the company for subscription agreements being about 93%, and we're seeing something just like that here in this business. And I think we're getting close to the end of the hour, maybe slightly over. We do have a question, though, on the line, we believe, from Stacy Rasgon, who's with Bernstein Research.

Great. I have one high-level question and one dumb question. For the high-level question, you talked about process control at like 13% of WFE and you kind of mentioned it's been there for a while. At the outset, you think that percentage should be going up, but it really hasn't until now. It does sound to me like you do think eBeam as a percentage of process control might be going up. But why do you think process control as a percentage of total WFE has been so stable? And do you think that percentage should increase going forward?

It's a difficult prediction to make in my mind, because it has been very stable for many years in that type of range. And I think what happens is there's kind of an economic cyclical effect in the industry, where any given year if a customer needs advanced tools for R&D and metrology and inspection, it might go up in that given year. But then they'll use those tools that are applicable for multiple years and then they'll cycle the spending, say, to the advanced lithography or the advanced etch or the advanced film. And so I think it's difficult for me to say that I could make a case that it's going to go up in a linear sustained rate. I think that it will be cyclic. When there are new especially logic and foundry nodes introduced, it might go up 1% or 2%. But as they build out that capability, they'll probably bring it down in outer years. So I don't see a long-term trend of it trending above what it's historically been at.

Got it. Now for my dumb question. So you talked a lot about like using backscattered electrons to find these very small buried defects. If the only way you can really see these is with BSE and maybe with cold field mission, how do you know that they're there in the first place? Like, how do you find -- how do you know what to look at in order to use the new tool to actually find them, if it's true that you need the new tool to find them anyways?

Yes, that's a good question. So what typically happens is that you'll set up an inspection of an area. And you'll do that inspection and you'll look at it and you'll go, "Okay, there are" -- "there's a bright variation in this area that we can't see on the surface, so it must be buried. So what's there?" And then what a customer will typically do is they'll take it to a destructive technique, which we call TEM, and they'll actually cut into the device and then they'll do the imaging in order to verify that, that is a real signal. And then they learn when they see signals that behave in this way, that's the silicon germanium remain or that's a void. And a cut customer, for example, told me when they first started looking at the G10 and doing review, they didn't quite know what they were seeing in those buried defects. And they said that they went to the TEM and they did 100 cuts down into the device to see what was actually there. And they said there was a 99% hit rate. And so it gave them great confidence that what we were able to image with the CFE and the backscattered electrons was real. You could imagine if we're showing them signals and they go do those cuts and they have a 5% hit rate, they would not adopt this. But I think this 90-plus type hit rate -- and when we say there's a defect there with either a G10 and now the PrimeVision 10 is convincing them that they can rely on that type of signal without necessarily having to do the cuts. And what's important about that is then if they can make decisions based upon the signals we're showing them, they -- and they don't have to do TEMs. TEM's one cut usually takes 24-hour turnaround, whereas inspection can take minutes. So we're saving them hours and hours and hours of time once we build confidence and the fact that what we're showing them is real. And that's what -- that's an economic driver to avoid those TEMs in the future.

Yes. That was actually a very good question. So we'd like to thank you for that. I'd also like to thank Keith and his teams in Israel and Germany for developing today's content. And I'd especially like to thank everybody for joining us today. As you sign off today, please don't forget to take a minute to give us your feedback on the survey. And we hope that you have a great rest of your day.