We better get started. My name is C.J. Muse, with Cantor covering semiconductors, semiconductor equipment. Thank you all for coming. I have the honor to introduce KLA. Always think about KLA-Tencor , but KLA.
Thank you.
Team.
Thank you for having us.
Rick Wallace, CEO, Brenn Higgins ,CFO. Welcome, gentlemen.
Thanks, C.J. for having us.
I have a number of questions, but I figured I would start with the all-important, you know, where are we in the cycle? Most management teams talked about kind of an up, mid-single digit number for 2025. You know, my sense is everyone thinks they're gonna beat that. Therefore, are we not in, like, maybe an up kind of 5-10% kind of world for the year, which, considering where expectations were, you know, three, six months ago, that's a real kind of positive surprise. I guess just first question, do you kind of agree with that assessment? You know, what's your early thoughts on 2026?
I'll let Brenn take that one.
Yeah. 2024 was a good finish for the company, where WFE levels were about $95 billion company. It was up somewhere in the mid-single digit. The company was up about 12%. We had a nice relative performance last year, and had a record revenue number too for the company, which, I think under the circumstances with some of the moving parts we had related to China was a pretty good outcome. You're right. You know, our view on 2025 is more or less a mid-single digit type growth year for the industry, which takes WFE levels up between $100 billion and $105 billion. Look, we'll see how that plays through.
Certainly for KLA, a return to growth at the leading edge, both in logic and foundry, but also in memory is a real driver for our business. We're excited about those opportunities. I'm sure we'll talk more about them. I think that is the biggest driver. It's a higher process control opportunity, particularly compared to the last two years, which were more legacy centric. Easiest way to think about it is that logic and foundry, WFE on the front end grows, in terms of leading edge grows, offsetting some digestion and export control affected business in China. The legacy parts of the industry are more or less flat. We see a little bit of growth off a very low base in the NAND flash market.
Investments in DRAM specifically supporting HBM drives the uptick in the year. Against that backdrop, we feel very good both in terms of our relative performance, both in logic, foundry at the 2nm node versus 3, but also the dynamics around HBM and how that translates into opportunity for KLA. Packaging is also a big opportunity in 2025. We talked at earnings about growth of at least $800 million, up from $500 million in advanced packaging. Another vector there of opportunity supporting what looks to be a pretty strong, not just this year, but as we go forward, high performance compute market. Just looking at all that, I think it translates into, you know, mid-single digit kind of growth. We feel very good about KLA's relative position.
We'll see as we go through the year whether that view changes. That's where we sit today.
Perfect. I think I'll kind of pivot now between some of the end markets and '25, '26 trends and key technology inflections for KLA, you know, this year and beyond. Maybe to start, optical versus E-beam, obviously you guys are the best in the world at optical. We met a few weeks ago, and the comment that you made to me was you are all in on E-beam, which seemed like a bit of an inflection. Can you expand on that?
The inflection is that it's showing up in the market. We've been spending on E-beam, as I talked to Ahmad for quite a while. The question was, when was it gonna actually start showing up and wins? We've always had the view that E-beam and optical are complimentary. What we needed to have was a fast enough E-beam that was also gonna be able to sort through a lot of the noise. It's funny when we talk about AI in our products, the first product to have AI was an E-beam inspector from us back in like 2018. We're finding now with the latest generation that we've introduced, that we're actually getting wins and we're seeing growth.
Some of our share gains this year, what we anticipate and, you know, as part of outperform is because of the strength of the E-beam portfolio. We are seeing, it surprised Brenn on the operation side that we keep getting increase in the demand for that.
Yeah. Yeah. It is, you know, it is interesting for us 'cause it has always been something that has been somewhat complementary. They are complementary markets to the optical franchise. Now you have other applications, particularly in terms of Gate All Around and characterizing transistors, and transistor performance. Over time we expect roughly 80% of the dollars in pattern inspection to be spent on optical technologies because those translate to production. E-beam has great sensitivity, but is incredibly slow. It is more than just pointing our inspectors and increasing the relevancy of optical inspection.
There are unique applications that we think we can differentiate at also. That is also a factor in how we see it. Now there are also other parts of the market. There is a metrology market where you are doing more in-die measurements for overlay metrology as an example. Then there is review. The dynamics are anytime you ship an optical inspector, you have to ship a review system to review and classify those defects. It is something that being able to do that faster is a driver because of all the, to Rick's point, all the nuisance that is there. It is absolutely a factor that we think is a share opportunity. Those are markets we have not had a strong share over the years. There is growth.
We will lay out a deeper analysis on this at our investor day in June, but we're certainly encouraged by what we're seeing in the market today.
Maybe just to get ahead of that a little bit. You guys exited CD- SEM, but you just talked about SEM review throughput.
In 2007.
Yep. Yeah, yeah, yeah. Not suggesting near term, but just trying to get the landscape of where you're playing.
Yeah.
It's SEM review and it sounds like throughput, there's some advantages there now. The other one, big one is EBI. I guess for E-beam inspection, are you attacking all three markets? Voltage, contrast, physical defects, buried defects?
Yes. And also reticle. I mean, we have a system that's being used for advanced reticle inspection too. Yeah. The answer is yes. I mean, part of why the strategy took so long for you to hear about was we had debates about this internally. Do you develop one segment and then grow, or do you develop for all the segments and bring them out? We actually went the second path, which is a longer, more expensive, but that's what's showing up now. We have a portfolio of products for our customers and we're essentially seeing it both in the logic, advanced logic, but also advanced memory showing up and getting really strong results.
Part of why I wasn't sure I was gonna be here is I was in Asia last week and I was in Taiwan and Korea talking to customers. It was interesting how their view of our E-beam has changed pretty dramatically from prior visits.
How so?
Very positive and wondering why they can't get more capacity.
What are the lead times for E-beam inspection?
From a new order now, it's either pretty consistent with other lead times. It's somewhere in the nine- to 12-month range. I think we didn't fully anticipate, you know, the strength of the demand. We are struggling a little bit to ramp, although I think we're in a much better position today than we were before. You know, the demand's been strong and, you know, we'll work our way around it. I don't think it's an impediment to the opportunity. It's just one of the usual operational challenges you face.
When you talk about optical and E-beam complimentary, obviously they're very complimentary in terms of, you know, as you talked about what you can see in throughput and balancing that. Is there an opportunity and an advantage for a customer to select you for both?
Yes.
Where they work well together.
Yeah, absolutely. Both from, we're now sharing some of the algorithms. So some of the AI work has been done on both. And we actually have linkage between the system so that to Brenn's point, you can point them, but also you can validate. So if you can find it on E-beam, can you then tune the optical to run it at higher speed and be able to continually monitor? That was always the question, what was the value of the portfolio? Was there synergy? Finally we're seeing realization of that, and customers are playing it back to us.
I guess is, is there a framework for thinking about, the driver of adoption from, you know, three to two to, I wanna get the name of my wrong, A16, A14.
Right.
A14. Yeah.
I think it still, the customers are exploring. I mean, there's a waterfall. If you look at, you know, a thousand steps in the fab and there's, let's say, 200 that are being inspected, monitored on a routine basis. And like a, you know, right now, for the most part, only two of those in the advanced logic are being used by E-beam. I mean, mostly everything else is the fastest tool that can do it. So whether it's a dark field tool or whether it's a BBP tool. And so the numbers are expanding as you go forward, provided we can have the coverage and the throughput, right? The challenge always for any of these technologies was if it's used as an engineering tool, it doesn't drive capacity because you're only debugging a line with it.
The answer is yes, we're seeing them implement. There's some layers now, and this was a shock to us that they're inspecting twice the same layer with optical and E-beam because they're seeing different things.
Wow. You also talked about reticle, with E-beam.
Yeah.
Obviously, you know, very strong 640ES, Teron optical, the 8X multi-column E-beam, but also soon the Teron 710X, . so as you look at that portfolio, is there a framework for thinking about, you know, your share for overall reticle inspection?
Yeah.
Should there be a difference between in-house versus mask houses?
Not a huge share. I mean, recall is a different application for sure, and then it's more gonna be optically based. When you say soon, I think soon on 710 is.
Sorry, 26 to 28.
Okay. It depends what your timeframes are for soon. That is, the answer is yes. Really, actinic for reticle, our view has been consistently that when it's really gonna be needed is when there's high NA lithography, EUV lithography, which has slipped out. You know, I think there might be a couple layers at A14, but it's out there a way. I think it'll more intersect with HVM for that. Later, right now the question on 8X has been, can we continue to provide more capability by adding more columns and getting the throughput up? That is, we're getting good feedback on that. The optical tools continue to expand based on algorithms and just what we've always done, provide more illumination and more algorithms.
We're covering most of the market with the optical systems, right?
That 90% of the EUV reticles in production are going through optical systems, given the nature of the lines and space and density that's on the reticles. We cover most of the production use cases. You have what's happening in the fabs where they're requalifying reticles. You also use a wafer inspection tool that does not show up in how you calculate reticle inspection, but it's a use case for wafer inspection to print the wafers and use that to validate the fidelity of the reticle. Those are the principal tools in the market. You know, we've got some advanced capability obviously with E-beam and actinic down the road.
We just talked a little bit about a partnership we have with a key supplier, with Carl Zeiss, as it relates to the optics for that system. The ability for us to make adjustments to the NA very quickly, and there is some synergistic value between this optics and the support that they do for the scanner and for the inspector. I think we are in a pretty good place when the market need materializes there.
Perfect. Maybe moving to advanced packaging, including HBM. You talked about $500 million in 2024 or $800 million in 2025, obviously great growth, but I believe that's largely co-ops. And so curious, as you think about going to HBM three and four, can you kind of walk through how the opportunity opens for KLA? You know, where are you in terms of having the right price, right gross margin to KLA tool to serve that market? And how do you kind of balance, you know, the margin requirements, requirements that you need, but also this high growth market that other players have traditionally supported?
Maybe I'll intro and Brenn can fill in the details.
I remember a meeting, it was probably 18 months ago, one of our major suppliers where they told us that we want your WFE, you know, your, your front end stuff in the backend. We said, I don't think you mean that 'cause it's too expensive, right? You, you, you're gonna say that and then we're gonna put it in there and you're gonna say that we, and they said, no, no, we're dead serious because what we're seeing in packaging is the requirements are looking a lot more like the front end. You take what was a low end in the front end and it becomes a high end in packaging.
A lot of our growth and success, in addition to some of the process stuff, but a lot of our growth has been that conversion of some of those capabilities and just the adoption of optical inspection, the adoption of some overlay, the adoption of flatness. Those are, what are those films? Yeah. Those are the reasons. You know, what they're telling us is the cost is so high. I mean, C.J., I always look at what is the cost of what they're making and what % can they add on inspection, right? If you think about a reticle, you can understand why they can spend as much as they do. The cost of the reticle is so high.
The issue was the cost of the packaging was not high enough to spend a lot of extra money on the inspection, but that is not true anymore. The cost of failure on a package, and so they are throwing a lot more process control intensity at it. What they are telling is that is not slowing down, right? They are trying to streamline the HBM process, but even as they do, they need more. The answer is, you know, we say $800 million, but we do not think it stops there in terms of growth. The beauty is we have got all the roadmap, we have all the products, right? Because all you have to do is continue to adapt the stuff that we have in the leading edge.
The challenge for the guys that have come at this market from the low end was that they're already stretched to the top of their capability. You know, when we talk about packaging as our low end is their high end and they're kind of at a limit, whereas we've got a huge roadmap plus the opportunity that we have because we have the processing SPTS division. Those are kind of the roadmap we see from the packaging guys is fully integrated to the conversations we're having with our leading customers on front end. I think that market keeps growing for as far as we can tell.
If we isolate to HBM, can you kind of walk through, I guess, bumps are getting smaller, wafer thinning, what are the tools that you see as an opportunity for KLA?
Those are the things that are happening, right? I mean, generally is, is that you'll run into height restrictions. That will, you see, you'll have, you know, thinning or smaller bumps, thinning wafers. Then you also run in eventually to hybrid bonding techniques. Right now, historically, bump inspection has been more of a legacy packaging market, but as you start to increase the capability there, it creates opportunities for us to address with higher value systems. They are the same systems and co-ops, but, but it's, it's, you know, inflecting more from a technology point of view. Most of our exposure today has been, to your point earlier, has been logic and, and, but we're, we're increasingly optimistic about the momentum in memory as we go forward for some of the things that, that you mentioned.
As you think about going to five or three nanometer on the logic die, for HBM, freeing up space there for adding, you know, incremental logic, is that an opportunity that you see as well?
Yeah. HBM four, you mean in memory or we're still talking packaging?
The logic die within the HBM four stack.
Yeah, yeah, of course. I think that they're, you know, back to the, every conversation we have, customers are frustrated that they're having to add process control as a % of their spend, right? That's a natural thing for them, but they're also saying they have to do it, right? We know that there's a driver and opportunity. They want cost-effective solutions to do it. You know, they want us to re-engineer some of our tools for those applications. I don't know what the,
yeah, I mean, one of the things, it's roughly 12 nanometer today. To your point, going to something more, more aggressive. If you look at where the biggest opportunity in HBM is, it's how it changes DRAM intensity on the front end.
Forget the packaging part of larger die because you have to drill the TSVs, you get microcontrollers, you have less redundancy. You gotta make, do you have to do this logic-based die? Then you have reliability checks to make sure that each of the chips functions, well, in the stack. All that's driving the bigger opportunity for KLA in terms of just overall intensity as it relates to DRAM. You take that HBM, do some of the connection stuff that we talked about earlier and integrate it in the interposer with the logic chip. That's the evolving opportunity, but we're really encouraged by what we're seeing just in terms of the DRAM die and the DRAM processing relative to advanced DRAM in terms of the intensity.
We hit on advanced packaging, but really focused on HBM. Are there other technology inflections that we should be thinking about for other parts of advanced packaging?
I think the biggest issue is, is that you start to shrink the interconnect density in the interposer, which drives you to need more sensitive solutions. To Rick's earlier point, we're addressing this part of the market with very high sampling rates near 100% inspection with our lowest end capability, but we have the portfolio all the way up to, you know, E-beam. Not saying E-beam's part of this, but, but we, with, as you start to shrink the density, then you're gonna need more advanced systems. Typically when you need more sensitivity, there's a throughput trade-off. It is why I, I'm less concerned in the long run about the margin, gross margin implications of packaging today, which is a negative from a put and take point of view to the gross margin because it is our lowest end solutions.
Over time, we would expect that we're gonna ship more advanced solutions, which will carry a, a stronger margin profile that, that I think turns it into more kind of normal KLA over time. Now, when we talk about these deltas, I mean, it doesn't change our longer term view of the margin profile over time, but certainly in the nearer term has been a bit of a headwind as those tools carry margins that are different than some of our advanced front end tools that make up a bigger part of our mix.
Gotcha. Maybe focusing on leading edge foundry, which you alluded to a little bit earlier. I think there's a couple areas where you're very excited. First one, going to N2, I think you've talked about process control intensity higher by 100 basis points. Secondly, as you move to HBM very quickly on five and three, you start talking larger die and requiring more process control insertions because you don't want to, you know, have the yields on those larger die, you know, be a disaster. As you think about those two trends, how are you thinking about overall kind of foundry for KLA process control intensity into the 2025, 2026, 2027 timeframe?
You can ask the number. Yes. For N2, and that is KLA share of market, right? We have talked about our versus N3 that our share of market expectation is about 100 basis points higher, and a piece of that is share. We talked about E-beam earlier, but also there are intensity changes. The nature of a gate all around device is complex. You still have, while you do not have an increase in the number of litho layers, you still have very advanced litho EUV layers or increase in EUV layers. They are scaling defects down in the interconnects of the transistor design that are still a challenge. You have new defect mechanisms related to residual defectivity, buried defects, and so on. All that we think is contributing opportunity.
The last time we had an architecture change in the industry, KLA's share of market approached 8%. I think when you look at the numbers we talked about earlier, I think we're more or less approaching that again overall, across the market. If you go forward, you look at the ramp of N2, particularly with the number of designs, some of the issues you talked about, large die, higher value die, all those things we think are conducive, less reuse because of the adoption of the node, the strength of N3. We are pretty optimistic about that, but also the roadmap going forward, both in terms of shrink, but also then you start talking about power distribution, other opportunities here over the next couple of years.
I like to say that, you know, you have all over the last five years or so, you've had a lot of things that have driven KLA share of market up, logic foundry mix, relative to memory, well adopted nodes, limited reuse. I think all those things continue, but now the mix is shifting. The mix is shifting to high value die, big die. And so that's creating, I think, opportunities for us to see it continue to grow. The one, continually surprising fact is the number of leading edge designs keep going up, right? And so, you know, you think about the conventional wisdom years ago was there'd be fewer because the costs are so high. Yet what we're seeing is, and this is a unique benefit for KLA.
If the number of designs go up, it means it's a higher mix fab, which means process control becomes more important. It's good for the reticle market because it means there's more reticles being produced. The cost of these are so much, back to my point about the need for inspection and measurement is also supported by the cost structure because you have to make sure that these are right. We also know that there's an undersupply relative to demand of 2nm right now in all the forecasts. We know that there's more stated demand. When you translate all the hyperscaler, you know, CapEx, they could come down a lot and you'd still be short of demand. That's a great driver for us. The other thing is that it's playing to our technology strength.
We're gaining share because the performance differentiation becomes more important. All the work we did on AI in our systems is showing up with our customers. You know, that's why we're pretty bullish on that, but also on the memory side of everything that's happening in high performance compute and AI, those memory chips are getting much harder to inspect. I mean, it's literally memory's more important, the logic's harder, and the packaging, right? We have all three at the same time. Memory, I was thinking about this, there was a time where memory was our big driver. Then years ago, people went to 100% redundancy. The reason they did that, and so, you know, if you had a problem with your computer, you just reboot it because it was a memory leak, right?
That's not gonna be okay in these high performance compute, right? Suddenly the intensity around memory manufacturing for inspection and measurement has gone up again. Because they're terrified of being the weak part of that very expensive stack. We see process control intensity going up there. It's kind of high performance compute is, if we had to design a technology to drive our markets, we couldn't have come up with a better one than everything is happening with AI and high performance compute.
Maybe a question on Intel. I know you can't be too specific, but I guess following on your comment around any which way you slice two nanometer, there's shortages. My question is geared around they either, you know, have to bring wafers in-house or they have to recommit to more wafers at TSMC. This is more of a high level kind of foundry logic question, in terms of, do you think the industry is ready already capacity wise to support that kind of binary decision or whenever it happens, we're gonna need to spend more money?
Somebody's gonna need to spend more money, right? I mean, it's not because the demand is already above the forecast supply. Nobody's allocated for that when we talk to customers and there's only a couple you could talk to. If there's success on those leading nodes, they're gonna need some level, whether it's 40,000, what do we think, 50,000, they're gonna need some level of capacity to support it. You know, the 2nm stuff is already spoken for. Now, you always plan for less than what customers are telling you you want. But yeah, if there's success there, there's gonna need to have more, more capacity added. That would, and that's upside to the model we talk about for 2025, if it were to happen in 2025. Yeah.
We contemplate different demand scenarios in terms of how we plan for our production, given our lead times, nature of the lead times, the intrinsic lead time to some of the components. We, and implicit in our model is carrying the flexibility to be able to support surprises and demand that are inside of lead time. I think we're in a pretty good place from a capacity and supply chain point of view to support the different scenarios that are potentially out there. We'll see how that plays out. To your point and to Rick's point, that seems like there are opportunities for upside out there and, you know, we'll have more to say about it when we think it materializes.
Just to clarify, Rick, you think that surprise could actually happen in 2025? My sense was there was not enough clean room space.
No, no, I'm saying they could recognize they need it, that it within '25. You couldn't bring it on in '25.
Yeah. Yeah. Gotcha. Makes sense. How about China? You previously talked about a $500 million headwind and now, you know, there's, you found other business kind of to offset it. Do you feel like you've completely de-risked, you know, any sort of headwinds there? What do you think kind of normalized China is as a percentage of your business over time?
I talked about it at earnings that we would come, our business % would come down from about a little over 40, I think 41% last year to somewhere around 29% plus or minus this year. A pretty significant reduction. There's capacity digestion that's happening. There's obviously the export issue you mentioned. The last two years have been fairly unique in that you had a drop off in investment from your non-China customers and your China customers with the greenfield facilities and a lot of public funding were able to basically fill that void. It drove the % of business to elevated levels, but it was really unique, I think, to the timeframe and the constraints we had in 2021 and 2022 in terms of being able to satisfy overall demand.
I think over time it drops down somewhere to 20-25%. There's self-sufficiency directives in China, over time. And, you know, if you look at an 80% target and where they're at, sort of 20-ish percent today, it does mean that you're gonna see a level of investment going forward, $25 billion-$30 billion of WFE. I think as you model that out towards the end of the decade, we see it somewhere in that range. Yes, we had the impact that the overall estimates didn't really change. We did see some marginal improvement at the leading edge, some additional strength in China from where we were when we sort of set those expectations. And we'll see what happens in terms of that impact over time.
You know, is there a licensing opportunity or not, you know, less clear. You can reallocate certain systems that are strongly demanded, but that's just a pull forward of business. From a gross impact level, I think we're comfortable with the estimate we gave plus or minus $100 million. Then we'll see if there are opportunities to mitigate that and that changes moving forward. At least that's where we see things today.
Makes sense. If I were to look at my bottoms up model, you know, I look at domestic China, Samsung, less so TSM or Hynix due to lack of clean room space as sources of upside for 2025. Would you generally agree with that view?
Yeah. Yeah. Sure.
Okay. How about gross margins? So you've guided 62% for the full year.
Yeah.
you've contemplated decline, I assume of higher margin China embedded in that. At the same time, you've exited, or divested flat panel business. I guess when you see an uptick in volumes, what kind of incremental gross margins should we be thinking about?
Our margins by region are generally the same. What drives our margin variability quarter to quarter tends to be product mix. Customers all buy different types of tools that carry different margins, outside of volume driven deltas. That is really the only difference across different customers. In fact, if you look at customers in China, given the nature of some of our older systems, they may actually carry lower margins. I do not think, you know, on a go forward basis that that is necessarily a headwind. We talked about the packaging dynamic earlier, so we feel like we are operating today about 62%, as you mentioned, and that is the guidance we gave for the year, at roughly $12 billion business levels. Our target for our 2026 model was 63% plus, at $14 billion business levels.
Given the mix we have, I feel pretty comfortable about our ability to continue to drive gross margins between 60-65%, which translates into a number in line with our target model. The exit of FPD will have a marginally better impact, probably, you know, 20 basis points on the overall. So it's a factor. Growth in our software business is also offsetting some of the mixed dynamics I talked about earlier. It's, I think, very consistent with what you've come to expect from KLA. The biggest dilution factor we have is our services.
Services. Yeah.
Services growing at a faster rate. We have a lot of focus in our service business about driving the incremental margins of service, but that's contemplated in that 60-65% long-term incremental gross margin target. One of the challenges in service is supporting a much broader footprint. You have to make investments more broadly, but you ultimately get leverage on those investments. We feel pretty good about the ability to manage whatever that diluted force. I mean, we love the business. It's a great business for us. Again, as I said, I think it's contemplated in that overall target.
Perfect. I think we've run out of time. Rick, Brenn, really appreciate it. Thank you very much.
Thank you. Thank you. Yeah. Go, go do that.
All right. Good morning still. My name is C.J. Muse with Cantor, covered semiconductor, semiconductor equipment, space. Very pleased to host Lam Research. This morning we have Doug Bettinger, CFO. Thank you for coming, Doug.
CJ, thanks for having me. I appreciate it.
I think Doug has an initial statement to make, and then we can proceed.
I always need to start with my safe harbor to keep the attorneys happy. Permit me. Today's discussion may include forward-looking statements that are subject to risk.