KLA Corporation (KLAC) Earnings Call Transcript & Summary

Good afternoon, everyone. Apologies for the delay. My name is Toshiya Hari. I cover the semiconductor space at Goldman Sachs. Very excited, very honored to have the team from KLA this afternoon. We've got Bren Higgins, Executive VP and CFO. We also have Ahmad Khan, President of the Semiconductor Process Control Business. I think Bren is going to start us off with some opening comments, and then we'll follow up with Q&A.

Sure. Thanks for having us, Toshiya. Glad to be here. Just quickly, just to reiterate some of the themes out of earnings last month. As we sit here in the middle of 2023, we see an industry environment generally with the WFE part of our business down about 20% overall. Our semiconductor process control business within that is set up constructively for another year of outperformance. We've got a good mix with customers. Customers are continuing to invest in road maps. We're exposed to some infrastructure investment that's happening. So we feel pretty good about our overall position there. As you look at our customers, if we just look at where we are today versus where we were at the beginning of the year, the leading-edge business has gotten worse from the beginning of the year to where we sit today. We've seen some memory has -- was expected to be low and has continued to be in line with that. We've seen some strength in the legacy parts of the business. We've seen strength in China, and we've seen some strength overall in infrastructure as it relates to silicon wafers -- the silicon wafer industry, but also in the reticle infrastructure, particularly in China. Our service business continues to grow. And overall, I would expect service to be up somewhere in the high single digits this year on a combined basis, combining our Semi PC business, but also our EPC business. Set up pretty constructively for growth over the next few years as we start to see a transition from warranty to contract or tools that we shipped in '21 and '22. We had very significant growth in those 2 years. EPC business is a tale of two stories, and one, the specialty semiconductor business, which is our process business, has a lot of exposure to power semiconductors and two, advanced packaging. And it's -- that part of the business is doing well. And against this backdrop we have in terms of overall WFE spend, we see that business is roughly flat this year. And then in the other parts of EPC, which are more consumer focused and more capacity centric, the slow -- the weakness in mobility is having an effect on that part of the business. So it's down meaningfully this year versus last. So as we look at the whole company, if you take our guidance in September and our commentary around a stabilizing market that we talked about at earnings, it sets up somewhere in the $9.5 billion range in terms of revenue, which is down high single digits from last year. So pretty good, I think, in terms of what we've seen happen in our end markets. And today, we also announced a capital structure action with an increase in our dividend, which was our 14th consecutive increase from $1.30 to $1.45 in our quarterly dividend and also additional share repurchase authorization of $2 billion. And that was reinforcing not only our view of confidence in the long term for the business but also the plan of cash flow return is 80% or more that we outlined at our Investor Day last June. So I think I'll end there, and we'll move on to your questions and happy to have Ahmad here to get into some of the dynamics affecting our customers and some of our products.

Sounds good. Thank you for that amazing overview, Bren. You talked about the construct of 2023 WFE, the pluses and minuses. I know it's way early to talk '24. But as you think about the big swing factors that you guys are focused on monitoring, watching, what are some of the potential pluses or minuses going into next year as well?

Sure, sure. So I think the biggest one is just the timing of recovery in the memory space. Hard to see how memory would be worse as you move into next year. But looking for the fundamentals of those businesses to improve, something I watch pretty closely is cash flow. And looking at our customers' cash flow, we're in a prohibitive financing environment. So I think we're going to need to see a recovery in those businesses that drives the confidence and the cash flows to be able to drive investments. So I think that's one of those wild cards into next year in terms of timing, whether we see it somewhere in the middle of the year or is it sometime later. Leading edge investment, I'm encouraged by what we're seeing so far and the adjustments that we've seen this year. I think we're fairly derisked it this year. It will vary potentially across the different customers. There's healthy competitive dynamics there that are encouraging, but we'll have to see how that plays out. On the legacy side, we've had a couple of good years of growth. So I think baseline is sort of a flattish on legacy. As it relates to China legacy probably, I don't see that falling off either. So I think it's probably baseline sort of flattish as well. Infrastructure maybe falls off a little bit. It had nice growth this year. But as that capacity starts to come online in preparation for a recovery in WFE, I would expect to see that flatten out a bit moving forward. So I think -- but the wild cards, I think, really in terms of WFE really are about the memory market and what happens at the leading edge. And obviously, some of that predicated on drivers of demand, which is more of a macro question in terms of not that bottoming out or correction, but what drives growth from here. So that's how it is.

And Bren, just to clarify, when you say infrastructure, you mean the mask shops and the bare wafer guys primarily?

Yes.

Okay. Got it. The process control market has performed really well vis-a-vis the broader market, I think, now during periods of memory strength and weakness, right? So it feels like on a permanent basis, the market has outperformed over the past couple of years. Can you talk a little bit about some of the drivers that are driving the market and your business specifically?

So why don't I let Ahmad start, and I'll add if necessary.

Yes. So process control intensity has gone up over time for KLA. One of the major drivers of that is all the market inflections that are happening in semiconductor industry. So first, I'll start with -- can you hear me now? Or -- okay. So first, I'll start with high-end logic. So if you look at it with the introduction of EUV, the device architecture changed quite a bit. That changes all the metrology and inspection modes with smaller lines being printed on the wafer. The criticality of defect inspection becomes more important, and we saw that a pretty large growth as EUV gets implemented. With EUV coming in, mass shops needed to make new investments in mask houses and mask shops by bringing in new systems that will be able to inspect these type of reticles. Once the -- once EUV comes in the pitch at the reticle changes and that drives new inspection modes. So basically, all the inflections that are coming out in the marketplace is really driving process control intensity increase. And this trend continues. With EUV, more EUV layers are getting implemented after that high-end EUV comes. Transistor architecture is changing quite a bit, as you know, away from FinFET now to going to gate-all-around and then also buried power rail is getting introduced, which is a new type of power delivery system, which brings wafer-to-wafer bonding in the front end. All this drives new inspection modes and metrology modes. In NAND, where there's wafer-to-wafer packaging also happening there, and that's driving process control intensity increase. In DRAM, EUV is getting implemented and that is driving process control intensity. So we see a pretty systematic increase in process control intensity from, I would say, high 5s to now 7% or 8% increase in process control. So I think this trend continues. One of the other things I would say is that about 4 to 5 years ago, we have many operating models in KLA and one of them is our strategic planning process where we look at what inflections are going to happen in the marketplace 3 to 5 years from now, and we do this every single year. And 3 to 4 years ago, we made a decision that automotive and legacy market is going to start to need new systems instead of the used systems that people were used to and we started releasing new systems, new Gen 4 system, new Surfscan system, new wafer thickness systems. And all of those were launched about 3 to 4 years ago. Of course, since then, as many of you know, automotive and legacy market has grown quite a bit, and that's driven also process control intensity to go out. So these are some of the factors that are driving it, and we think this continues for some time.

That's great. And then from a competitive standpoint, you guys have done a really good job maintaining share, if not extending share, despite maybe a handful of companies over the past decade or 2 coming out with pretty bold targets. But year after year, you guys come in and maintain it on extend share. Can you talk about some of the differentiating factors between you guys and the rest of the pack? And I think at your Analyst Day, you talked about the red team, green team analysis that you guys run through. So maybe for the benefit of the audience, kind of describe how that works internally at KLA?

Yes. So I'll start with very close customer relationships. And this is not about the quality of the wine we bring to the dinner. It starts with years of delivery that customers expect us to do, right? So I have been with KLA for near 20 years. My staff has been around for another 20 years in each of them. So we have a pretty long tenure. And the relationships that we have built with our customers on the engagement side has been critical to our success, which essentially, they measure it very simply. Were you able to predict the problems that I'm going to have and have solutions ready and available when that inflection happens because many times, our customers don't know what problems they're going to run into. So we have to predict those and invest in R&D to ensure that we will have those solutions available. So it's years of delivery, which has really developed a very strong customer trust. The second area is consistent reinvestment back in the business. KLA has done a fantastic job ensuring that we invest in the right things and not on the wrong things. A third area is really a portfolio approach. We're not a single product company, as you know. Just in Wafer Inspection, we've developed 5 different inspection systems. And essentially, you can plot sensitivity and cost of ownership on that curve. And we go to the customer where our best competitors. We go to our customers and say, okay, you have this sensitivity at this price point and this sensitivity at that price point. So the customer can make a decision based on their process. All processes are not the same. Everybody is different, and everybody has different requirements. So they are able to use based on the best performance and cost of ownership that they need, what systems they would implement in their fabs. So all of those together, I think, it's a very small world, right? I mean things catch up to you if you're not executing. So collaboration, innovation and execution is our flywheel that we work very closely on to ensure that, that continues. And for that reason, our customers are willing to reinvest back in us by giving us the market share, which enables us to invest back in the research and development to solve the future problems.

I think with our market position, we're able to invest at a level that our competitors struggle. And so we're investing across the portfolio, and we're allowing customers to make these trade-offs that we're fairly agnostic too, whether they have technical challenge and they're trying to balance technical requirements, which matter a lot in R&D and maybe matter a lot in a fab ramp phase, but matter less in a production phase. And so then they can trade off across that balancing active sensitivity and throughput to meet their economic priorities. So I think the ability to invest to be able to deliver products faster and invest across the portfolio makes us a moving target. We also have 800-plus applications engineers in the field that work with customers to help them with recipe optimization to get full value out of the tools. And so you can have that broad footprint of really skilled resources that are intimate with the tools that also helps drive incremental value. That's why the market share generally is the same across all regions and across all customers and all segments. So I think that there's some differentiating competitive advantages we have here. And year-over-year, I think it plays out, and it's reflected in the results that we put up.

I guess as a quick follow-up to that. I know you guys manage a very broad portfolio of products. What does the product iteration cycle look like? And you guys talk a lot about value-based pricing. How does that work? What does that mean? Obviously, it's leading to industry-leading gross margins. But how does that go about in the marketplace?

Why don't you start, Ahmad?

Yes. So another one of KLA processes that I really enjoy is our product life cycle process, PLT process, which really ensures that we are working to determine what the market requirements are, work closely with the customers to understand them and then deliver on these products to make sure that the customers get what they want. So that process really enables us to make sure that we don't make too many mistakes. And why is that important? A substitute of product, meaning you have an iPhone and you're releasing a new version, it's about 2 to 4 years to get that product out. A new product, which is meaning we're not in a segment or we're releasing something that has never been done before, that can take about 7 to 9 years. So development costs are high, but that's why our collaboration with the customer is very important, right? Because we have the largest amount of market share in the largest suite of systems, our engagement with the customer is really deep so that we know that what we need to develop and then have that release. On the value pricing front, I mean, our customers understand that process control is not like process where the number of units is quite significant. And process control is limited amount of units. It's increasing quite a bit, of course. And therefore, the research and development required in order to develop these systems is quite high. And for that reason, they have a very good understanding of the pricing that we propose. Of course, pricing is always based on a product that they already have been purchasing. So we innovate to ensure that the cost of ownership and pricing of the new product that makes the other product less attractive. But of course, because complexity continues to increase, we don't see units degrading. We see units increasing. So that's -- I think if Moore's Law continues, which it has shown that it's extremely resilient, then I think this trajectory continues.

Ahmad mentioned the Moore's Law dynamic, and I think that's an important one to understand is that we actually have a good pace of Moore's Law. It's not the -- if the industry had been on the original pace of Moore's Law, we would be spending meaningfully more to try to keep up. And so you'd see the R&D as a percent of revenue, which today is, I don't know, 12%, 13%, something like that would have been meaningfully higher. So because we've had the right cadence, which is still driving enough compelling attributes to drive the end markets, it has allowed us to optimize the platforms we have and effectively extend them over time. So we're not spending as much developing, which is when the big investments happen when you transition from one platform to the next. So we've been able to increment, we've been able to leverage things like AI to try to get more out of the hardware. And so that cadence has allowed us to drive higher levels of profitability, better leverage in the overall model as we've moved through this period of nice growth. But I think it's the extendibility of these platforms and then the ability to even take them back with new capability to sell into the legacy markets. Two things are true. Customers are getting a lower cost of ownership, and we're getting better pricing. So we're getting more today on 10,000 wafer starts of 28-nanometer than 10 years ago as an example in terms of the revenue contribution and the customer by buying the newer capabilities getting a better economic equation for them. So it works out both ways.

Got it. Got it. That's interesting. I guess one of the perception that investors had and still have today is process control intensity is very high in R&D, HBM less so. But I think over the past couple of years, several years, the intensity has grown on the HBM side as well. I know the increase in design starts is one thing that has played into that. Can you talk about the forward as you think about process control intensity specifically in an HBM environment?

Yes, I can start, and Bren, you can add to that. So if you go back 5 or 7 years ago, right, a lot of our customers used to do R&D and complete a certain level of R&D so that you have achieved a certain level of yield and then do the process transfer. As complexity increased, this has been much harder to do. Complexity in the design process itself and the number of devices that a customer needs to make in a particular fab or a particular node. So this basically drives a lot of variability in the process. And because there's a lot of variability to process customers increased process control in HBM. And therefore, a lot of our customers today get the yield to a certain level, and that level is pretty low, and then they transfer it to HBM. This happens in memory and now very readily happens in logic devices. And therefore, you need quite a bit of process control inside HBM to drive yields up. So that's just one segment, which is a process complexity has increased and variability has increased and number of designs are increasing, and all of that is driving process control intensity. The other thing is some of our processes -- some of our process control tools operate like a process tool. I'll give you an example when EUV was implemented, many of our customers have not implemented pellicles on reticles in a heavy way. So now if you don't have a pellicle, which is kind of like a surround wrap on top of your reticle, inside the wafer fab, you can add a defect on this reticle, which will print on every single wafer, right? So it's pretty catastrophic if you have that situation. So what we developed was a new application on Gen 5 called print check, and that is implemented in all HBM fabs now. And what does this system do is it tells you only the defects that are coming from the reticle. Of course, it can tell you what the defects are due to process, but this particular application is for that. But you can't skip that. So every so many wafers, you have to ensure you check the reticle. And if you don't check the reticle, you will have a catastrophic situation. The cost of the wafers are quite high. And generally, customers are building 100,000 wafer starts per month. You can do the calculation of the cost of the mistake. So a tool like that becomes closer to a process system because you are now making sure that it's in line and it's constantly checking. There are several other examples like that in HBM where process control tools are utilized as process tools, and that's driving high intensity. Of course, in infrastructure world, which is wafer manufacturing and reticle manufacturing, the process control intensity is very high because you don't ship a new brand, new wafer and not inspected on a KLA inspector. So high intensity in that area as well.

Got it.

Yes I guess the only other thing I would add to that is that with the proliferation designs, you're seeing customers also backport technology. What that means is that they're taking -- they may have an issue with the new design and new process iteration, for example, N5 or N7. And we'll have a yield problem with that particular design. And so we'll find our customers that will be buying new capability that they'll be adding -- even new capability may be designed for N3 that they'll be adding into that process flow to do a certain inspection because of some particular challenge related to some new design. So the robust design environment is a factor that when you've got multiple process iterations along with designs, they're going to test design rules in different ways and create some of these yield challenges. And all that is driving customers to inspect more. The final thing is, as you start to move in to larger die, larger die carry intrinsic higher defectivity, statistically, it's that way, larger die, same number of particles, your yields are lower. And so larger die that you have higher defect challenges and that die that are more valuable, like GPU die, for example, customers have a lot more risk given the cycle time of producing those chips. And so the inspection requirements are generally higher as well.

Got it. Makes sense. And the trend toward larger chips and the increase in the heterogeneity of semiconductors would -- it's more of a permanent shift than something transferred.

Yes, I see it's a more permanent shift and all of these are additive steps to make that same chip that you needed before. So therefore, there's more process control. Wafer-to-wafer bonding, which is happening in NAND, it's going to happen in front end in logic. It's happening for sure in heterogeneous integration all of those are driving additional process control steps.

Got it. And maybe double-clicking on a couple of the growth drivers that you guys talked about gate-all-around and backside power delivery in particular. From your vantage point, obviously, you guys see these inflections early. Where is the industry on those two transitions? And how should we think about the opportunity set for your business going forward?

Yes. So I think, again, it's another dimension of process complexity that is increasing. If you want to add more transistor at this point, doing it the FinFET method is not going to work. So customers are going to gate-all-around technology and also backside PDN or backside power delivery systems to ensure that your RC performance is pretty good, which drives wafer-to-wafer bonding. So first, in just adding gate-all-around the number of steps to pattern that transistor is significantly increasing. And we just finished releasing our new version of Gen 4 system, which is a slightly longer wavelength and therefore, it can do buried defects as well as surface defects. So we see a growth in that area, and we had to release a new non-upgradable system because it has far more capability to do these types of layers. So we see that. And also for high-K metal gate and the number of deposition steps is increasing quite a bit. And for that reason, the gate control measurements are going up significantly. So that's just 2 steps out of -- it's 2 steps -- 2 mega steps, right? There's many, many, many loops inside them. So these 2 steps by itself are going to cause growth in 2-nanometer. Of course, wafer-to-wafer bonding because you have to attach to these 2 different wafers, there's growth in overlay, there's growth in inspection to make sure that these wafers come in contact very well, all that growth is there. I think another way to think about it is that process control intensity is increasing. And the other thing you can do is you can look at the cost of a wafer fab, a 3-nanometer wafer fab versus 5-nanometer wafer fab and a 2-nanometer wafer fab, the cost of the wafer fabs are increasing. And of course, that goes to KLA as well.

Got it. A question on e-beam versus optical. This used to be a pretty loud debate. I feel like you guys have sort of settled the debate somewhat. They're complementary technologies. I think you guys are doing well in both. But as of today, where do you stand as it pertains to e-beam versus optical? Any evolution in how you think about those 2 technologies going forward?

Yes. So we have maintained this sentence. I think we started, I don't know, in the 2018 Investor Day or it's the first one we had, I don't know when it was. I always say 2018. So what we said is the ratio of optical to e-beam is going to be around 80-20, meaning 80% of the layers are going to go to optical and 20% of the layers will go to electron beam systems. And if you go back in time and chart this out, and we do that because we have to look at the market back to my strategic planning comment. So we look at very regularly, and we don't see any change, meaning the bulk of the market is really optical inspection. And it's not that optical can and cannot do it. I mean we are the sub-resolution kings of the industry. We are able to not image clearly and find the defect, and this is what KLA does because our algorithm technology and our optics technology is so strong that we're able to look at with 190-nanometer wavelength a defect that is 10-nanometer of size. But that's where all the innovation comes in. So we feel that is the case, and it's going to continue to be the case. If that is not the case, the economics break down because you wouldn't be able to have a fab big enough to store the number of e-beam systems you would need. It just doesn't work. So physics, of course, keeps us all in check. So I think the 80-20 continues. Now we are still investing in e-beam technology because we believe it's a complementary technology to optical. For example, today, the bulk of the e-beam systems are used to verify what optical systems do. So we are building a connection between our electronic beam systems and our optical systems. Of course, someone can say that they can do that as well, but KLA has the widest suite of inspection systems and the number of wavelengths we have is the largest, right? I mean most other companies have single wavelength systems. So because of that, our complementary access between optical and e-beam is quite strong, and that's how we believe the synergy with electron beam comes in the marketplace.

Because the challenge is it's just incredibly slow, right? So there have been lots of aspirations about how to speed it up and go into hundreds and hundreds of beams to try to make it more viable from a throughput basis to cover more of the optical applications. And it's pretty complicated technology and electrons repel and so getting consistent beams across the wafer to be able to get actionable results is hard. So it really is to, Ahmad's point, since again, back to the portfolio approaches is that, look, you want to be able to inspect the full wafer, you want to see signatures across the wafer, you have defects that show up in different places, you can't really do that with an e-beam tool. Certainly you can't do that in production when it's 1,000x slower and take you a week to scan the wafer. It doesn't work for production. So it is one of those where they're complementary. I think there are ways that we're starting to speed up e-beam to make it more attractive. But I think at the end of the day, what you're really looking for in the optical systems is you're looking for contrast. And then you're using that, then you're going to review in a review system to do the review and classification steps. So we're investing in it. We think we can use it to point our inspectors to be a baseline for our inspectors, and we can do that with a network effect across the portfolio that gives us an inherent advantage. That's where you're going to see most of the investment, I think, as it relates to -- from us in terms of EBI, EBR is an established market about reviewing tools, which is e-beam review. And then there are some emerging metrology applications where you're starting to see some opportunities there as well. But when you add up all the dollars, 80% to 85% of them are getting spent on optical technologies because of, I think, these inherent technical trade-offs that customers have to make.

Got it. That's super helpful. Shifting gears, localization is a big theme, and many, many governments are focused on bringing back production, if you will, in semis. Does that impact how you think about the broader strategy, how you think about resource allocation? We often get questions about competition from local regional companies. But outside of the big multinationals that you've competed with for many, many decades, are there any emerging that we should be mostly watching or not so much?

I think the competitive environment has been pretty stable. As you regionalize, it's requiring us to invest. We've got to invest in resources to support customers' application service. We have to have depots. We have to invest in inventory to make sure we have the parts. So you're going to be making those investments. And over time, as those facilities start to scale and they go into a service stream, you start to get leverage on those investments. But from a competitive point of view, I don't think it's really changing the market opportunities. In fact, typically, when you have regionalization, our customers are relying on us more. When they're concentrated, they can deploy their resources in that concentrated environment much more effectively. And so when they're spread more thinly across the world, they're going to need more support, and that tends to translate into higher market share. So I think as a scale player, we bring that expertise. And I think we can help our customers more in that scenario than perhaps where they were concentrated, and they have more confidence in their ability to do some of these things on their own to rely on us to do.

Got it. In the last couple of minutes, we'll take questions from the audience. Can we get mics, please? The second row at the very front. I think -- it's webcast I think.

So when you were talking about the -- and we hear a lot about the architectural and see change happening in data centers, right, and the different -- of course, it's all related to AI or most of it's related to AI. Your leading-edge business didn't grow as much this year. But if we talked to you a year ago, it was like 75% of your revenue or some number like that, right? Now I know it's come down a lot. So is it fair to say that a year ago, everyone was buying for some of these new chips that are getting spun out this year? And just -- I guess this is more of a cycle question. So wouldn't that set up well for '24, '25, if that trend of Gen AI and these changes in data centers continues?

Yes. No, so it's a very good question. And you're right. We saw some announcements in the market of some strength in part of the business. And our customers had the excess capacity to be able to absorb some of that. I think on -- the utilization rate has actually been higher in that part of the market. And I think it's a result of that. So yes, customers have adjusted the spending levels, and we'll see how that ramps. It's still a pretty small percent, although growing faster of the overall investment as mobile has been soft, which tends to drive the bulk of the investment in -- at the leading edge. I think that, that is a -- that's probably the biggest factor in terms of how that market recovers as we move into next year. But if you look at the expected growth, you don't have to go too far in the future where you start to see, okay, driving leading edge, which mobile tends to drive the first introduction of the leading edge at a certain node that you start to look at wafer starts that are commensurate to support this evolving market. So it does create a demand potentially over the next few years of a lot more wafer starts earlier in the node ramp than what we've seen in the past. You've got multiple drivers of that leading edge node. Ahmad, is there more?

No, I agree. I think right now, it's somewhat limited, but it's increasing. And I would say, in 2 to 3 years, there will be multiple drivers for advanced nodes and elements will drive that for sure. And that's just the GPU part and then the HBM part is increasing quite a bit. The complexity in HBM is higher, and there's going to be more EUV implemented there. Then integrating those onto a chip with interposers and also packaging, there's additional growth there. So that's the lower end of the growth, but the higher end would be the GPU and these are very large dies. And that's where KLA does really well because it's very hard to inspect large dies because you can't compare it to another neighboring die because it's a single reticle field. So that's where we do quite well. So we expect that this is going to grow. So I think 2 to 3 years from now, there will be multiple drivers for advanced nodes. There's going to be enough capacity there, yes.

I wanted to hit on so much more services and your capital allocation model, which you hit on, Bren, at the very top. But anything else that you want to highlight before we let you go. Anything that you feel like the market collectively underappreciates about the KLA story, the markets you play in? I think it's pretty straightforward. And you guys are doing really well.

Sure. The company is well positioned. I think from a share point of view, we gained share last year, and so we think we're well positioned for the future. We've got products that are in some of the fastest-growing markets of WFE. We're seeing our share of the market increase. Our financial model is the best in the industry, and our capital allocation is consistent and explicit. So I think that -- I think it's a pretty compelling investment thesis.

Awesome. Really enjoyed the conversation, and thank you so much for coming. Really appreciate it.

Yes, thank you for having us.