KLA Corporation (KLAC) Earnings Call Transcript & Summary

Good morning. We're going to get started. I'm Tim Arcuri. I'm the semiconductor analyst here at UBS and very pleased to have KLA. We have Bren Higgins with KLA. Very happy to have you. Thank you, Bren.

Thank you for having me.

So first topic I wanted to talk about was export controls, and you made a comment yesterday about there being $500 million impact. It sounds like that's incremental to what you said already. You said China would be down 15% year-over-year. It sounds like that sits on top of that. So can you go through some of the dynamics and sort of how you arrive at that number and what it includes?

Sure. So yesterday and based on new export restrictions that dropped on Monday, we provided an update that incremental to our views for 2025. First of all, we reiterated through an 8-K before market, our comfort with the guidance that we provided for the December quarter and affirm that. But as it relates to 2025, incrementally, given our preliminary review of the regulations that we thought it would have roughly a $500 million incremental impact to our calendar '25 expectations. Underneath that, it's about 70% systems, about 30% service. And to your earlier point, so if you look at our overall China business year-to-year, it takes the China down roughly in 2025 about somewhere in the low 20% range based on what we see today. As a percentage of business, the numerators and denominators move around here, but as a percentage of business, we're going to go from roughly 40% in 2024 down to somewhere in the high 20s, I think, in 2025 based on what we know right now. We'll see as we work our way through the regulations, whether anything changes here. There are mitigations around licensing that's potentially a mitigation or reallocation of certain long lead time tools that would mitigate that. We'll see as we go. But for now, as we looked at what was very specific around certain customers, the impact was consistent with that $500 million plus or minus $100 million range that we talked about.

And in terms of service, so you were thinking service would grow sort of low doubles next year, that would bring it down to what, high singles?

Yes, high singles, maybe low doubles. I think we've been trending this year at the higher end of our target range of 12% to 14%. As you take that headwind in the short run, yes, it would pull service back to somewhere in that range.

Got it. And can we just talk about 2025, just in general, what are some of the puts and takes? You haven't given that much specificity really on '25. But I think you still think '25, even after this, you still think '25 is up in terms of WFE?

I do. I do. I think that there's probably a few billion dollars of WFE that comes out related to these new regulations. And 2024 WFE looks like it's somewhere in the high 90s. And so I would expect some modest growth into next year. Obviously, China is down. But one of the things that we're pretty excited about what drives KLA's business really over time is expectations of reinvestment or new investment at the leading edge. And so we're excited about what we're seeing around the 2-nanometer node, both in terms of customer interest, our customers' customers' views in terms of designs and potential design starts, high leading-edge memory is also something that we're excited about, mostly driven by high bandwidth memory which is good for process control. And then finally, advanced packaging that is being driven by high-performance compute and what that is doing in terms of the integration of this very high end, either GPUs or custom silicon and how it integrates with the memory requirements in advanced packaging. So those are all very good drivers. The relevancy, I think, of KLA is also increasing given the design start environment, the tight capacity of 3-nanometer that we see our share of WFE increasing at N2. We saw it increased at N3. But at N2, we see it increasing meaningfully, I've talked publicly about at least a 75 basis point increase in daily share of market. Part of it is relevancy overall given the nature of the transition with new architecture, but also share position as well. So we feel pretty good about what's in front of us there. So I think that, that offsets. I think the rest of the legacy market is more or less flat, doesn't change much. I think it offsets what we see in terms of the impact in China, and we end up with a year that's, we'll call it, mid- to high single-digit kind of growth given expectations for WFE being modestly up.

And you think -- so you think you can hold WFE share next year even despite there's some China headwinds, you did very well this year in China. NAND is going to be a little better next year. But do you think that the power of that leading-edge foundry/logic is enough to allow you to hold or maybe even gain, I don't know, WFE share next year?

I think we do. I think there's the possibility to gain. I don't think we're going to gain much, but we'll call it 20 bps or something like that. Well, today, this year, it looks like it's somewhere around 7.6% depending on your denominator for WFE. But I think it probably ticks up a little bit despite what is likely a mix shift, a slight mix shift to memory, right? Where logic/foundry goes from, we'll call it, 70% -- low 70% or somewhere in the high 60% of the total mix of WFE. So normally, that would be a bit of a headwind for process control because relative process control intensity is higher in logic/foundry than it is in memory. But given what we're seeing, both in terms of the leading-edge focus on leading edge in logic, but also the strength in high-bandwidth memory that we think that, that will compensate for the mix impact, and we're in really good position for, I think, a modest outperform as we look at next year.

And then how do you think about WFE intensity in memory? I know, like I used to think of it as your capture rate per dollar is 50% higher in foundry/logic than it is in memory. Now that's changing a bit because you are -- you've come out with some new technologies in memory and your capture rate's a little higher, I think, in memory than it used to be. So can you talk about sort of how you see the gap between your process control and intensity and foundry/logic versus in memory, both...

Yes, it's been getting better in memory, but there's still a delta there, right? So a high mix leading-edge foundry would be more mid- to high teens as a percent of WFE. And we'll call it DRAM as somewhere low double digit, so we'll call it 10% or 11%. And look at Flash, it's probably somewhere in that 9% to 10% range. So hasn't -- we haven't seen a lot of new investment in the Flash market. So I think that we'll see if that changes. I think as it relates to memory, though, particularly with the introduction of EUV into DRAM which was a driver for our business, but then also the nature of a high-bandwidth memory device, both in terms of the size of the die, you have to drill the TSVs, the logic circuitry in the die, the reliability requirements of them because of the size and less redundancy, which typically you have more redundancy, particularly in DRAM, that all those have been good factors for process control intensity. And so we're pretty excited about that. And of course, then that goes into the leading edge logic piece, into the packaging piece, which is also something that's been really strong for us, right? We're about $300 million in packaging revenue in 2023, it's greater than $500 million in 2024. And I think in 2025 right now, increasing momentum, somewhere around $750 million. And it seems like each month, I keep taking that number up a little bit just because of the strength of what we're seeing in terms of demand, particularly on the logic side and CoWoS by our major foundry customer. And their capacity requirements but also share momentum as well. So it's a pretty positive story. And I think over time, as more capabilities required in that part of the market that it does start to drive more high-end tools, which carry typically higher margins and typically, tools that there's a trade-off between the sensitivity tool and the speed of the tool. So as you add more capability, it tends to slow the tool down. So for the customer to keep similar sampling rates, then there's an ASP component but potentially volume component, too. So we're pretty excited about what that looks like moving forward.

But you still think that rough math, every dollar of foundry/logic is -- spending is 50% better for you than a dollar of memory spending. That's still just the 50%.

I don't know about...

But it's -- we'll call it, if it's mid-teens versus like 10% or 11%...

Yes, okay. So somewhere in that ballpark.

Okay. Got it. Okay. And then can we just talk about China. So my contention is that there's going to be some forced consolidation of a lot of this mature node spending, a lot of these mature node projects. Some of it government-led, some of it's natural business forces. Do you see signs of this happening that this litany of all these new entities in China that we don't even know who half them are that there's going to be some consolidation of those entities that could further pressure the market even as you look out into '26, when we're trying to figure out what like steady state is for China WFE and what your China business is.

Yes. China has been elevated for some particular reasons, right, in the business. If you go back to '21 and '22, we booked a lot of business that given the supply constraints at the time, we had to deprioritize given the nature of the demand from our other non-China customers. And so because a lot of those projects were greenfield, they weren't really being driven by the supply-demand dynamics in the market. So as the market corrected into 2023, we saw -- we were able to pull a lot of that business forward. The funding was there. These were new projects that we're doing R&D or risk production, so not really being driven by general market dynamics. So that really provided a nice cushion through this period of time at an elevated level, right, over the last couple of years. I think there was also a concern around, a, there's the concern around restrictions, which I think drove some buying activity. I think there's also concerns around, okay, we were deprioritized once, we could be deprioritized again. And so I think that there was a general rush to take systems. And it provided a nice cushion to what could have been a certainly more challenging downturn from an operating point of view. But as we see this year, we had planned even before the export restrictions that we would start to see some digestion in that spending. And so it was going to come down from 40s into the 30s. And I think the long-term trajectory has it coming down into the low 20s or maybe 20%. I think this normalization, another way of talking about what you said, a rationalization of capacity, you can only invest ahead of revenue for so long. And that over time, next rounds of funding would be either driven by business requirements or meeting technical objectives. And so I think any time there's strategic investment happening like that, there's a rationalization that likely comes. I'm not saying -- I don't think we're seeing any of it yet. But our view of the long term is that they would have to rationalize. And so if you back up and look at our view of the long run, right, at KLA, we say, look, there's an expectation for semiconductor revenue growth. That's about end markets, not necessarily about what's happening in China or other places. There's a view on capital intensity. And our view on capital intensity is it's not going to rise a lot, but it should -- it's been rising, it should be slightly better than semi revenue. So it should allow the WFE market to grow faster than semi revenue. Given the growth rates in semi revenue that are tied to logic and foundry, that should drive a higher mix to logic and foundry, which is, as we talked about, more process control intensive. So that creates an opportunity for process control to grow a little bit faster than that. And then if you look at what's driving process control from a market point of view, we have the strongest share in the markets that are growing the fastest. So that creates an opportunity for KLA to grow faster than that. So that's our overall view, and that takes you into this kind of -- if you start with, we'll call it, 7% revenue growth for semiconductor revenue. And then that puts KLA somewhere in that 9% to 10% from a systems point of view. We feel pretty good about that trajectory. Now look, if there's more inefficiency in the market. That's fine, whether it's driven by regionalization, whether it's driven by strategic investment. We've proven we know how to ramp the company to be able to take advantage of that. But how we're going to drive the business size it, make investment decisions and think about the market opportunity is more consistent with that framework. Now the service business is going to continue to grow. You'll take a headwind from these kinds of restrictions. But over time, if we're shipping these systems, we still feel very good about the target range for service of long term at 12% to 14%. And we have some other markets we're in, where I don't think they'll grow as fast as the semi business, but could be kind of mid, high single-digit growth. And so when you add it all up, we think it's a pretty compelling story that translates into our current public view of our top line of 9% to 11% revenue growth over time. And of course, the financial model, which is one of the best in the industry, the operating leverage on that at 40% to 50% operating margins and the ability to grow earnings per share at 1.5x the revenue growth rate is our plan. So that's the long-term view. Of course, all these dynamics in the short term, we'll manage our way through. But in the long run, that's how we're thinking about the company and the opportunity in front of us.

Great. Can we talk about gross margin? And it sort of struggled to get much above 62% recently. So can you talk about some of the puts and takes. I think China is for not just you but for everybody, a bit higher margin. So that's a headwind coming down. That's a mix headwind. Service is a little lower gross margin for you, I believe. So that's -- maybe that coming down a little bit as a better offset for you. But when you look at the puts and takes, it seems to be probably a little challenging for you to get gross margin much above 62%. Can you just talk about sort of how you can push gross margin up?

Yes. 62% is pretty good margin.

It is, which is great.

Particularly for a hardware company. So yes, we've been operating in this 61% to 62% range for a while. I think our margins are more impacted by what we sell versus who we sell to. And when you're a value seller about where your pricing is really based on the value you offer to customers the return on investment in process control, then you really are fairly careful around pricing because it's not a cost-plus mark. And there's a fair amount of transparency around pricing. There's volume-driven pricing. That's natural. But outside of that, our pricing is fairly consistent. Our customers all have different strategies in terms of how they adopt and how they inspect. So that will drive -- we sell a portfolio of products. They don't all have the same margins. Advanced packaging being an example where advanced packaging margins are -- while the gross margin dollar, which we are optimizing for more than the percentage in the company is higher. The actual gross margin per tool is lower than some of our higher-end systems. We also made a number of investments in our own capacity and supply chain capacity over the last few years to position the company for the growth rate implied in the plan that I talked about earlier. And until we start to see, I think, what is kind of a headwind today, we start to see a ramp in the business. If you look at our '26 plan, we have a $14 billion target in our '26 plan presuming WFE has significant growth. And in '26, we really have built the company to be able to achieve that plan. And so I think what is a bit of a headwind today turns into a tailwind as we start to move forward. But our incremental margins are generally between 60% and 65% at the industry scaling above sort of a normalized growth rate. We tend to be at the higher end of that range. If not, then we're at the lower end of that range. But I don't see that changing all that much. So at $14 billion, we had a 63% target. I feel like that's reasonable to think about. But I think given the incremental in our business and where we're operating from, there's opportunities to improve it, but we're talking about basis points of improvement over time, right?

Yes. I guess I asked because one of your peers that doesn't compete with you in all your [ tool types ], but the other major player in process control, they've been talking about pushing pricing higher and pushing gross margin higher. So I just wonder, I mean, if there's anybody that can push pricing higher and sort of set the margin, it's you. So I kind of wonder if there's some thought of well, maybe there is an opportunity for us to sort of reset pricing and margins a bit higher. Why can't you do 65% or 66% because the customer doesn't have any alternative in most cases.

Well, but it also comes down. The alternative could be nonconsumption, right? Because it is -- we don't actually make chips, right? So I think that there's a return assessment of the value to the customer. And at some point, there's probably a lid on pricing because it's like, okay, how is that return assessment impacted by that pricing? And if it turns out that, you know what, I'll take more risk. I'll use a less sensitive tool. I think customers will think through that. So there is a bit of a limit. We do have a lot of discipline. It's also not really an elastic business. They don't typically buy more because the price is lower, right? And so you have to have a lot of discipline around understanding that because they're not going to just necessarily discount it. So I'll buy a second one. The other thing is that the baseline for pricing as you think about the next generation is the previous negotiation for that tool type. And the next generation at a baseline level of capability, there's a cost of ownership improvement. And we share -- typically try to share that cost of ownership improvement with the customer as we move to the next node. So if you make a bad deal, that could be the gift that keeps on giving for a long time because that's the baseline for future negotiation. But that's generally how our pricing model works, and I think that the discipline about understanding that is fundamental to the company. It's one of the things that we really believe that we add in the KLA operating model when we buy companies is that a deep understanding of understanding return and thinking about return and how we price. We think it's fundamental to our go-to-market. And when we buy other companies, we see opportunities to understand that and to optimize that better. And so I think we do a pretty good job of it, and I think the gross margins reflect not only finding that right balance to optimize for gross margin dollar, but also ensuring that we're delivering the innovation and differentiated capability that our customers are looking for. And the sure sign of differentiation is to look at gross margin, right? And given our gross margins, I think it shows why our share is what it is and the value that we think we bring.

Bren, I've always thought of KLA, and I've covered KLA for 26 years. So I've always thought of KLA as much of a software company as exists in the semiconductor equipment world, I sort of think of you as maybe even more of a software company. I mean, yes, you sell hardware, you don't sell software. But your secret sauce really is in the software as much as it is in the hardware. So we don't talk about that as much. So I'm kind of wondering if you can talk about just from a value proposition for people that don't know the -- that don't understand the like software piece of your story. Maybe if you can talk about that.

Yes. No, it's a good question. And while there's a lot of unique capability in the systems and the optics and the sensors that -- it's the data that they generate and how we manage the data that is really fundamental to KLA's value proposition. Nobody generates more data in the fab than KLA, right? And so -- and one of the biggest challenges for our customers is managing and trying to understand signal to noise. And so we invest a lot and most of our engineers are software engineers, we've been doing physics-based AI for years to be able to -- and given the cost of our systems and the computing requirements, we could charge for it. But to be able to help our customers speed because really what we do is you're improving yield, but you're also improving yield learning cycles and time to result as a customer scales a $20 billion fab. So that -- and when you look at the capability of the hardware today, you find everything, right? So there's a lot of noise. And so trying to find the signal to noise, trying to vary how inspections work relative to a design that to target specific what we call care areas on a device where the topography may be denser than in other areas. These are all things that we do to try to ultimately find defects that matter and help speed those yield learning cycles. So it's fundamental to what we do, and it works, I think, part and parcel with the hardware capability that is also, I think, very unique. We do sell stand-alone software also. And within the business, which you're right, I don't talk about it very much, but we have about $175 million of stand-alone software revenue that has effectively doubled in the last 4 or 5 years, probably a contributor of 50 bps of margin for the company, gross margin for the company over that timeframe, where we're allowing and selling whether it's data analytic capabilities or simulation capabilities for our customers to try to leverage that data in more effective ways. So there's how the systems actually perform themselves, but then also some of the other software packages that help try to leverage that information is also a focus for the company. And I think that continues to grow. I think customers used to do a lot of this on their own, and I think they're looking for us and understanding the interoperability across the portfolio of tools which has a network effect of collecting data that is pretty important to our competitive position that allows us to really collate that data and think about how that might impact process or on device performance.

Yes. I guess I ask you because NVIDIA has this product cuLitho and they're working with ASML and TSMC and are using AI to basically predict what -- before you ever start the design to predict how the design is going to actually map on the wafer. And this is a massive opportunity for you. So I would think that there's much more of an AI opportunity for you where you can bring your software leverage to bear, maybe work with NVIDIA, I don't know. But rather than kind of the old world was, okay, let's produce the wafer, let's look at it, let's see the defects and figure out what was the problem and which knob do we have to go turn and it was more of like an iterative process back and forth. But if we can -- if you can have products that make it so that you don't even have to start the design before you fully simulate it, and you know exactly where the defects will be, you know exactly how it's going to tape out.

Yes, there's a lot of that's actually still -- that's happening today. And we are a partner, given our image computing requirements, a partner working very closely with NVIDIA. In fact, we've moved our architecture and our most advanced systems from a CPU-based architecture to a GPU-based architecture that's created significant cost benefits. So we're big believers in it as a driver given our workloads. And maybe similar workloads and a high-end AI server. But the robustness, and we do the software work to change the architecture, which was engineering within the company. But now we've got a much more robust algorithm environment that allows us to do a lot of things I talked about earlier. And then of course, we're also using AI capabilities to try to speed some of our time to market and code development as well. So we have a number of use cases there that are very interesting and we're going to take to our other systems that drive not only cost but capability and productivity. But what you're talking about is exactly what you're doing more of today is how can you, before you start wafers, make investments? Can you leverage the information you have? Can you do simulations? And we've been doing a lot of that. I think that's a big part of what's been driving that part of the business and the growth that we've seen in that in our software revenue.

Great. Can we talk a little bit about the backlog and the RPO. Because just another thing that's changed also with the company in the past 3 years, 4 years, really since COVID, I always think of KLA. KLA always had 5 months of systems backlog, give or take, and now you have doubled that, double plus. And so we always have this conversation, is that cyclical? Or is that secular? Is there like something different in the business? And is it you're running the business differently? Or is it your customers are planning their capacity differently? So can you talk about that? Is that like a permanent thing where you think we should expect you to have 10, 11 months worth of systems RPO? And is that really -- should that give us a sense of confidence that you have even more visibility than you used to have? Or is that just sort of like out there? And maybe it happens, maybe it doesn't?

Well, so there's 2 parts of it. I think one part is cyclical that was unique to the timeframe, particularly as it relates to all the greenfield investment related to China and the supply chain constraints, which drove it to a higher level. So that's part of it. But I think from a secular point of view, there's also some changes that are also happening. If you're a customer and you're building a fab and you have a need for equipment 1.5 years or 2 years down the road, it's in your best interest to give us orders so we can plan around that and plan to be able to deliver when you want those systems. And because the last thing you want to do is finish that project and you're waiting for tools. And of course, the customer is trying to coordinate across all the different providers to make sure they can align that when I open that fab, I have what I need. And so when you had a lot of greenfield, particularly from customers that were new customers, for example, in China, that they are more than willing to give orders. We also took deposits because it didn't have a long-standing relationship, so it was important to secure that business with deposits. And of course, when you ship that business, you ship under letters of credit. So I think managing the risk, but also have the certainty that -- and the visibility to what's required. So that was certainly a factor. The second factor is that we do have certain products that have very strong demand. And are fundamental to not just production but also to R&D. And so our optical inspection products, our Gen 4 product in particular, had lead time -- has lead times today of 18 to 24 months. And so if you're -- any of our customers who want those systems and want them in a certain delivery window, you need to get in line and we want the order so we can plan around it. I think the third factor, which I think is more secular is that our biggest customers weren't always great at giving us orders, and we would do a lot with forecasts. And of course, we have very close relationships. And so it worked. But I think once we went through a period of short supply, I think they got some religion around, hey, look, the visibility is important. And so those customers are now giving us more orders than they used to give us. In the past, historically, I think they've just come in and sort of force what they wanted. But then if we give one customer all the capacity, then we shortchange other parts of the industry and then ultimately, everything slows down. So we have to try to manage it very broadly. I think -- so that part, I think, is secular. I think we're going to get more visibility so we can plan. And I'm doing the same thing in my business with my own supply chain because lead times are great until things are under strain. And when things are under strain, then all of a sudden they extend out. So as much visibility you can give to ensure that we're prepared and we can meet their needs, I think there are some merits to that. And so I think customers are likely to, I think, continue with that. Now I think the piece that was the first part, that's coming down because I think that part of the business is in this digestion phase. But we're not going to be at 6 months anymore. I think we're closer to maybe it's 10 months plus or minus moving forward. And I think from a planning and predictability point of view, I feel pretty good about taking those orders and ensuring that it allows me to plan capacity, I run manufacturing in the company also. So being able to have that visibility is important.

Great. Maybe just one quick last question. So there are some pretty big stimulus in Japan from the government. I mean, big, big numbers. There's a big new foundry in Japan that we can debate whether they'll be successful or not, but they're going to begin to be impactful for your business even next year. So how do you sort of like think about the emergence of -- we think of the chipset all the time, but either the Japanese government is doing the same thing, even like bigger dollars.

Yes. So look, regionalization, I'm generally a believer that over time, the industry has done a decent job of balancing supply/demand. Now sometimes because of the lead times and what it takes to build these projects, they'll overshoot, but then they correct and that's why we have the cycles we have. So even though I expect a lot of structural growth moving forward, I think the cycles, as I remind people inside the company, it's cyclical growth. And so you're still going to have some of that. Now regionalization, so I think the regionalization dynamic creates really an opportunity for us more around how do we support those customers and it drives opportunities within the service business because our customers are spread more broadly in terms of their footprint. I think it creates an opportunity for us from a share point of view because I think customers are relying on us more to help manage their ramp. And so I think, in general, managing the broader footprint can be -- will be a positive. Now you have to invest ahead in that, which can be a problem, particularly when these projects slowdown or delay because of training cycles around service people. We've got to position parts to support. So in general, I think for a scale provider, it's an opportunity, and we'll see if it creates some permanent inefficiency in the market that translates into more systems. But I think in the long run, I think that it likely stays relatively balanced and customers will rely on us more to ramp a broader footprint.

Great. Thank you, Bren. Appreciate it.

Yes. Thanks for having me.