Hi. Welcome back everybody. Again, I'm Joe Moore. Happy to have with us today the management team of KLA, Rick Wallace and Bren Higgins. Thanks, guys.
Thanks for having us, Joe.
Maybe we could just start out with you know, big picture look at this. You know, I guess this time last year I thought we'd be looking at WFE last year of like mid-$60s billion, ended up more like mid-$80s billion, you know, now we're talking about $100 billion. One of your competitors at this conference last year actually sort of threw out a $100 billion WFE number, and it was so contentious that they kind of walked it back, and then here we are a year later, that's the number. You know, maybe just talk to that. Are you guys surprised at how strong that is and, you know, how sustainable do you think that $100 billion level is?
Well, thanks for having us, Joe, and thanks for such an easy first question. Look, I think that a year ago we didn't. You know, you go back two years and we were questioning whether or not the market was gonna tip over, right? At the beginning of
Mm-hmm
... of coronavirus. I think that what happened in the last year, of course, is digital transformation happened. All these drivers we were talking about in our Analyst Day in 2019 were there, and they were all starting to happen, and then they got accelerated through COVID. I think part of what we started to see was a lot of the end market demand was driving our customers, and then the profitability of all our customers, when we look at their expenses, their CapEx investment relative to their revenue, for the first time I can remember, they're all making money. It's all kind of rational from that standpoint.
Mm-hmm
With a couple of exceptions. I think there are some projects in China that are not really driven by the revenue that they're generating, right? Those are strategic investments. By and large, when you look at most of the spend of WFE, it is rationalized, and you have the player that's trying to regain technological lead investing at a higher level, right, with Intel. I think it kind of makes sense that we're seeing these levels and we've got a huge amount of pent-up demand now because the supply chain has extended it. You know, part of what we're seeing, I think the sustainability of it, on the second part of your question, I don't think there's an equipment company out there that's gonna be able to ship what we would like to ship this year.
Yeah.
We can't get the. There's plenty of customers to take it. I think, yeah, I think that we're in good shape relative to the growth of the industry, relative semiconductor growth, and the capital equipment, and I think for WFE, it's supportive of that. Yeah, of course, we were surprised. I mean, three years ago when we put out our model, we didn't have this big step up.
Yeah
... in terms of overall investment. You know, I look at the profitability of our customers kind of, it's supportive.
Yeah. Makes a lot of sense. You mentioned the supply chain, and it's kind of interesting you guys have so much higher semiconductor content in your tools than everybody else, and yet you actually kind of have been the cleanest on supply chain until, you know, it did kind of start catching up to you a little bit. Can you just talk about where the bottlenecks are for you? You know, you mentioned nobody's gonna get what they need for the whole year. You know, that, I would think that when you're selling tools that are this expensive, you can kind of force your way to the front of the line and get the pieces that you need to kind of keep that supply chain in order. Can you just talk about that?
I'll start and then give it to Bren, who does the heavy lifting on this as ops reports into Bren. I think that the biggest overall for KLA, because we have a high mix, low volume in general, and we have a lot of unique kind of parts that we buy, whether it's optics or other electronics, we have pretty good handle on the supply, so we're not buying commodities, so we're not typically competing as broadly. But in the case where the few things we've gotten bitten on are very low volume and not particularly specialty. Yeah, we had a couple, I would say pretty minor on a relative basis.
I think that the other thing that happened in the first quarter and even a little bit at the end of last year is I do think that Omicron kind of screwed up some of the supply chain, too. Suddenly you had a number of people out-
Mm-hmm
that were, you know, they were fine, but they were out for a couple of weeks and our suppliers, and so that was what we think kind of a hiccup. I think in aggregate, you know, it is absolutely more because of the increased level of business than it is necessarily because of the challenges from the pandemic.
Yeah. To your last question, right? If we had thought it was gonna happen, we would certainly had more parts. That's true. Your question, so around semiconductor content that's driving our image processing requirement, that's been less of an issue, and that tends to be more leading edge centric. Our challenges as it relates to semi, just in general, tends to be around commodity parts are in demand by lots of industries, right? Even beyond what our requirements are. In some cases, getting visibility to that when it's deep into the bill of material is not always clear or it's in a subsystem that's a supplier's supplier build of material. When we have visibility to those issues, we tend to be able to go work them and deal with them.
It was certainly an issue that affected our guidance into the March quarter. As Rick said, we have certain components where we have very clear relationships with these suppliers. They have a certain amount of capacity, that capacity drives a certain amount of output, and it's fairly predictable. In those cases, we just have a shortfall relative to demand. We're trying to manage our way through that. We have a second group of suppliers that tend to be single source for lots of competitive or strategic reasons, and they're just feeling the pressure, like Omicron was a factor, where they're running these facilities 24/7. They're running their equipment. They're running up to legal limits as far as overtime goes with people. If you have a disruption, it just slows everything down.
Eventually, we think they catch up, but it does take some time for that to happen.
There was a mix of those last couple of areas that affected the guidance into March. That being said, while it's still challenging, we think that the way we saw things are just generally still intact, and we feel pretty good about about the guidance we've given in terms of March being probably the low with sequential growth through the year, and that the overall year for the company was probably somewhere in the neighborhood of + 20%.
Yeah.
Still managing the issues. I wouldn't say it's necessarily, you know, getting easier. It's just, you know, we're battling our way through it, but we did some, you know, risk adjustment in terms of the guidance we provided, and we feel pretty comfortable with that.
The big things, the ASICs, microprocessors, DRAM, stuff like that, you're basically okay. It's the smaller raw materials.
It's down the list.
Okay.
versus some of these other challenges.
Yeah, yeah. Okay. The question of, you know, remaining constrained through the whole year, you know, and every equipment company has said that, so, you know, it's consistent. You know, I guess it is surprising to me, just given the multiplier of, you know, that a $5 component costs you so much revenue, that presumably, if there's any kind of auction situation, you know, you win that auction.
No, yeah. I correct, but that some of it's just timing.
Yeah.
Right? If you're short and it takes you. This is what the hiccup was. It took a little while. The constraint is not on those. The constraint for the rest of the year is these other things, the long lead items, let's say optics, where we've dramatically ramped up. You know, we'll satisfy Bren's plan. We have that capacity for that, but beyond that, if you could do more, would customers want it? Of course. We're constrained by literally the build time on some of this optics.
Mm-hmm.
which has been ramped up pretty dramatically from two years ago. The fastest growing product line, we think $1 billion more in WFE, is our broadband plasma inspection systems, and those are maxed out. Those were sold out through the year, and that has grown dramatically in the last three years, that product line.
Yeah.
Yeah, one of the fastest-growing markets, if not the fastest, in WFE. One of the differences about this, you know, the last couple of years is we've always had, you know, steep slopes in our business, and it's happened, and there's always been then a follow-up. I think one of the things here in this case is we've had sustained growth over a long period of time. The duration's unlike anything we've seen. Where you have a fair amount of strategic buffer in the system, whether it's long lead time material or your suppliers, or it's or they're carrying long lead time or we're carrying it ourselves, we've worked through that over the last couple of years.
In some cases, our suppliers are shipping directly and we're getting these parts and we're waiting for them in the build cycle, and they're going into tools. It does make us a little bit more dependent on not just the predictability of timing, but of the quantity and the volume. In some cases, you know, you do get surprised here and there. It's a battle, but like I said, I think we assessed it and feel pretty good with our assessment.
Yeah. I mean, I think you've navigated it really well under the circumstances. Maybe you talked a little bit to the geopolitical rationale for some of the spending in China and also domestically. You also said that it's backed up by profitability for the most part. Can you just talk a little bit about those drivers? You know, do you see that persisting, where there's, you know, we have the CHIPS Act in the U.S., equivalent in Europe, obviously China has their own incentives. You know, do you kind of continue to see that as a growth driver? And separately, any risk of export controls that could end up disrupting any of that?
Sure. I think that the regionalization effort that's going on, that's been talked about many times publicly by a lot of players, I think the U.S. wants to get, as you know, above the 13% that we're roughly at, and maybe more towards something like 30%. I don't think it's additional capacity, it's just where the next capacity is put, right? I don't think it's that there's suddenly gonna be overbuild because of that. It's just gonna be geographically managed so that there's a balancing. TSMC, for example, has announced, obviously, Arizona, and they just announced Japan, and so they're gonna spread their footprint out too. Similar with Intel and what's going on both here and in Europe.
I think you're gonna see it just back to this, whatever the WFE, it just gets reallocated more broadly and not so concentrated just in Taiwan. China's got a different plan. China's well under the plan they talked about a few years ago, but still pretty significant part of the overall WFE. Most of that is, almost all of that is now trailing. Export control is not such an issue when we talk about the nodes that they're building there. Even a lot of the growth on some of the, you know, more mature nodes of technology, that's. We don't believe that's going to be, you know, what we've heard so far is that's not the real focus. I think it's on the leading edge.
I don't think that those factors. I think that if anything, what COVID did is reminded people of the critical nature of supply chain having, and especially as it hit other markets, automotive may be the best documented, how much they lost because they didn't have semiconductors. I think you're gonna see that play out. I don't think it overbuilds, and I also don't think it is at particular risk. I think from a customer standpoint support, it's also not particularly hard because
Yeah.
We have plenty of lead time. We can hire and it's really only service people and some apps people that we need to support.
It seems like politically, export controls would be rough right now too, because we are dealing with a supply chain crisis on foundry capacity.
You know, we have conversations. We feel like the way it's being managed right now is around leading edge.
Yeah.
Some of those actions have been taken.
Okay. Maybe if you could talk about process control. I mean, you've continued to outpace WFE overall. I guess some of that is intuitive to me, where when people are making investments in 3 nm and beyond, that there's gonna be quite a bit of process control. You've also done really well in trailing edge nodes, which, you know, I have been pleasantly surprised by. Can you just talk generally to what you're seeing in process control nodes with WFE now?
Yeah. I think, a couple things that maybe we were hoping for but certainly didn't foresee that helped, drive that dynamic. Number one of the things that we pay a lot of attention to because it impacts process control so much, is the number of designs at particular nodes, the number of unique designs.
Mm.
Remember, process control benefits the most in a high mix, high change environment. If you're a manufacturing person, what you want is to build the same thing and it never change, right? I was in once upon a time there, and then I can dial my process. Process control is really necessary when I'm building a lot of different devices and a lot of different process flows because I can't, I don't wanna overbuild and I don't wanna under-build per lot. Those are places where they tend to have the highest process control. A couple of things have happened in the last few years. Number one, we've had a resumption of scaling because of EUV. The cost of design has come down enough that there's been a dramatic increase in the number of designs. Just one example.
28 nm was the mega design node. Still there were designs, even recently there have been designs. 7 nm has now exceeded 28 nm in the number of designs.
Mm.
Over 500 designs at 7 nm, and there were some 400 at 28 nm. That means remember traditional wisdom a few years ago.
Yeah
... with every node there were gonna be fewer, and by the time we got to 3 nm, there were gonna be 3 designs left or something, right?
It's not cheap at 7 nm.
It's less expensive than it was.
Yeah.
Certainly more than it was foreseen to be. You have over 500, and we also already see a huge number at the next node, at 5 nm. What that drives, it drives a bunch of aspects of our business. It drives the number of reticles that are being built to support that. It drives change in the process because, you know, if you go on TSMC's website, I think it's still true, you can pick your design flow, and there's lots of different design flows. That's really good for process control.
Mm-hmm.
The other players that are trying to be in the foundry business that also wanna provide lots of flexibility in designs, this is a driver for process control as well. Scaling has also been a big driver. Back to the BBP or broadband plasma, part of what happened there is we anticipated that Gen 5 was really gonna grow, we talked about that a few years ago, with EUV. It started growing before EUV, and it has been growing dramatically. What was amazing for us in 2021 is Gen 4 on a revenue basis was bigger than Gen 5. We're still seeing, you know, this broadening across all of these advanced designs. You have scaling resumed, you have more designs, and you have more people on the leading edge as a result.
WFE, as you know, has gone up pretty dramatically in foundry logic the last few years. The trailing edge, some of these 28 nm designs are trailing edge. You know, part of what's happened is a lot of what's happened in the trailing edge because of automotive and others. We're seeing two things that are happening that are good for KLA. One is more focus on reliability because of markets like automotive. We talked about that at our Analyst Day a few years ago. Also again, there are more capacity constraints and more advanced designs for them, right? 28 nm in an old fab that wasn't doing 28 nm is good for KLA. Bren, am I missing anything on that?
No. Well, given the overall demand in the environment, particularly at those nodes, the ability to go out and get old equipment and used equipment has dried up.
Mm-hmm.
You see demand for new equipment from a lot of those suppliers or customers. Even to the point where we're restarting some of the old lines-
Right
where you know they're old products, but at the same time, you know, it meets some of these requirements. That's driving some operational challenges for us in terms of redesigning components in these parts. At the same time, they're having to do new investment. The other thing on the leading edge is as you have that design proliferation, what ends up happening is that capacity doesn't get moved to the next node. At 5 nm, for example, the 7 nm wafer starts are still roughly flat, maybe actually increasing a little bit. As customers add 5 nm, they're not reusing any of the 7 nm.
What we saw with the limited number of designs at 20 nm and 14 nm/16 nm, you didn't have a technical driver, the delays in EUV, to buy new equipment, and you didn't have end markets driving the follow-on needs to those nodes, so that capacity was migrated. Today, they're having to equip the fabs fully and have a technical challenge along the way with EUV that presents new challenges from a process point of view. That factor was a big factor in terms of how it's affected, I think, the overall WFE and foundry logic, but certainly KLA's business.
To the extent, I mean, I don't wanna put you in an uncomfortable position of, you know, talking about any one of your customers.
Thank you.
I'll try anyway.
Okay.
We've made the case that you're one of the bigger beneficiaries of Intel's resurgence. You know, I guess, can you just talk generally about when a company accelerates the process migration as rapidly as they are that, you know, what happens to the process control intensity? You know, Intel used to be a very large customer for you guys. It's become somewhat smaller. You know, again, I don't wanna make you uncomfortable about that customer, but the process control intensity roadmap seems pretty clear.
I think that for anyone that wants to. You know, we used to talk about Moore's Law being the doubling of transistors in 18 months and the you know design rule shrink. Now it's really, it's harder to know what it means, but it's process evolution for sure. The reason it's harder, by the way, is once we went to FinFET, scaling is a different concept altogether, right? Because of the way the transistor is laid out. Anybody who wants to progress have to solve, I used to use this example. You know, let's say, if I wanna bring up a new node in a fab, I have to solve 10,000 problems. I have to solve all kinds of problems about integration, and I need data to do that.
What KLA really provides often is the insight around what's going wrong in that process. Somebody once likened it to a debug tool. It's just I gotta keep looking all with our equipment to understand why is this node not working. I wanna have the maximum amount of insight, and a lot of our metrology and inspection tools provide that. What happens is, as people are progressing the nodes, if they wanna do it quickly, they need more of that. Part of what we benefited from. They often front-end load it because, you know, I want at the beginning to understand, how do I design my process? The thing that a lot of people don't recognize is the process margins with every subsequent advance in technology have gone down. In other words, there's just not as much leeway.
That's why I have to measure more things. The number of points that are being measured, you know, 25 years ago, you would measure five points on a wafer to see registration. Today, it's literally thousands of points are measured because I'm tuning this process so much. That's where we get it. If you're taking longer in your node migration, you don't need as much equipment because you're taking longer to do it. If you wanna accelerate it, and this is why we get pushed, Bren gets pushed so hard on deliveries for some of our advanced tools, because not only do they need them, they need them at the front end. The answer is EUV. If anybody's adopting EUV, that's been a big driver for us.
Mm-hmm.
Because suddenly there's a whole new set of problems that they didn't anticipate that they've got to debug. Later, that's in the development and the pilot stage. In production, this stuff comes and goes, and we've had customers where they had a resist change, and suddenly they started losing wafers because then they realized they had to put more monitor points. That's what happens to us in capacity. Anybody who wants to accelerate their nodes, they're gonna need to figure out how to understand what the problems that they're trying to solve are. That's what drives the capacity, and Intel has said they wanna accelerate their nodes.
Yeah. I appreciate that. Then, excuse me. The other thing we hear a lot from several of your customers and several fabless customers is the importance of advanced packaging. I feel like the acquisition of Orbotech kind of becomes clearer as every sort of semiconductor CEO talks about how what a critical advantage that is with very complex multi-chip packages and tiling and through-silicon vias and things like that. Can you talk a little bit about that business for you guys, and how are you gonna be able to drive the growth in that business going forward?
Right. It was the driver. I mean, a large part of that driver was associated with more than Moore, so a little bit to the packaging, but also frankly, power, automotives and some of the work that we do in the SPTS division there. Our view, because we had the ICOS business, we've always had a view a little bit in the packaging. ICOS kind of opened up our view that there was gonna be more work going on. Even though you have a resumption of scaling, the combination for every one of the major semiconductor companies is now about packaging, in addition, and some of the CapEx you see is split in a way that's different. Also, some of the same people now are in the packaging that used to be in the front end.
Mm-hmm.
A lot of them know KLA. What happened with Orbotech is suddenly we started having conversations with those customers that Orbotech told us on their own they wouldn't have those conversations. Because a lot of those customers, they wanna rely on a supplier that they can count on and hold accountable for their results. I would say two things. One, we've seen a growth in that market, and the other, we've seen a huge growth in the engagements we've had about stuff that's coming out in the next couple of years. Bren, what would you say packaging was last year for us?
It was probably $300 million, $300 million-$400 million or so.
That grows, we think, over time because of these drivers. I'd say we're early on, and we're early on in the product development associated with Orbotech because it's only been a year or so that we've been having these conversations, so we're still in development. We do have new products that'll come out to satisfy that market.
Yeah. Okay, great. So I'll ask one more question and see if we have questions for the audience. Can you talk about memory, process control dynamics in memory as robust as they are on the logic side? You know, do you see that in the context of higher layer counts in NAND, that being more process control intensive? As you look forward, and you look at things like 3D DRAM, you know, where do you see that?
Yeah. Two things. One, I think finally, we're seeing EUV being applied in memory, and so in DRAM.
Mm-hmm.
That drives with it a number of opportunities for us. I think that the registration requirements and metrology requirements are pretty high. In 3D, yes, an increased number of layers, but the process control intensity of memory has always been lower for two main reasons. One, there's less process change.
Yeah.
The other is they repair, right? Because they can repair, they can run with a higher defectivity count. Bren, in terms of the percentages that we're seeing?
I believe high single digit, you know, 9%-10% process control intensity.
Mm-hmm.
If you know, a foundry would be sort of a mid- to high teens, a high-mix foundry, you'd have memory somewhere in that range. A little bit higher with 3D memory, particularly on the flash side.
You know, you do have redundancy there, so it's in more commodity parts, so there's less intensity overall. We are seeing opportunities certainly with the introduction of EUV and DRAM. That's creating, at a minimum, just infrastructure requirements. We are seeing some investment there. You know, it's not that there are problems that we can solve. We can solve them. There's opportunity for us moving forward. It'll never look like a foundry sort of logic environment necessarily.
Yeah.
At the same time, the introduction of EUV and some of the increased layer challenges will create, you know, some new opportunities for us.
One other thing we saw, and we're still seeing now in bare wafer inspection from the people that make wafers, as you have higher stacks, the flatness becomes more critical-
Mm-hmm
and the cleanliness. Actually, it translates to some of our business in the bare wafer manufacturers as well.
Okay.
Driving up intensity and their need for precision.
Okay. Are there any questions from the audience? All right, I'll keep going. Can you maybe talk a little bit to the competitive dynamics? You know, I feel like it's pretty well appreciated how strong your position is in these markets, but there continues to be focus on, you know, next generation e-beam, as well as some of your optical inspection competitors. Can you talk about the strength of your competitive position?
Yeah, sure. When we had the Analyst Day in 2019, we talked about our 2023 plan, and part of that we were a little off on 'cause we said WFE grow 6%-8%, and it's been higher. But the part that we nailed was what we thought we would do with our market share. At that point, we said 250 basis points, right, over,
Yeah, over the timeframe of the plan, the 2023 plan, yeah.
We've gained some share, and clearly, we feel good about the investments that we're making. There are always, you know, because of our margins and because of our market share, we always have competition. I think when we look ahead, we continue to think that we'll continue to gain share, but it won't be at a huge rate. I think it'll be in the 50 basis points a year kind of, you know, level. The areas that I think are most critical for us, optical inspection was one, and we talked about this at the Analyst Day, that, you know, many people had written off a few years ago.
Mm-hmm.
E-beam was gonna take it over. I took a look at 7 nm flow, and there, if you look at the inspection points, I was talking before, you do the debugging and then you put it in the section. There were like, what did we say? 250 inspection points. 150 were with what we call the laser scanning systems.
Okay.
About 60%-80%, something like that was BBP, and 1% was e-beam. If you go back 10 years, probably 1% was e-beam. The reason for that is the price performance of e-beam relative to optical lends it more and more toward the debugging end of my analogy there, and less into the production.
Mm-hmm.
That limits the volume that you're ever gonna sell because how many do I need? Yes, we'll sell some for inspection, but mainly in what we call engineering analysis, not in the scaling of production. That's been why that market dynamic's always been like 80/20 between optical, and that's been true for-
Every year.
Yeah.
This was a conversation 15 years ago.
Right. It's because it's literally 1000 x faster.
Yeah.
Now we develop e-beam technology too. In fact, we have an e-beam inspection tool that is uniquely positioned in the market, but it's for reticle, and it's what we're using right now for EUV reticle inspection.
Mm-hmm.
We think that that's a limited inspection that you're doing because it's a reticle. You don't have to do and process thousands of wafers, and we think that's the right technology for that particular application. I think competitively we feel good. We're investing at a high level. We've historically built on our share gain that we think will continue to, you know. That's our plan, that's our forecast, and that's what we think.
Yeah, I think share will be flat year-to-year. Really, the movements in share, both pluses and minuses, have to do more with the relative growth of certain parts of the market.
Mm-hmm.
You know, we have a portfolio business. We have some part businesses that in a given year may not grow as fast as overall. WFE is an example, and some that have grown much faster. Depending on where you are, your movements or the growth is a little bit different. When you aggregate it all out, we think share is roughly flat, maybe up a little bit, as we look at 2021, which was a similar result in 2020. Not much change overall for the reasons Rick talked about.
Great. Well, I didn't get to all my questions, but I guess we're out of time. Thank you very much, guys. Appreciate your time.
Thanks, Joe.
Thanks, Joe. Thanks, everybody.