ASML Holding N.V. (ASML) Earnings Call Transcript & Summary

Good morning, and good afternoon to everyone for joining us for this fireside chat, with Pete Convertito, ASML Investor Relations. My name is Didier Scemama, I'm the Head of European IT Hardware and Semiconductor, Bank of America Securities. Warm welcome to Pete. We know each other for quite a long time.

I'm going to start my first question, Pete, with regard to EUV. That was the big surprise and the upside surprise in your calendar year '21 revenue guidance raise at the time of Q1 results. Can you maybe just run us through the bridge between the sort of EUR 4 billion to EUR 6 billion EUV revenue bucket that the company has sort of printed over the last 5 to 10 years? And the current sort of EUR 8 billion plus that you're guiding for, if you could talk to us about what's driving the upside? And how much of that is cyclical elements versus maybe secular elements, maybe as a starting question?

Sure. Thanks, Didier, and thanks for having me. Great to be here. Yes. So in our quarterly earnings, Peter and Roger talked about 3 trends going on right now, right, a cyclical trend, a secular trend and then the tech sovereignty trend. And the uptick in our guidance for the year, to your point, was around DPV. And part of that falls in the cyclical trend being catch-up from last year for decisions that were made last year, whether it be to hold off buying equipment because of profitability concerns with pandemic or just inability to get tools into people's fabs or take tools in, I guess. So part of the cyclical part is just decisions from last year that they're taking more equipment in this year, and that was primarily DPV. And the uptick, if you know, it was mostly around Memory. And secondly -- but there was some Logic uptick because we did raise that as well. I'd say in the Logic portion, it's around mature nodes running longer and stronger. And if you went back to our Investor Day of 2018, we talked about the secular drivers of 5G, artificial intelligence, autonomous driving, high-performance compute and all the memory that goes with it. That secular trend continues. And in fact, it was probably accelerated by the pandemic with all work-from-home and remote learning. But it also enabled, call it, Internet of Things or I think some customers are calling it edge compute. And those are really a couple of nodes back, not leading edge nodes. It's more of sensors and cameras and things like that, that are in our homes, thermostats that you control from your phone, cameras in your home, all kinds of content in cars, especially electric vehicles that are coming into prominence now. And those are a couple of nodes back, which don't require EUV but requiring quite a bit of DPV. So the big uptick in DPV was driven by the cycle, the upcycle we're going through in Memory and also these mature nodes running longer and stronger. So on top of the normal secular drivers that we were talking about in Investor Day and that was being accelerated a bit.

Yes, that's great. It makes sense. Good color. Coming to core, if I rewind the tape a little bit. So roughly 6 months ago, Peter already sort of said, "Hey, we're thinking the EUV is going to grow now as opposed to shrink because ArFi is not falling off a cliff. It's not being cannibalized as much as we thought by EUV, which obviously is a big difference. And the Q1 call, Peter went as far as saying, "Hey, we're thinking about increasing DUV production capacity, double digit. Right now, we're selling everything we can. So we're giving -- we're selling about EUR 2 billion give or take per quarter." Can you give us an update on your thinking around this DUV unit production capacity? I mean, have you already sort of made up your mind? And Peter said, we're thinking this would grow double digit. Is that sort of dried effectively for next year?

Yes. I mean, I think if you look at it, we're already from bringing out what we can out of the supply chain, our own cycle time, our supplier cycle times, any inefficiencies out of the supply chain, we're already up significantly this year over last year. We think that cyclical part I talked about carries into next year. And we don't -- our gut feeling is this longer, stronger, the older nodes does not go away. But we're going through that analysis now. We do kind of our annual modeling, and we're looking at that. We're looking at what capacity also do we need per EUV beyond 2021 or 2022, sorry. And all of that, we'll make decisions about what we think we need to do from a capacity standpoint. We've been able to do a considerable part of that, again, just through bringing out the supply chain. But in the next, I'd say, few months, we'll make a decision on do we need to either invest in equipment or manpower or square footage, whether it be at our own facility or in the supply chain. But those are going on now, and I think you'll get an update either in the next earning call, if not then at our Investor Day in September.

Got you. And if I can sort of move this debate a little bit further. So as I said, in the past, DUV has been a EUR 4 billion revenue bucket roughly for ASML in a sort of downturn and then EUR 6 billion in an upturn. And this year, we're talking about EUR 8 billion. You mentioned, obviously, a number of secular drivers, number of cyclical drivers. You mentioned, in particular, ArF and Memory. What I wondered is, are we moving now into, as you said, stronger for longer, are we moving now in a new range, which is maybe, I don't know, EUR 5 billion to EUR 9 billion or EUR 6 billion to EUR 9 billion? Or do you think that the circumstances that we are in now are exceptional and even the secular drivers can be overwhelmed in a case of a downturn, in which case, we should really still consider EUR 4 billion as a sort of downturn number?

Yes, it's a good question, and that's kind of exactly what we're going through now is this longer, stronger of the older nodes a transient thing? Or is it really a secular? And our gut is that it's secular. But that's what we're going through now, and we'll make some decisions in the coming quarter. We always think there's going to be cycles. They're obviously at -- as whipsaws. They were as drastic as they've been in the past, but there'll always be cycles. But our gut feeling is that this is -- this leading edge secular drivers continue, but it's enabling another market of, whether you want to call it specialty silicon or Internet of Things or edge compute. It's more content in devices like this. It's more content in your home. It's more content in cars that is enabled by leading edge but doesn't require leading-edge technology to do it. So it kind of piggybacks on. And that's the decisions we'll make in the next couple of months. But to your point, at our Investor Day, we'll update the longer-term model. So I'm sure that's where we'll talk about our parameters of -- 2018, we talked about leading edge and logic nodes. Wafer starts on our mid-market moderating 10% per node in a down market, moderating 20% per node in a high market being flat. Does that still hold? We'll update on that. I think one of the things we didn't catch in our 2018 was again this phenomenon of longer, stronger with the older nodes. So that's probably something we'll be talking about as well as does -- how does that contribute? Because that's all DPV.

Yes. Cool. So DUV and the cyclical elements that we sort of talked about, I'm just changing gear now to EUV. So to recap, you said, we've got 40 EUV unit production capacity for this year and then you hike that 45 to 50 to 55 for calendar year 2022 at the time of Q1 results. I wondered, Peter, if you could maybe remind us what your lead time, excuse me, are for EUV at this time? And also how you managed to effectively increase the production capacity for next year by 5 to 10 units? And did you manage to reduce the cycle time effectively to manufacture those tools. So if you could touch a little bit, effectively why you found those efficiencies to get there, that would be helpful?

Yes. So our cycle time, we are running about 20 weeks once we get all the modules in our facility, and we continue to work that down. But the real -- the gate is in the supply chain, if you will, or the restrictions in the supply chain right now basically in 2 modules, in optical modules and drive laser. The -- in the optical modules, it's really the long lead time of going from, call it, growing raw glass to getting it to our door and then we can affect the cycle time a bit on the end of the integration and test on our end. But what really restricted us in terms of this year and even next year a bit is supply, getting the optics in. So we're planning for 55. Our suppliers are planning for 55 next year. We could probably squeeze a tool or 2 more out if we can get some more of the modules. But right now, 55 is the plan, if we can do better. It's greedy. The planning now is about what do you need for 2023? Is it 10 more tools, 15 more tools? Then that requires some investment in probably in the supply chain and in ourselves a bit. Is it 5 more than you? You say, okay, what's the return on investment? Do you just kind of say, hey, we can do what we can do and accelerate a bit more and then go through. But I think what people also have to keep in mind is that -- and I think Peter mentioned it -- Peter Wennink mentioned it on the call, we're going from Cs to Ds this year. Next year is all Ds. So 55 Ds that's really like shipping 65 Cs. So you -- from a wafer output capacity standpoint. So you're getting more wafers out with fewer tools. Now obviously, we get paid for the value we provide in a higher ASP. And then in late '23, we have another E version coming out, that will be even faster. So put more wafers out. So you got to factor in what the wafer out component can be with the newer tools when looking at the number of tools you're going to need. So again, that's an area we're discussing with customers now is what's the outlook in 2023 and what capacity we need to be at with a 15- to 18-month lead time on -- or call it, cycle time or lead time on optics end-to-end out the door on a system. Those are decisions we need to make fairly soon to be able to affect 2023.

That's great color. On the -- you touched on the bottlenecks, optics, the laser and drivers, et cetera. Can you maybe just give us a little bit more details on this? And related to that, back in the days, you bought fiber because you felt like not owning the full value and IP in-house was inhibiting you, in fact, from hitting your milestones. More recently, you also made some small acquisitions like Berliner Glas, et cetera. Do you feel like you could do more or you would need to do more, and in fact, to increase the cycle or reduce those cycle times, I should say? Or do you feel like, okay, you know what, the land is really outside of our competency. So we should not touch on that or maybe some other bits that you feel like you may or may not want to be involved?

Yes. Well, I think Cymer was a little bit different story. It was in a technology development stage where it was accelerating it was investing more in, call it, basic R&D. So making that acquisition was about from a financial standpoint and, call it, fiduciary responsibility that Cymer had at the time of investing more in the technology and doing on their own as to making the acquisition accelerated that as evidenced by what we were able to do quickly once we acquired them. With ZEISS, not that we didn't have this with Cymer, but with ZEISS, we have, obviously, a very good partnership with them. We own 25% of them or 24.9% of them. Would be -- I think we even -- Roger and Peter in the past would say, "Hey, if they'd sell it. We acquire it. But I don't know that we need to do that." This is more about getting raw glass, and it's not about technology so much. It's manufacturing, lead times and capacities and we can help, whether it be if they need capital to get other equipment or build square footage, we don't have to acquire them to do that. So we'll help them -- how we can, at this point, whether it be if -- you could get more equipment or more square footage. Manpower, obviously, we -- as everyone needs more manpower to help these things along, but we'll see how that goes in the next couple of months.

Great. You touched on the 2023 EUV. So I wondered if you could give us a bit of color on your order book? How it's sort of looking like, shaping up in terms of foundry, logic and then the DRAM? If you can give us a little bit of insight into that? And do I understand from your point that we're taking orders for 2023 that you're fully booked for 2022 already?

Well, no, we're not fully booked for 2022 yet. You saw our orders in the first quarter, but we expect a steady stream of orders over the coming quarters to fill it out. I mean, I think we have quite a few orders on 2022 already. But again, we're not so much an order-driven company as we do these volume purchase agreements, and we're constantly upgrading forecast with customers. In terms of the mix, I think if you look at last year, we did about -- it was about 90% Logic, 10% Memory and EUV. It probably goes to more like 80% Logic, 20% next year. So we're somewhere in between this year. But I think we expect a steady stream of orders that come in the coming quarters.

Got it. Got it. One of your customers has publicly stated its ambition to become a major foundry in the sort of medium term. Should you give us a sense of whether you could accommodate effectively volume capacity on EUV for new player in foundry given your constraints at the moment and I suspect demand from your other guys?

Yes. I mean, I think there's a couple of things there. I don't know that it's clear that company, what nodes they plan to do for those foundries, whether they require EUV or not. Certainly, they're going to require for their leading edge fab that they're starting to ramp. And we'll see orders. We see some orders, and we're seeing orders for that. In terms of the foundry ambition, yes, like I said, I think it's unclear what nodes they're going to do on that yet and whether it requires EUV or not. So that's a bit TBD. But we talked about the lead times on EUV. So that gives you an idea of they want them in a certain time when they need to start ordering them.

Got it. I wondered if you could double-click a little bit on EUV in DRAM. You've got Samsung that deployed Hynix following and Micron, I guess, next. Can you maybe give us a sense of the number of layers that you're seeing adoption for EUV at the current node? And how that could shape up over the next few transitions?

Sure. Yes. A number of those have been pretty public about their plans for EUV. At least one of them is already doing 1 layer at 1z and talks about maximizing their use of it at 1 alpha. Another player is roughly a year behind them from an adoption standpoint. But they're coming along. We talked at our Investor Day back in 2018 about roughly 1 to 2 layers to start, which we're seeing that at a couple of players. Going to several layers at the next node called 1 alpha, which could mean somewhere 4 to 6 layers. And then beyond that, probably 6 to 8. And the third guy is certainly looking at EUV. They've also been a customer that's been very good at pushing a wavelength longer. So we've seen that. They're going through their, call it, analysis and evaluation, and they'll bring it in when they think it's economically feasible or makes economic sense. There I say at least one of the players that can do both Logic and DRAM had a bit of an advantage in learning on the Logic and seeing the benefit not only from cost, but from a yield and device performance standpoint. So when you do that and then you have the ability to run some Memory layers on that you get past the spreadsheet and analysis and see what it really can do. So they probably had a bit of that benefit over others as well.

Got it. Closing the chapter on EUV, I wanted to ask you a little bit about High-NA. For the audience that may not be super-technical or only looking at technology in semiconductors, can you maybe please remind us what High-NA is? And secondly, maybe talk about the challenges of bringing EUV to mass production?

Sure.

So High-NA and EUV.

Yes. So High-NA, just like when we're making the transition from immersion to EUV, that transition took place because you run out of resolution capability on the immersion tool. And they had to go to other techniques like multi-pass patterning, which basically is breaking the pattern up into different sections, if you will, and printing one section and then going back and printing another section in between because of the resolution capability of the tool. With EUV, you'll get the same thing. You get to a point where you're shrinking a few nodes. And then you run out of a bit of resolution capability. And you'd have to go to multipass patterning with 0.33 and EUV, which customers would prefer not to go too far down that road if they don't have because it's expensive, both from all the added steps as well as the -- any time you add more steps, you're adding complexity and opportunity for yield loss. So with High-NA, it's basically a larger collecting power of the optics that give you better resolution. So unlike the change from immersion to EUV, where there was a whole ecosystem change in terms of new resists, new masks, going from atmospheric to vacuum environment. This is really more evolutionary going from 0.33 NA to high numerical aperture. So the major changes between 0.33 and High-NA is a bigger optical system, which I don't want to trivialize. It's quite complex. But it's evolutionary. It's a lot of brute force engineering as opposed to pure R&D, which you had from going from immersion to EUV. So same basic source, it's just evolved a bit, higher power. In fact, it will be the same -- exact same source we're doing in the -- whatever the 0.33 system is at that time. Everything is scaled up a bit. So stages, same basic technology, but scaled up same source. Again, most of the development is in the optics area. So a lot of development in engineering, but not nearly the change in all the ecosystem that you had going from immersion to EUV.

Can you remind us sort of -- I think you said like introduction maybe 2023, 2024 and then mass production expected 2025, 2026. Is that the right ballpark? And then maybe if you could talk about how should we think about gross margins as we first introduced the first -- as you first introduced the first 2? Is it a 0% gross margin, a bit like UV? Or as you said, ArFi effectively you've got a learning curve, but your gross margin doesn't start quite as low?

Yes. So -- yes, on the margin standpoint, EUV, we started negative gross margin. We were shipping tools that weren't meeting spec yet. We were shipping them, taking the cost when we ship them, but we couldn't take revenue until we demonstrated it in the customer's facility. So a lot of deferred revenue there. You won't have that with High-NA. The anticipation is that we'll ship these. We'll still have to prove them out in their fab. So we won't revenue -- recognize the tools when we ship them. It will be after they're in their factory, and they've been accepted. But it will start positive. Now it won't start it where EUV is now in the low 40s, going to the 50s, but it will certainly be positive gross margins and then grow from there. Yes, something else earlier before the gross margin, therefore...

No, I wouldn't you just remind us what you're thinking in terms of first tools like beta tools introduction? And then perhaps when you think -- maybe can you talk about a little bit about the node, why you think that mass production might be using High-NA?

Yes. So we're in the process of building modules and testing modules for the first tools. We took orders for the first 4 tools back, I think, 2, 3 years ago. Those are progressing well. We'll start integrating the system next year with plans to ship the first system in 2023. And those first 4 early production capable tools for customers to get going with. And then there'll be a follow-on version that's, call it, the high-volume production tool probably in the 2024, 2025 time frame is when we'll probably see it cut in and start to be used.

Got you. Maybe we move on to inspection. I wanted to ask you sort of a geopolitical question around China. I'm sure everybody is asking you that. I wondered, how do you think about the situation currently and whether there would be further escalation of tension and a complete embargo on litho tools or semi cap or something related to that, that would effectively prevent you from shipping the EUV tools to China? How would that handicap the company?

Yes. So obviously, we always abide by whatever the export laws are for the reason we have to ship into. Right now, the only thing we cannot ship is EUV tools. We haven't received an export license from the Netherlands for that yet, but we can ship DPV tools. We were a bit conservative coming out of last year about guidance on China a bit. Because, in particular, we had a customer that went on the entity list. It didn't prevent them shipping, but our U.S. peers needed to get export licenses. And if you can't get your dep and you can't get your edge or you can't get your metrology equipment, our concern was they might hold off and say, "Hey, let's hold off on taking litho equipment. We haven't seen that yet." I'd say we derisked that a bit in the last quarter because even if that does happen, there's demand elsewhere that could take those systems. I think there's been discussions about hey, should we cut off, say, hey, at a certain technology. I remember when I came into the industry back in '96, there was hey, you couldn't ship anything less than a quarter micron technology. Would they do something now where you can't ship below a certain technology, let's say, it's 14 NAND out or something like that. Below that, you still need immersion. So how do you decipher between once an immersion or below or above. So I don't think they've gone there yet. Right now, we don't see any restrictions on the horizon for that.

Okay. Got you. Maybe let's touch on inspection. Just give us an update on what the business is? How the business is doing? Clearly, it's doing very well and growing very, very fast. What's your long-term ambition in inspection?

Yes. So in our metrology and inspection, on the metrology side, we have YieldStar metrology, which has grown pretty well with immersion as well as EUV in terms of overlay metrology and overlay control of our tool on the -- we also have e-beam metrology that we're working now. In terms of e-beam inspection, when we acquired HMI, I'd say, a big part of their core business was voltage contrast inspection with e-beam. That was a bit muted over the last couple of years because Memory had gone through its cycle. And these voltage contrast tools are used for looking down through these deep contact holes, particularly in 3D NAND, but also in DRAM, to see if you cleared to the bottom. Again, it was a bit muted over the last couple of years because of the cycle memory was in, but we expect that to step-up in the second half of this year. So I think we'll see increasing business there. On the multi-beam side, we shipped our first 3 x 3 beam eScan 1000 last year. We shipped several of those. They're basically learning and evaluation and process integration tools, and they're in various modes of learning and integration at customers. We'll come out with another tool later this year that's a higher number of beams. And that platform is probably the first platform that is more of a volume -- a production capable platform. So we'll start shipping that later this year. And we think we'll start to grow revenue in that throughout next year into 2023. But it's another area we'll talk more about at Investor Day in September.

Got it. Maybe one final question for me before we take questions from the audience. Installed base management has been a steady heady. Can you talk a little bit about the business model evolution with EUV? How you're going to monetize this installed base business moving from a transactional business with cost to a more, let's say, recurring revenue with cost as well. But more directly linked to number of wafers being produced and how that could change the sort of profile of the business, perhaps from a growth standpoint, but also maybe from a margin standpoint? Maybe explain if you want the EUV versus DUV today?

Sure. So on the service part of installed base, because installed base is a combination of our upgrades business as well as service. Our old model was, call it, a break-fix model. The customer calls up service, tool went down server guy goes in. If he can get it up and running pretty quick, everyone's happy. If you need -- if it's down from -- it needs parts, may have to get from a warehouse or maybe even a factory, which could mean several days of downtime, which is not what a customer wants to hear. It's hard to plan production around that. And under that model, we're getting 1% to 2% of the ASP of a tool in revenue. But we weren't -- this tool was sold to a certain uptime capability. And if we're meeting that uptime capability, we're not incentivized to do any better than that, if you will. So even though you get these long downs once in a while, we're still meeting the uptime capability. The new model is really an output-based model. And the genesis of it was really out of Cymer who had that type of model for the laser source. They called it OnPulse. Basically, they -- you get charged by consumption of pulses of the laser module. So it's -- from that standpoint, it's not something that's shockingly new to customers. They've seen it there. But with EUV, it's basically, there's -- has agreed upon output planned for the tool. If we exceed that, we get paid more based on how we exceed it. If we come in lower than that, then they pay less based on the output of the tool, but we're responsible for maximizing the output of the tool for the customer. And under that model, we think we can get 5% to 6% of the ASP of a tool per year in recurring revenue. So basically probably changes the slope, if you will, of the trajectory of services for us going forward. We've talked about installed base management having a CAGR of about 10% per year. I think you've seen last couple of years, it's been a bit higher than that, but we guided roughly that for this year. That's because, again, it's a combination of upgrades and this service revenue stream. And the upgrades business is a bit cyclical, and it's kind of a bit cyclical for a couple of reasons. One is the upgrade product cycle, if you will, every 2 years, roughly, we come out with a new tool or new version of a tool, and those upgrades are backwards compatible. So you get a new product cycle, if you will, every couple of years. But also, it's really more dependent upon the customer's utilization in their factory. In times of high utilization, they're basically printing money. So they don't want to take tools down for a long time to put upgrades in. If you have a software upgrade, which can go in quickly in a couple of hours, be tested and back up and running, they'll take those during times of high utilization, which is really what we saw in the first quarter. That drove our beat in the first quarter was customers just really clamoring for capacity. And we had some software upgrades that can help them get a bit more capacity on the tool with a quick software upgrade. So they said, "Hey, we'll take as many as though as you can get us." But again, it's a year of high utilization. So hardware upgrades, which take a bit more time to go in, they're going to hold off them until they have time to do it. So the last couple of years, you've seen Memory kind in that down cycle, less utilization of tools. They could load them up with upgrades. So those were high upgrade years. This is probably less of an upgrade year, but still a 10% CAGR. So there is a bit of cyclicality of our installed base management with the service part growing monotonically with the installed base and then this cyclical up tide bit of the upgrade business. From a margin standpoint, the EUV service, we started -- just started generating revenue last year. And we have quite an army of people out there servicing. I think we commented that we went -- we broke even on the EUV service gross margin in Q3 last year and went positive in the fourth quarter. And Roger Dassen, our CFO, has talked about growing that to corporate average over the next couple of years. So that's certainly a lever for us on margins going forward.

Okay. I think with that, we are out of time. Pete, thank you so much for your participation and for your kind answers, and many thanks to everyone calling in. Have a wonderful conference.

Great. Thanks.

Thank you.