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

Hello, everyone. Good afternoon, good morning. Welcome to day 2 of the TMT conference. I'm delighted today to introduce Roger Dassen, CFO from ASML, joining us today. Welcome, Roger.

Thank you. Good morning, good afternoon.

Obviously, lots to discuss following the CMD and also the commentary at the Q3 results. Just a note to all the investors listening, as you would have heard many times, if you want to ask a question please drop that down at the bottom of the webcast. It should hopefully be fairly self-explanatory and we'll include that in the conversation.

Maybe what's really a good starting point is just looking into 2022. Obviously, you've described a scenario where supply is the real constraint from your perspective. So is it fair to say that sort of the building blocks into 2022 include around about EUR 8.5 billion in EUV sales. A similar number was around EUR 8.5 billion to EUR 9 billion in DUV and about EUR 5 billion into IBM, and then maybe to broaden that out, within those system sales, how are you seeing the state of demand between logic and memory?

Yes. Thanks, Dom. And I think that steps directionally sort of where I think we're going to land. And of course, we'll give a bit more color on context in the January call when we discuss the Q4 numbers and give a more fulsome outlook on an expectation for 2022. But I think directionally, it's as you discussed. So EUV, I think what we said is we're looking at about 55 systems and the ASP for EUV, all of them will be Ds, is around 160. So indeed, that gets you to, let's say, about EUR 8.8 billion of revenue on EUV system sales. I think DPV, I think as we said, at this stage we're probably looking for next year at about the same level in DPV and apps combined as we're having 2022 about the same level as this year, and I think that's going to be approximately a little under EUR 8.5 billion, is what I think we're probably going to land at this year. And I would expect that to be there next year as well. As you know, we are building capacity on DPV, but as we also mentioned this year in order to get to the level little under EUR 8.5 billion that we discussed, we really needed to deplete all the buffer stocks that we had lying around there to get there. And of course, you can only do that for a short period of time because at a certain point in time customers really do need their parts because the systems are breaking down. So that's why we need to replenish those buffer stocks again and safety stock again. So quite a bit of the incremental capacity that we're building next year will be used to replenish those safety and buffer stocks. And I think you're right on the installed base business. Installed base this year I think is going to land approximately EUR 4.5 billion, and as I said on the call a couple of weeks ago, the CAGR of 12% that we've indicated for installed base business all the way through 2025, I think that's the right proxy to look at. So that indeed gets you to approximately EUR 5 billion. So yes, I think those are the building blocks in terms of your question, the different market segments, logic versus memory. I think nothing really changed there in comparison to when we last talked. So on the logic side, very, very strong, I would say, both on the mature side and on the more advanced side, strong end market demands based on the digital transformation. IoT very much driving the mature nodes as well and also the catch-up that is currently going on in terms of the chips shortages, I think all of that is very much driving also the sales level for 2022. On the memory side, I mean there's a lot going on there, but I think the way we look at it, we're still, for 2022, looking at a bit growth. I would say, high teens for next year, that's what we hear from our customers. And then of course next year we're also looking at the further introduction of EUV into DRAM manufacturing in particular. And as I mentioned before, I think more than 20% of the 55 tools, EUV tools for next year will go into DRAM manufacturing. So also that will give a good boost on memory. On that front, memory, just to round it off. I think at this stage, if we look at the utilization of our litho tools, they are at pretty high levels. I would say, historically high levels, and that in combination with the bit growth that we hear. For us that means that we're probably still looking at healthy demand and actually further growth of the memory business into next year or so. So those are the moving parts, but the most important point is the one that you made. I think we're very much still looking at being supply constrained. Demand is very, very strong and it really is looking at how can we further increase capacity in order to cater to the demand next year and beyond that point.

And already we have a few questions actually from investors just on that supply chain. So maybe we'd take that now because you also said that. So just as what you just described, could you talk about how the evolution of that buffer stock is evolving into next year? And what are the major constraints? That's grouping a few questions together for investors.

Yes. So on the buffer stock, it is all over the place. It is in a number of areas. For instance, it's on optical systems -- on the optical components, I would say, that that's one of the elements of the buffer stock, where we really depleted that in order to put them into new systems this year. So that's one of the areas where we are building it. But it's on quite a few other fronts as well. And in terms of size, how much is that? I would say it's somewhere between 700 million to 800 million is probably the right number to call out there that we need to rebuild in terms of buffer stock. That's sort of the number to look at there.

So turning maybe to the longer term. And obviously, being supply constrained, we obviously say that, okay, well, let's look at the longer term in that context. Obviously, one of your customers on the logic side has recommitted their business model to advanced manufacturing and equally to potentially pursuing a foundry model. So just curious whether you are seeing a difference in the visibility you're getting from maybe your [indiscernible] out from the logic and foundry side in aggregate from that industry versus what you were having previously from your conversations?

Yes. I would say on the demand side for them. So the demand that they are looking at, I wouldn't say that the level of insight there is dramatically different. I mean it's pretty clear that they are a supply constraint as well, right? It's clear that many of their end markets are booming as well. So no, I would say the level of transparency there hasn't really changed. It's just very demanding for everyone, that's clear. I think the one thing that probably did change is that the one customer that you talked about the North American customer, I think they have changed their approach in working with their partners including their main suppliers including ourselves, and that means that the level of transparency of the dialogue that we're having with them on their journey, on their road map, I think that has increased. So I think the level of insight, where are they, how are they developing, what's working, where are they seeing change, I think that has increased, and also now our ability to help them as much as we're helping all of our other customers. I think that has definitely gone up. The main message that we're getting from all of our customers, be it in -- and definitely on the logic side as I mentioned is, a lot of activity, a lot of demand in a number of instances also, quite frankly, sometimes a bit surprised about some of the developments. I think the resilience on the more mature nodes is clearly something that I think has surprised also some of our customers. One element that we gave you some time ago is, for instance, the dye size of image sensors, which has increased quite a bit, and as a result of that, now because the dye size is larger, obviously, more wafers need to be produced in order to produce the number of image sensors that are required. I mean those are sometimes surprises also to our customers leading to increased demand. But I think that's what all of them are experiencing and sharing with us.

I am also curious around how do you -- when you take a 5-year or long-term perspective on a potential customer that's growing a foundry business from 0, how do you yourself manage the capacity growth that, that implies, both from an OpEx perspective, but also from a physical capacity and supply chain evolution, because it increases the level of uncertainty, because you don't know, I guess, or maybe you do know and you want to share, but the size of what that business could be.

Yes, and Dominik, maybe I'd even expand that example, because in fact, the model that we apply is to say, based on the developments in the end markets, what do we think is the global demand for wafers? And then the global demand for wafers, as you know, is then being translated into the capacity that is required for us. So that's how we model also the 2025 numbers. And we do that on the basis of an expectation that the market is as efficient as it is today. So also the way those wafers are being produced is as efficient as it is today. Of course, that dynamic changes a little bit. For instance, if you have a third player that steps into the market. But also, if you have a number of existing players determining that they're going to build fabs, not just in 1 location, but are actually going to build fabs, not just in their primary location, but also in a number of other players, and I think that's what you currently see happening, right? I mean you see TSMC, you see Samsung with fab build plans outside of their main territory, also into the United States. We see players interested in building fabs in Europe. So that's the level of, you might to a certain extent call it inefficiency in comparison to the status quo, if you like, that we have to cater to. And that's why we said at the Capital Markets Day, we model the demand for our equipment based on the assumption of continued efficiency the way it works today, but we do recognize that, that might change, and that's why we're actually building and are considering to build more capacity also towards '24 and '25 than what you would get merely out of that efficiently produced wafer model that is the basis for our assumption. So that's why, in a high market scenario, we put in the 70 EUV tools by 2025, but we reckon that the demand might actually be higher. So that's why we're looking into that, trying to get an estimate, how much more might there be there than that 70 tools, and then making sure together with the supply chain that we are able to cater to that demand. So that's a CapEx decision. It's a bit of an OpEx decision. On the OpEx, I'm not too worried because frankly, Dom, even if the demand were only there for a very short period of time and it would then fall off, I mean the growth that we're going to experience also on the basis of high-NA, I mean, we need people, right? So from that vantage point, I think the OpEx side of life is the least of my concern. On the CapEx side, I think what we need to do in order to get to the 70 and also beyond the 70 will require a little bit of CapEx, but it won't be dramatic. A lot of the increased capacity will actually come from improved cycle time. So we will have to do some CapEx, but it won't be earth shattering. And I think that the payback period for us will actually be very, very short. So on our side of the equation, I think it is worthwhile to expand capacity to 70 and beyond that point and now we're working with the key players in the supply chain to make sure that they're tuned to that as well. Let's talk about EUV. I think the key elements in there to look at, I would say, Dom, are on the optic side. So that's working with Zeiss, and also on the drive laser side, which is another German company called TRUMPF. So those are the 2 main areas, the 2 key players I would say in the supply chain that we're working with, saying what does it take to get you to that desired level and those discussions we're currently having.

So taking that 1 step further, obviously, at the end of that horizon, we're looking at introduction of high-NA and high case to talk about 5 unit volume in terms of revenue recognized, which I appreciate is the accounting perspective, not the real-world machines shipped. But it does really sound like or at least the impression is being given that there is a real emerging fight for many of your key customers to get their hands on those initial high volume tools, not just the R&D tools, but the high volume tools there. So what kind of capacity are you expecting at that time frame and beyond?

Yes. I think around that time frame, and there will be a gradual slope obviously, but we think in a steady-state situation, I would say, that's probably not the case yet, in '25. But it will be the case, I would say, in '26, '27. In a steady-state time frame, you're probably looking at 15 to 20 tools, that's the capacity that we're aiming at I would say in the second half of this decade.

And then just turning to a slightly shorter-term topic. We also have -- one of the customers has historically used half nodes between the major nodes at the leading edge. Now partly that's from a cost perspective, probably let's say improve the speed, characteristics, et cetera. As they're doing so again in this instance, is that just a revert to the norm of this is the normal product cycle from your perspective? Or does that have any impact from ESMO's perspective in terms of like a layered option, et cetera, that the EUV layer count is getting a bit much. So in order to bring the market in from a consumer perspective, you need those half steps.

No, I think it's not a fundamental shift. I mean indeed you're right, we have seen that before, these subnodes, as we have come to call them. And the subnode really is an opportunity for the customer to look at how can we, within that node, drive down costs; how can we, within that node, maybe increase the energy efficiency of that node. So that's really what's going on. Sometimes you also see that with the introduction of that subnode that there is a modest increase in layer count there. So yes, it's all about making the node more efficient, making it more cost efficient and in that way making the node transition and also what our customers have to offer to their customer, make it worthwhile and make it to transition that from a device characteristic perspective, but also from a cost perspective make it attractive. But I would say it's not a departure of the past. I think it's a continuation of what we've seen those customers do in the past and I think that's what we see here as well.

So obviously in the past, let's say, quarter or past few weeks, 3D DRAM is the inescapable hot button topic for everyone on the wafer fab equipment side. You've been fairly clear about aligning ourself to customer road maps. So it'd be interesting just to refresh and describe for everyone what you're hearing from customers and how you approach this phenomenon and whether this is really more of a 2030 versus 2025 story.

Yes. Good question, Dom. So indeed, what I said before on 3D DRAM, there are 3 questions. Is it going to happen? If it's going to happen, how is it going to happen? And if it's going to happen, when it's going to happen? Those are, I think the 3 main questions to be answered. And exactly to your point, I think what is primarily important here is how our customers think about it. So first, if I look at the road maps of our customers, I think, the ones that are public about it are pretty clear that geometric shrink and X and Y is an important element of what they're doing. I mean, starting with Micron, they're pretty clear. They're introducing EUV into high-volume manufacturing in 2024. And of course, they are not doing that for 1 year. I mean, clearly that is something that is such a fundamental shift, they would of course only do that if they think that, that makes sense for quite a few generations in succession. Otherwise, you wouldn't do that. So they're pretty clear about this being an integral part of their road map for the years to come. Actually this week, as you probably know, Samsung had their investor forum and they had a few interesting comments and actually a few interesting slides on EUV. And they made it very clear that EUV for them is an integral part of their road map all the way through 5-nanometer and even beyond 5-nanometer and that's still a long way to go because currently, as you know, we're at 40-nanometer. So 5 nanometers and below 5 nanometers nicely gets you, I would say, into the next decade and they made it very clear that EUV was an integral part of their road map to get there. So what that tells you that is that customers believe that geometric shrink in using EUV and actually also talking about High-NA as a prerequisite to get to the nodes far out are an integral part of their road map. That said, of course, they're not just looking at geometric shrink. Of course, all the customers are also looking at device architecture at advanced packaging and what have you in order to continue to make innovations cost efficient. So of course, they're looking at 3D DRAM. I mean is 3D DRAM going to happen. I think the jury is still out. If you listen to Naga, the Deputy CTO of Micron. He recently gave an interview where he said, I know a lot of research is being done, but I cannot connect all the dots yet. So there is still quite some work to be done before you can realistically say that this is going to happen. So that still needs to be confirmed. And if it's going to happen, there are still many ways to get there. And there are many different choices and options that you have in order to get to 3D, and the ramification for litho, no changes from 1 option to the other, so that's the second one. And if you look at the current research papers, you do see all those multiple options currently being studied with different degrees of implementation difficulty. And then the third one, when is it going to happen? And there I think we made it very clear, also on the back of the comments I just made on the road maps and the role that EUV and geometric shrink plays there, that we believe it's extremely unlikely to the point of being not going to happen, that this will kick in, in 2025, as allegedly some people were suggesting. That ain't going to happen. As a matter of fact, I think the rhetoric of some of them has also changed in the past couple of days, as you've probably noticed and they're now talking about beyond 2025, which is a very broad time frame obviously, and I think that is realistic. I would be quite surprised if anything like 3D is going to kick in well before the end of this decade. So all in all, I think customers are emphasizing that geometric shrink will remain an integral part beyond this decade. I think that's important to note. So whatever is going to happen, it's not going to blow away the field in DRAM. But people are looking at 3D-type structures makes a lot of sense both from a technical and economic perspective, but there are still many unknowns in that regard, if it's going to happen, how it's going to happen and when it's going to happen? From my vantage point will not be before the end of this decade.

Very clear. On gate-all-around, that's another topic that people ask about from an investment perspective. Could you just elaborate on your comments about the lack of a change in litho intensity as you move from FinFET to gate-all-around, just confirm that, and whether you've embedded conservatism on your midterm road map in terms of how gate-all-around actually is introduced.

Yes. So gate-all-around, it's 1 of those device architecture elements that we talked about before that go hand-in-hand with geometric shrink and circuit scaling and all other things that you need in order to get to innovation within semiconductors. So gate-all-around will be introduced by one larger customer, as you know, one foundry customer at N3, and one other customer is considering to introduce it in around 2 months. It has not been confirmed, but it is under study by the other customer, so N3 and N2. In and by itself, gate-all-around is not more or less litho intense than FinFET. So fundamentally, there is no difference in the litho intensity of that technology. The only thing that could happen is that when a customer decides for a particular node to go to gate-all-around, they could decide that at that point in time they're not going to combine that with dramatic shrink. And we've actually seen that in FinFET. When FinFET was introduced a couple of years ago from planar into FinFET, we saw that customers were taking a different approach here. So some, at a point in time when they introduced FinFET, decided not to go for a lot of shrink; others decided to combine it. And this is just a risk management decision you need to take. Am I comfortable enough to combine the introduction of gate-all-around at this instance with a considerable amount of shrink. If a customer were to decide not to do that and just to introduce gate-all-around and not shrink, I mean that's a very small event that would only pertain to that particular year, because in all likelihood, on the next subnode, we talked about subnodes before, on the next subnode, they would probably then go for more shrink. So then they would have a few more layers that they could potentially put in there. So gate-all-around is not in any way any more or less litho-intense than FinFET is. The only implication from going to gate-all-around could be that at that point in time that could be a little bit less shrink, but in all likelihood that would then be compensated in the year following the year of introduction. So all in all, the impact on us, I would say, would be fairly benign.

Turning then to China. Maybe you could just provide some color on how the business in domestic China has been performing in 2021 and how you think it can grow into 2022, given your current visibility? And then the second sort of related question on that would just be around, you say you're fully booked, so that's taken as a given, but even should you face any cancellations, the existence of Chinese demand, does that basically hoover up any cancellations to say, any cancellation that does happen maybe on lagging edge technology is going to be find a buyer in China.

Yes. So currently, if the lands -- I think we have pretty strong growth in China this year. I expect that the sales into domestic China will probably be a little in excess of EUR 2 billion this year, EUR 2 billion. And that's nicely distributed, I would say, over logic and memory, a bit more logic than memory. But in contrast to last year, when most of the sales was logic, I think this year is a pretty strong memory contingent in there and memory is both the DRAM and NAND in this regard. For next year, I wouldn't expect dramatic increase in that number, but we would expect that number to sort of be sustained into next year. So also next year, pretty healthy sales into domestic China, about the same level I would say based on what we see today. In terms of, if demand is going to fold away in some place, would it all go to China? Well, not necessarily to China. I do think, given we are supply constrained, if for whatever reason demand folds away at some place, given how supply constrained, I think the market is at this stage, I'm pretty comfortable that if some part of the demand folds away that we would be able to find a customer for that and that customer is not necessarily in China, it could be some place else. I think the reverse is also true. If for whatever reason, something were to happen on the China front, for instance, more stringent regulation or whatever, the same would apply, right? So to the extent that something would happen to demand in China in the current markets and how capacity constrained we are, in all likelihood that demand would be hoovered up to some place else. So yes, that's the beauty I think of being supply constrained at this stage that demand doesn't seem to be the gating factor for next year.

So obviously, since we're talking on the regional aspect, globally, we're seeing this massive trend for silicon solvency and everyone trying to localize and you already mentioned that sort of inefficient level of spend. So I guess in the near term or even midterm that can be a very big boost towards the 2025 time frame. But could you just describe or give some color to the conversations you're having, maybe, for example, on the European side, particularly around how to actually develop this market, because near term, it's a boom. But then beyond that, you maybe have the hangover of saying, okay, well, if you do really end up with this inefficient spend, the government-directed spending may not end up in the best place, but obviously as a result of a political wish rather than the direct economics.

Yes. And I think what we are emphasizing in our discussions is to that while we understand the tax sovereignty, we understand the desire of the United States and Europe to have more semiconductor manufacturing on the ground, we completely understand, appreciate and support that drive. But we also believe it is important that, a little bit to your point, that that's the players that are considered to do that, the players that are considered for building those fabs. We think it makes a lot of sense to primarily look at the incumbents. And what I mean by that is that we think, if you look at advanced semiconductor manufacturing, we believe it is wise to primarily look at the key players that you have there today and ask them to, or invite them if you like to, to build fabs in those territories. So to be looking at completely new players, which maybe politically some people might want to do, right? So while in Europe, why don't we ask the European players to do it. Realistically, if you ask players that do not have the expertise, that do not have the capability for leading-edge node manufacturing, if you were to invite them to do this, I mean you would be a decade out in all likelihood before you could really see that translate into meaningful manufacturing, which from our vantage point would not be taxpayers' money wisely spent. So we think it makes a lot of sense to look at the incumbent players. And also if you look at the incumbent players to make those investments across the globe, then Dominik, I would expect that they're probably going to be wise enough to not allow too much overcapacity and inefficiency to be built into the system. So yes, that could lead to some incremental spend in the next couple of years. But I think they're probably wise enough not to allow that capacity to be so much over the edge that really a couple of years from now, we would all of a sudden be confronted with a huge air pocket because of the overcapacity that is being done, and that's really the way we look at it and also the way we advise policymakers to look into the matter.

So one of the main surprises in terms of your earnings through 2021 has really been the lagging edge, particularly on the DUV side, and you talked explicitly at the at Capital Markets Day around growing the capacity for dry tools, specifically within that envelope. You obviously mentioned the image sensor dynamic as one area of driving it. So I appreciate that would be part of it. Is that the biggest surprising factor? Or are you also expecting that actually there's a drag or pull up from the advanced nodes, that they also involve, obviously, some dry layers or memory. What's the sort of mix of where that extra dry demand has come from?

Yes. I think it's a combination of the drivers there, and I cannot really give you solid numbers there. But I think it's a combination of the things that you just mentioned, plus one. So I think indeed dye size has been a bit of a surprise and leading to more demand. I think in general, the number of applications around Internet of Things. If you could bail that to the way we look at that today versus the way we looked at that maybe even 1.5 years, 2 years ago, I think that has really exploded, and I think Peter threw out this number where he said, by 2030, we're going to have 10x number of devices connected to the Internet. So that means that every year you would have to build the current installed base of those devices again, and again, and again, and again. So that is, I think, something that has been higher than what we thought. But you're right, I mean also on the advanced nodes, you've seen that the reference size for capacity that we increased that in comparison to 2018. So that means that also on the advanced nodes, we're looking at more capacity than we expected based on the developments in the end markets and also some dry tools have to go in that direction. And then it leads to your point finally also on the memory side, we're looking at a continued strong demand there. So it really is a combination of those that has driven up the dry demand far higher than we anticipated for the years to come, but also in 2025.

And then through the second half of this year, what you've planned already in Q3 and then the outlook for Q4 is a much better installed base management performance. And obviously, we have 2 dynamics there on the service side and the upgrades. One can imagine, in a very hot market next year, maybe upgrades are a bit constrained because of the fact that everyone wants to use all the shiny new litho tools that they bought and maximize cash generation. But particularly curious, where sort of the incremental EUR 350 million of revenue, what particular drove the higher installed base into year-end on that side? Because it's obviously probably actually one of your most visible line items because it becomes a recurring type business, which you can actually put a high visibility to.

Yes, correct. So clearly, the incremental revenue that you talked about really comes from the upgrade business and that has surprised us quite a bit. You might recall when we were talking about the Q1 numbers. This year we said that we pulled in a number of upgrades because that's what we thought, that some things that were actually planned and scheduled for Q2 and Q3, they actually got dragged into Q1. For the reason you mentioned, it was an easy way and a quick way for customers to get more productivity out of the tool and a lot of that is software based. So these were software tools that didn't take a lot of machine time in order to get them to get on process. But Frankly, it's been like that for the entire year. So rather than pull in, it actually turned out that there was continued demand. So there's a combination of, on the one hand customers really looking for productivity upgrades, and this is the only way to increase capacity? If we are supply constrained in terms of turning out new machines, then the only way to get to quick capacity increase of course is to increase the productivity. So that's a quick way to get there. And secondly, we're more and more focused on making these upgrades and doing them fast because we know that machine time, to your point, is precious for the customer. So putting a lot of emphasis on how can we do this rather quickly, and I think that combination of us trying to do the upgrade is a little different from the way we did them in the past, and also the high demand from customers has led to what also was a surprising uptick for us. And this is what we try to continue to do into next year, and that's why I said on the call a couple of weeks ago that, for next year, and also as I mentioned at the beginning of this call, expected 12% increase that I mentioned there for next year as well to get you to sort of EUR 5 billion installed base for 2022.

Then just turning to sort of capital allocation and strategic thinking around M&A. So historically, the companies were very, very focused on consolidating the immediate core competencies and the adjacent areas. So we think about lithography, supply chain to the lithography machine and metrology around that. And, by the way, congratulations to the team for buying and selling [indiscernible] glass for the same price and removing costs that you needed. But my question is, how do you think about or contemplate whether you would ever go down the avenue of buying maybe non-litho exposed businesses that maybe aren't immediately adjacent. If you were to ever perceive that the likelihood is that litho intensity is actually going to step down, because today we're in a world where we're looking at litho intensity anywhere from 20% to 25%. You guys obviously have probably the best view in the world of where litho intensity is going. And the longer-term, 2030 view, is sort of assuming that it stays at that level. But should we have a decline, how would you consider sort of capital allocation at that point?

Yes. So it's implied in your answer, but it's still a long way out, right? Because we still believe that we have some pretty cool plans within our existing business, even keeping litho intensity constant, which you could argue is a bit of a prudent assumption from '25 to 2030. We're still looking at growing the business from EUR 14 billion to EUR 40 billion in a decade, which I still think is pretty cool, right? So from that vantage point and also beyond that point, as we also mentioned at the Capital Markets Day, even beyond High-NA, we still have some ideas about how to continue to go down fast. So I think that will require quite some management attention for years to come. At any point in time, if we think there is a huge opportunity, because it would have to be a huge opportunity, where really we would be able to be synergistic with the capabilities and the technology that we have today, that could be considered. I mean not short-term, because short-term, a little bit to your point, we're very much focused on getting High-NA up and running. So for the next couple of years, we think we're still very, very busy and focused on what we do today. But it might be a point in time when we get a little bit more management time, if you like, to maybe look at a number of alternatives. Firstly, I think that's still a while out before that kicks in. But then I think we would never do it out of negativism. We would never do it saying okay litho intensity is shrinking, so now let's turn to something else. We will do it always on the base of strength. And strength is, is there something out there where our capabilities or our technology can really be helpful to get something to a higher level and get it to a level where our customers and whatever the market would be, would actually be excited about us stepping in. So I think our philosophy would not be, we're just going to buy ourselves into an existing market and then run it in that way, we would always do it on the basis of the strength of our capabilities and the strength of our technology, so we add something to what is not there today. And if we can identify something like that, I would say, in the years to come, that could be of interest. But for the foreseeable future, all the way through 2025 at least, I would say, we're still so focused, so busy getting High-NA up and running that I don't think it's realistic to see us make any move there.

Excellent. Well, I think, unfortunately, that brings us to the time allotted. But that's been very illuminating and a very exciting ending there to talk about the many growth drivers. So thank you very much to Roger Dassen from ASML for participating in the webcast. Thank you to all the investors who have tuned in, and we hope you enjoy the rest of the conference.

Thank you so much. Have a great conference.