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

Good evening, good afternoon, everybody. This is Sandeep Deshpande. I cover the semiconductor space here at JPMorgan, and I'd love to welcome Roger Dassen, the CFO of ASML. Thanks, Roger, for joining us today.

My pleasure, Sandeep.

I -- just before going into my Q&A, just to tell the audience if you have questions, please, go on to the conference website. I'm happy to ask Roger questions on your behalf here. Just moving into -- before into the questions I prepared for you, Roger, I mean, maybe you can give us a quick update. Anything changed since your results in April? How have your customers behaved since your results in April at all?

They've behaved very nicely and politely, Sandeep. We cannot complain. I think pretty consistent with what we've seen in the past couple of months. So no real strong developments on the customer front. I think on our end, I think as we mentioned to you, key objective for us is to see what we can do in terms of increasing capacity to cater to the demand. And I'm sure you're going to raise some questions on that. But that's the key emphasis now. So looking at short-term ability and medium-term ability to further increase the demand. I think that's been the key objective. Other than that, I think, customer behavior, since we last spoke about a month ago, has been fairly consistent. The markets are generally looking very favorable, and customers continue to be quite optimistic about what the future brings.

Understood. So moving on to capacity. I mean, this is the big question in the market. Firstly, let's just quickly talk about EUV. I mean you have talked about being able to do 55 EUV tools. And I guess -- I mean, I assume at this point that you should be able to do that many EUV tools in 2022. Do you need to add further capacity than that at this point?

Yes. So the 55 is the capacity that we're working towards for next year. And I think that's about what we think will be feasible. As you know, Sandeep, it's not necessarily a matter of our capacity right here in Veldhoven. We have quite some flexibility to even further increase capacity. It is particularly in the supply chain that we're looking at. And then within the supply chain, I would probably call out, in particular, the optical columns. So that's size and also the drive laser. Those are the 2 main components of the EV machine that we're looking in to increase. So having conversations with those 2 suppliers to see how they can get to 55. But also indeed, having discussions beyond '22, what can we do to further increase capacity there? So that really is the main focus of us at this point.

So do you think at this point that in '23, you will have more capacity for EUV as well?

I think the first question is, do you think you will have more demand in '23 than 55? And the answer at this stage to that question is yes. It's not like we have POs for that amount. Of course, not. But just given the conversations that we have with our customers, we have reason to believe that the demand for '23 would extend beyond the 55. So that's for us now the exam question. So indeed, if the demand goes beyond 55%, what can we get done to get those manufactured. And as I mentioned, that's not necessarily question for us here, it is a question in the supply chain. So we're very actively working to see, in the next couple of years, what does it take to further ramp the capability of our suppliers, particularly the ones that I just mentioned, to get to that level. Optimistic on the one hand, because we still have a bit of time, particularly if you look at '23, but of course, to the extent that it would require significant investments at our suppliers. Of course, that also takes a bit of time to get that to get that realized. So that's really the journey we're on. What can be done for next year? What can be done for the year thereafter? And is there anything additionally needed for the year beyond that point?

Understood. Excellent. So essentially, what you're saying is, because my next question was that post this 55 tools, will there be a period of digestion. You're saying, no, in fact, it goes up in terms of demand beyond '22, et cetera.

Yes. I mean, we believe, again, based on conversations that we're currently having, that if we were able to get out more than 55 systems that probably we would be able to sell them in 2022. So that means there will be a bit of a carryover effect of '22 into '23. I think the '23 demand, if you just look at the secular trends, Sandeep, if you look at the effect of regionalization, we'll probably come to talk about that, if you look at, for instance, the insertion of EUV into DRAM and the implication that, that has in this time frame, I think we have reason to be optimistic about the demand for EUV also beyond 2022.

Understood. Moving on to the tool itself. I mean, 3600D is going to be coming into production in second half of this year as such really. There is an ASP uplift versus 3400C as such really. I mean, has that given your position in this industry, et cetera, that has the pricing power shifted? I mean historically both you and Peter have always said 50-50 that productivity increase 50% to yourselves, 50% to the customer. Does this shift at all?

I wouldn't say so. I think the pricing is very complicated as you know, Sandeep. Pricing is a -- we really look at what is the value that this tool brings to the customer. So what's the incremental value of a new tool to the customer and then incremental value, we have a formula to arrive at that. But in essence, the key parameters in there are imaging quality, overlay and, to your point, throughput, so productivity and productivity improvement. Those are the key parameters that go into a model. And based on that, we determine what the price is and what the incremental value is, and that then determines how we shift that value to the customer and to ourselves. If we look at this particular tool, if you look at the D tool, as we said, we think for the detail, you're looking at low-10 percentage uplift from the C tool. The base is there for the C tool, and that's something that we also shared with you in Q1. I mean, so far, typically, we refer to the ASP of the C tool at about EUR 130 million. But I also mentioned, too, just looking at all the options that customers are ordering for the tool. Typically, we do see that the ASP really has moved from EUR 132 million above EUR 145 million. And I think if we look at the specification that we think our customers are going to demand for the D tool, I think it's fair to say that, that low 10% uplift that we talk about over the C tool can actually be applied to that EUR 145 million number. So that gives you a bit of an indication of what you're looking at. If you look at the throughput increase for that tool, it actually takes it from EUR 135 million to EUR 160 million. So that's a 19% improvement. So the low-10s improvement over the EUR 145 million that I talked about is a pretty -- is still a pretty good deal, I think, for our customers, still a very fair distribution of the incremental value.

Understood. Moving on to DUV, Roger, I mean when you look at -- I mean, at your units you've done in dry and in immersion in the first 2 -- Q4 last year, and Q1 this year, there seems to be some sign of a capacity limit. I mean I'm just reading out the numbers to you. Immersion, you did 24 and 24, Q4 and Q1; dry -- new dry tools, you did 42 and 39. So you seem to have hit some kind of capacity limit there. So -- and you've talked about in the past that you're considering whether to add capacity in DUV starting next year. Have any decisions been made on that at all?

Yes, Sandeep. We are increasing capacity. By the way, 24 on immersion is a pretty good number, right, because that gets you to 96, which sort of is -- would be sort of an all-time high, I think, in immersion. It's not bad for technology where, at the Capital Markets Day of 2018, we started to think about how is this technology over time going to be sort of replaced by EUV, recognizing it wouldn't be completely replaced, but at least we were looking at that stage at a reduction of, let's say, 90 units in 2018 to sort of 40-ish in the 2025 time frame. So pretty good extended life of immersion, and I think that is true for all of DPV. It is remarkably resilient. But yes, I mean -- so 96 would be a pretty good number for immersion. But we are looking at -- and we are working on capacity expansion. And again, same story as for EUV. It is not just extending or expanding our capacity here. It's also working with the supply chain. But yes, that is happening and that will, therefore, increase capacity for the full year. And then, of course, that increased capacity we would also carry over into next year and would then be available for the entire year.

So you think you can increase that DUV capacity in the second half of this year itself, potentially a little bit?

The answer is yes. So the effects of the capacity increase that we're working on will, of course, fully mature in the second half. And therefore, we'll also carry that fully into 2022.

Okay. And the question a lot of investors ask me, clearly, at the moment, the DUV, TSMC is building a new 28-nanometer fab in next year. I mean, I've never heard of in the past new fab in older technology being built. I mean, I've covered this sector for 20 years as such really. And -- but it is happening as such really. But is this DUV demand, clearly, everybody knows that EUV is a sustainable demand, it's a new technology, but is the DUV demand sustainable or is this kind of short-term big increase and then it will collapse later?

I think there is reason to believe, Sandeep, that some of this will sustain quite a bit longer. And I think what you need to do in order to really understand it is look at what are the driving forces behind the DPV demand. And let's just peel the onion there. I think one element in DPV is the China sales, right? So China sales this year is high. Of course, all of that is DPV because as everyone knows, we cannot sell EUV into China. And that is -- I think that is pretty sustainable what is happening there. If we look at our forecast as to how the sales into China is going to develop, I don't see any slowdown there. As a matter of fact, we see continued growth there. Of course, I think we all appreciate that's dependent on regulatory developments. But if there is no severe limitation in comparison to where we are today, then we think there is good demand and demand increase also in China. So I think that part of the demand seems healthy to us. And then there are 2 other main drivers, I would say, on the DPV side, what is on the mature notes. I think that's -- Sandeep, I think that's the -- if you ask me, if you compare to where you were in 2018 and where you are today, what's the most shocking element to you? Probably, it's this one. I think the demand, to your point, the fact that TSMC is building a brand-new 28-nanometer fab. I mean, if you told me that even 2 years ago, I would have said, Sandeep, come on. Time for you to move on, look for another industry, right? But it's happening, right? It's clearly happening. And that's, I think, because of everything related to Internet of Things. I think the proliferation of more mature Internet of Thing-type applications throughout, so many things happening at a very high pace. So if you look at domestic applications in terms of mature nodes and Internet of Things type, it's is pretty high. I know that's for a fact because we recently renovated their house, and my wife proudly announced that she now has 26 apps on her iPhone to control the house. I mean, it's just one example, which clearly tells you that there is a lot happening on this front. If you look at the semiconductor also on the mature nodes, the semiconductor value in a car, just look at the competitive nature between automotive manufacturers as it relates to their infotainment system. Everything they're doing there is becoming more and more of a competitive tool and, of course, that leads to increase also in the more mature nodes. So I think there is just this proliferation of more mature node application in a whole bunch of applications that we didn't quite frankly see 2 or 3 years ago. So I think that's an important dynamic. And then on the advanced side, the fact that we're seeing such growth in EUV capacity that is being built. We're talking about continued growth in unit numbers in spite of the fact that the -- that every generation, every next-generation of EUV tools actually delivers quite a bit more performance, right? We just had a C to D discussion. You also note that if you're looking to the D to E step that that's another big step in terms of productivity uptick. And in spite of that, in spite of those increases of productivity per tool, we're still looking at increased numbers there. And that is because the wafer capacity that is being asked by the markets is at a much higher level. And given the, let's say, the attach rates between an EUV tool and an immersion tool and other tools, because that is still pretty elevated. Of course, that means that also on the more advanced nodes, you do see quite a bit of DPV sales. So if you combine all of those, China, development on the mature side and developments on the logic side, and you add to that the continued developments in, for instance, the memory market, and I would be pretty surprised to see DPV sales fall off the cliff any time soon.

Excellent. Maybe I will move on from the capacity questions to one that one of your competitors raised in their Capital Markets Day earlier, and I've got so many questions on that over the last 3 weeks, which is, I mean, they say that by 2025, the market is going to go to a 3D DRAM. I mean, we are just beginning to start with EUV in DRAM. So I mean does ASML have any initial thoughts on what was said there?

Yes, we do have some initial thoughts, Sandeep. And the key thing to mention here is that, first and foremost, we are seeking our guidance and information from our customers, and we think we were probably well advised to do that. And if we listen to our customers and we look at the road maps that they are looking at and actually the road maps that they publicly present, they are talking about 4, in some instances, even 5 nodes out that are dominated by geometric shrink in x and y. So if you multiply it at 4 to 5 with a 1.5 year or so cadence, that means that you're looking at 6 to 8 years of road map of our customers that is dominated by geometric shrink. And that is consistent with the behavior that we see, right? Because, as you know, 2 of the 3 are heavily engaged in a transformation from DPV into EUV and are publicly acknowledging the benefits that, that gives them, not just on the cost side, but also on the yield side and the base performance side. So that's what I'm advised by. And then if I look at 2021 and I add something like 7 or 8 years to that, it gets me to 2029 where geometric shrink will dominate. Sandeep, don't ask me what's going to happen thereafter. I mean of course, there are many ways to increase performance on all accounts, including on Memory. So I cannot predict what's going to happen thereafter. But if I listen to the customers and I look at what they publicly disclose, I don't see a road map that says in 2025 EUV is going to be scrapped, and we're going to 3D DRAM. And that probably tells me that EUV is there to stay for quite a while in DRAM. And as I mentioned, what's going to happen in 7 or 8 years' time, I mean, I cannot tell you. But pretty comfortable that with the credibility that we're building up with our customers on EUV insertion into DRAM that, that will have continued value to that.

Moving on to the order intake and your sales this year. I mean, maybe you've talked about that the strongest growth is going to be Memory, but in absolute terms, Logic is going to be the bigger -- then in terms of absolute dollars will be...

A higher base, yes.

Yes, the pie is bigger this year. So can you just talk through, I mean, where is the spending occurring in Memory at this point? Because there's this percentage terms, you're still saying that Memory is your bigger percentage whereas some of the other players in the industry have now already shifted to saying that Logic is driving them as such really.

Yes. And I think the reason for that, Sandeep, is that we actually have 3 growth engines for us in memory. So you're absolutely right in terms of relative increase, Memory is growing fastest. We said Memory, we expect to grow at 50%, taking it from EUR 2.9 billion to EUR 4.3 billion this year. So the growth -- the largest relative growth of all the segments that we operate in. And those 3 growth engines. On the one hand, of course, there is the cycle. I mean we clearly see a recovery of the Memory cycle, in particular, DRAM, I think that's important to note. Because for us, DRAM is quite a bit more significant than that. So I'm not really talking NAND now, really talking DRAM, and that the recovery is happening. And I think there's also a few data points, also looking at spot prices in Q2 that tell you that the DRAM market is pretty healthy. So that gives us a bit of tailwind there. We were already forecasting that for the second half of last year. It came in the second half of last year, albeit less so than we anticipated, but now we see with a bit more force to push that into this year. So that's one force. The second force is again China. So of the over EUR 2 billion domestic sales into China, that used to be last year, a lot of that was -- the lion's share of that was Logic. Now a significant part of that is also Memory. So it's also a Logic, but it's also a healthy Memory contingent in there. This time, it's both DRAM and NAND. So that's the second development. And third development is the one that we just touched upon, which is the insertion of EUV into DRAM manufacturing, which might be specifically to us more so than many to some of our peers. But that is going on, as you know, and our 2 largest -- 2 large customers have been talking about that. And that has manifested itself. And of the total EUV shipments that we're going to have this year, somewhere between 15% and 20% of that will go to DRAM. So that's already starting to become an important part of the DRAM equation. So if you add it all up, those 3 engines, that's the reason why I think we are relatively bullish about the development of Memory. And probably we have some specific reasons that I just mentioned as a result of which now our pattern might be slightly different from what our peers are showing.

Just quickly, I mean, one of your big logic customers has said that their future manufacturing strategy is very EUV-dependent. And what we've heard is that they're going to be more EUV-intensive in their next process, and they were planning to be ahead of the new management change, et cetera. So have you been much more -- your engineers and management team as well, much more intertwined with that customer to help them get off the ground with that process?

Yes. I mean, it's very clear. I mean, they have made public comments that you just referred to about, indeed, doubling the layer accounts on EUV. I mean, it's a customer that we're also -- that we're always very close to. So it's not like we ignored or neglected them in the past couple of years. And they also were a happy DPV customer of ours, and we cherish all of our customers. But clearly, they have communicated this significant change in strategy. They've also shared some very strong ambitions. And of course, we need to work very, very closely with them as we do with our other customers to make sure that they can turn a corner. I mean the ambition is clear. It's also clear that they have a number of issues that they need to deal with. And just as we do for all of our other customers, we help them as best we can. And to the extent that they're fully open and they seem to be fully open to the support that we can provide, we're absolutely there also for them to help them navigate and get to the best possible answer.

Just a message to the audience. I mean, if you have any questions for Roger, please send them through. I've already asked a couple out of there, but I'm happy to ask more as they come through. Moving on to the next question is on the Installed Base Management business. This year, that growth is dilutive to your overall growth as a company. Does that change next year? Or does -- because of this capacity shortage, that is these upgrades are not happening this year, which is probably the reason why it is dilutive and has not been in the past.

Yes. Yes. It is dilutive to the overall growth. I should still remind everyone that we're still looking at a 10% growth. So it's not like there is a standstill, but there is a 10% growth in there as we anticipate for this year. So I think it's good too -- Sandeep, to your point, to first distinguish between the main engines in Installed Base, because there are 2 main categories. One category is the upgrades and the second category is the regular service. Regular service, of course, is heavily dependent on the Installed Base and the growth of the Installed Base. And there, I think it's important to remind everyone that with the continued increase in EUV, you will see a continued increase in the service revenue there. And just the way we've talked about that, if you have any EUV tool that is out of warranty and is working at the target productivity, if you like, then about 5% to 6% of the ASP of that tool translates into recurring annual service revenue on that particular tool. So in that way, I think it's easy to see how, over time, this grows. And that becomes a substantive number, right? I mean, if you're talking about, let's say, 50 D tools, over time, 50 x 6% of EUR 160 million, it starts to become -- that starts to become a meaningful number. So that is accretive, I think, will be accretive year after year to the bottom line and to the top line. So that's one element. So over time, you're talking about hundreds of millions per annum that will be added to it as a result just of that. And that we will see next year as well. So therefore, I think you will continue to see increase at least on this front. Then on the upgrades front, you're right. I mean, it's interesting in terms of capacity constraints, all customers have to answer the following conundrum. On the one hand, I want the tools to be as efficient as possible. On the other hand, I don't want to power down the tool. Ask the ASML team in, tell them here is a powered down tool, good luck with it, and I'll see you in 4 weeks' time, right? I mean they counted the 4 to have the team -- to have the machine out of commission for 4 weeks. So what the customer is looking for is quick upgrades, things that we can quickly address that will give a demonstrable upgrade to the productivity of the tool without tool being out of commission for too long. And that's what we're focusing on. You saw in Q1, you saw that we had a big uptick in the Installed Base revenue. That was also the reason why the gross margin was so high in Q1 as we mentioned on the call. And that was one of those instances where we had a software product that, without having to power down the tool for long, it could immediately increase productivity. That's what isn't very heightened. So we're now really focusing on what are quick wins for customers and what are upgrade packages that can very, very quickly improve the productivity. So we're looking into that. I'm pretty optimistic that we'll be able to come up with a few packages that customers will find interesting even in those days. And that should also lead to healthy sales in next year. So I can't promise you, can't give you a percentage. But again, on the IBR side, it would be looking at continued good growth into 2022.

Understood. There is a question from the audience. Actually, it was related to the earlier one. So I will ask it here. What is your -- is there any implication for ASML as the industry moves to gate all around? And will it have -- will we see more 3D structures in Logic, which have an impact on ASML?

Yes. 3D structures is -- is a little bit further out and a little bit theoretically. So I'm going to confine my answer to gate-all-round versus FinFET. And I think indeed, we do see one customer moving towards that. So one customer moving for the next node to gate-all-around instead of FinFET. We see another large customer not doing that. So you see differences there. But over time, I think it's fair to say that in order to further accommodate the geometric shrink that our customers are looking for, I think it will be inevitable for everyone to move to gate-all-around/nanosheet-type applications. So that will be important. That does not have an immediate impact on litho intensity. So if you look at layer account or whatever, that is not materially affected by that choice. So I think the development into gate-all-around, I think, is positive, because it will continue to allow innovation on the Logic side. So from that vantage point, we welcome it. The immediate effect in terms of litho intensity is negligible.

Understood. Overall, I mean, semi-CapEx to sales over the last few years has increased. Does ASML have a view on whether this is structural or this is cyclical?

I think it's -- if we look at that percentage, it's a very rough percentage because it includes -- if you look at the composition of the wafer equipment that is in there or the fab equipment that is in there, the composition varies quite a bit. I think on our side, I think we continue to see litho intensity increasing node over node quite substantially. I think we had that conversation also at the Capital Markets Day. Of course, we'll update that in September when we have our next Capital Markets Day. So we'll give an update there. But I think the trend towards increased litho intensity node over node, we on that we continue to see that develop. So yes, I think node over node, you're going to see continued innovation there. On the other hand, also the wafers that are being produced are becoming more and more valuable. And I think overall, the Moore's Law -- the empirical side of -- the economic and empirical side of Moore's Law, which is a reduction of the cost per function, I think, still holds. So in spite of the fact that litho intensity and wafer fab equipment goes up, I still believe it makes economic sense to the entire industry to continue to innovate. And I think that's important, Sandeep, to continue to see that pattern further evolve.

Understood. You talked about the wafers, but you don't remember at your Capital Markets Day in 2018, you had talked about a decline in the number of wafers from node to node. And I think you have stepped back from that view in recent conference calls. Do you have a view on what that is going to be? As such, a number of wafers, say, added at 3-nanometers versus 5-nanometers. So do you have a view already or that is going to be updated at your Capital Markets Day?

I think that's a really good idea. Do not steal the beans today, but have that conversation at the Capital Markets Day. In fairness, I think, Sandeep, you're right. I mean the -- we talked about 3 different scenarios, a high, medium and low growth market scenario. And in all instances, we were looking at the decline node over node. And I think it's fair to say that we have not observed that behavior in the past couple of months -- past couple of years, in particular, as a result of the end-market dynamics that we just talked about a few moments ago. But translating that into new assumptions for the 3 different, the high, moderate, low, to make that distinction probably still make sense. But the translation of what exactly does that mean, I think we need to give each other an update on that in the September meeting. But in all likelihood, it will be more optimistic, I would say, than what we presented in November of '18, I think that [indiscernible].

The usual China question, I mean, anything changed in your perception of what is happening? Particularly investors are worried, not of EUV. We know what is happening on EUV, but whether anything changes on deep UV in terms of shipments to China?

Yes, Sandeep, obviously completely dependent on regulatory changes. And I've come to appreciate that they are hard to predict. So I think the reality today, just to level set everyone, as you mentioned, for EUV, we cannot ship because we don't have a Dutch export license there. So we cannot do that. On deep UV, we materially have no limitations there. So we can still ship entire systems from the Netherlands to China. So no real limitations other than a few spare part issues that come out of U.S. The question is, to what extent are we going to see limitations on that front? Personally, I don't think we're going to see some kind of outright ban on DPV shipment into China. I think it's highly unlikely that we'll see that. What I hear more and more and what you see speculation on also in public reports that we see, particularly in the U.S., is that people say maybe we should limit shipments into -- of WFE equipment to such an extent that it will keep China, let's say, at arm's length of the latest and greatest in the industry development, i.e., if the industry is at, let's say, industry and 5, then wafer -- than equipment that allows China to get to, let's say, 10-nanometer or 14-nanometer or whatever is okay, but 7 or 5 is not okay. If that's the -- and I see more and more written about it in that way. If that indeed is the way regulation is going to change, and in reality, what it probably would do for us on the lithography side, is that maybe the latest and greatest in immersion would not be feasible. So 2050, maybe we would not be able to ship to China. But the older immersion technology like in 1980, that has been out of orders 4 or 5 years would still be okay to be sent. I mean I'm purely speculating right now, Sandeep, because I don't know what the regulation is. But I do observe that more and more people are talking about it in this way, keeping China 2 or 3 nodes away at -- 2 or 3 nodes distance from the latest and greatest. If that's the case, then I think the consequence for us would be moderate.

Understood. Moving on to margin. Progress on EUV margin, clearly, we are seeing your margin overall in Q4 and Q1 much better than it has been in past years. Are we already seeing the future because when do we see this crossover between EUV and deep UV?

Yes. So 2 important points as it relates to EUV is EUV system margins. So one is that this year with the introduction of the D, we're going to see that for the D model, we're going to get to about the corporate gross margin. So that's the first checkpoint, so 50-ish, low 50% gross margin on the D sales. So that's the first checkpoint. And the second checkpoint, indeed, is what do you see the crossover between EUV and DPV. And our expectation is that, that will happen with the introduction of the successor to the D, so the E model. So the E model, which as you've seen in the road map, eventually gets you to 220 wafers per hour, maybe not immediately, but with some upgrades, it should get you to that level. So that's a significant step up from the 160 that we talked about for the D. So of course, that will have a commensurate effect on the ASP. Also, the cost of goods of that one will be higher because there will be quite some part commonality between that ETO and High-NA. So we're going to have quite some cost bulk on the anti there. But still, the gross margin on that should be more favorable than on the D tool. And that's the one when that's going to be introduced somewhere in the course of 2023, that will get us to deep UV gross margins.

Excellent. Any last questions from the audience, please put it on, we've got 3 minutes left. I'm just going to go with one question for -- with Roger. In terms of the shareholder returns, I mean, there is a big ongoing share buyback, of which -- EUR 6 billion of which EUR 2.8 billion is utilized. I mean, how should we look at -- I mean, you're producing huge amounts of cash at this point really. So I mean, I mean, is this the way to return cash along with the ongoing dividends improving every year?

Yes. Sandeep, you know our capital allocation policy. And again, this will be a topic of conversation in September. But in essence, it is, as we communicated at last, first, we're going to see what cash do we need in the business. And you see that we're not modest in terms of our R&D expense, right? It's clear that we push down the accelerator in terms of R&D with the commensurate results, right? Because it does generate, I think, good value to customers and to the shareholders that we're doing this R&D and the fact that we're stepping up there, both in EUV and the DPV and coming in High-NA, the CapEx to go with that. Then the next agenda is M&A. On M&A, once in a while, you see us make a small acquisition sometimes in the supply chain, but I wouldn't expect anything big on that front. Sometimes people say, with all the cash that you're generating, isn't the time for you to enter into some kind of adjacent business. And sometimes, we're obviously looking into that, thinking a little bit about it. But frankly, we say first, we want to make sure that High-NA gets up and running. I mean this company has become the company it is today because it's undivided attention on what we do and do that well. And I think that's the way we establish opposition in DPV. That's the way we brought EUV to where it is today. That's the way we're going to bring High-NA to where it is today. And that undivided management attention we think will be required for another couple of years. So 2 or 3 years from now, once you see that High-NA is going to enter into buy volume manufacturing and that we're satisfied if that's going to happen. That might be a point in time that we're going to think about is there a next big thing for ASML that we're going to dive into either a new technology or maybe an adjacency. Way too early to speculate about that. But for those 2 to 3 years, I think it's safe to assume that we will continue to play back the cash that we generate to our shareholders. Beyond that point in time, if there is a great opportunity that we think will be value enhancing, we'll do it. But for the time being, we think many of our shareholders are pretty pleased with the returns that we give them.

There's a last question from the audience. I think we've almost run out time. So actually, I will leave it there. Roger, thank you so much for your time. And always a pleasure to see you. And hopefully, now we'll see you in person in September at your Capital Markets Day.

Very much look forward to that. Thanks so much, Sandeep. Have a great conference all. Thank you. Bye-bye.