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

Technology conference. I'm Rob Sanders, the technology hardware analyst in Europe. And I'm delighted to join -- to be joined today by ASML and the Head of Investor Relations, Skip Miller. Skip, thanks for joining. And for those who are joining by webcast, we are going to do this audio only. Skip? I was just saying welcome to the conference. Can you hear me?

I can hear you. Thank you, Rob.

Great. Great. So Skip, thanks so much for coming again to this conference. You guys have come pretty much every year, I can remember, to the U.S. Technology Conference. And I guess you have had a phenomenal year. You've said to grow sales by, I think, 35% in euros. I guess one question that may be of interest. I'm afraid investors agree to. They want to know is there any more upside to the 2021 outlook. Or are you pretty much at your full capability would be my first question.

Yes, Rob, I think if you listen -- like you said correctly, we were guided the year in terms of sales, up 35%. And if you listen to what we said at the start of the year versus what we've guided in April and then again in July, we continue to take up our view on the year across all fronts, logic, memory, installed base. And so we're now at a point where logic is at up 35%. We see memory up 60 and installed base roughly 15%. So the demand is quite strong. And so we're doing all we can on the output side. So we'll see how hard we can push in terms of getting more systems out. Last quarter, for example, we were able to get another EUV. We took it up a bit from where we were prior. In terms of the growth year-on-year, we went from 30% to 35%. So we're able to stretch EUV and we've taken DUV or the non-EUV business up quite significantly over the course of the year. So it's all about what we can drive out of the factory. So we'll look to continue to do that because the demand is clearly there and see if we can stretch that in time, but that's currently our view today.

Great. And is it too early to talk about 2022? I think you guys have given some vague indications. We do know the EUV units are likely to move up to 55 from 40 this year. Is there anything else you can say about 2022 while we just get that out of the way, the near-term outlook?

Yes. I think 2022, it's a bit early to say any guide on 2022. I think we have talked about a plan to increase EUV capacity in 2022 towards 55 systems. We're also looking to increase our DUV capacity. So we've talked about both those. And you have to keep in mind, again, in 2022, I think from a lead time perspective, EUV being quite long, that we have that pretty much set up at 55 in terms of capacity. We're now looking out at 2023 and seeing what we need to do there and maybe being north of 60 machines, what we're talking about in terms of capacity. But we're putting together demand. We're looking at our supply chain and seeing what we can drive in our supply chain. On DUV, the same thing. We see demand being quite strong. So we're looking to increase our DUV capacity. This year, as we've indicated, it's been quite strong in terms of the demand and what we're able to get out. And to do that, we're also doing everything we can to clean any inventory -- any buffer inventory out to really maximize the output. So if we're doing nothing next year, our output would be lower as this inventory -- obviously buffer inventory is not available so we have to build that back up. And we need to really work with our supply chain partners regarding our DUV output next year and understand when there's a timing factor, whether you're talking cycle time reduction, which is a much shorter-term activity; medium term, it's adding people and equipment into an existing facility; and lastly, if you have to build more walls, and that has a longer time horizon. So we need to put the timing aspect into this as well. And we plan -- as Peter mentioned in our call in July, we need another quarter to do that. So maybe we'll give a bit more of an update on this at the end of the quarter at Investor Day.

Great. And that brings me neatly to the Investor Day, which I think is on September 29, for those of you who don't know. Obviously, highly anticipated I guess what have you said about -- just so everyone understands, in the past, you gave a 2025 sales range. And I think the kind of indication on gross margin, et cetera. How are you going to think about it now, now that we're only 3-and-a-bit years away from 2025? Will you give a longer range outlook? Or will you just so kind of narrow the range or -- I mean, obviously, you're probably going to increase the range, the midpoint of the range, but will you kind of rethink the scenario that you did last time, which was quite wide? I think investors were a little bit like you could drive 3 buses through that range. So maybe if you can give an update on the 2025 thoughts when it comes to your CMD.

Yes. So yes, first off, regarding our Investor Day, as you pointed out, we're still planning to hold it on September 29. We were hoping to have it in person and we tried very hard to do this. As we know, many we're looking forward to having some in-person interaction and get away from the virtual. But recent changes in the COVID-19 restrictions have led our staff to make the decision that we need to change the format of this event from an in-person to fully virtual. So we communicated that this week. So unfortunate, but we'll make the best of the virtual world, and I know many of us had to do that over the past 1.5 years. In terms of what we plan to update, yes, we plan to provide an update of our longer-term strategy, obviously, which will include an update of our current view on 2025 scenarios that we first communicated in 2018. And as we get closer to 2025, I think it's reasonable to expect but the range of these scenarios would narrow compared to what we provided in 2018. So that's one. And regarding longer-term view, meaning beyond 2025. I think it's still a bit -- it's being discussed on how much we will say beyond 2025. So that's still in discussion. And so if you just be a bit patient in only 3 weeks away, we'll have clarity on all the details here.

Great. And it does look like when you think about up to 2023, as you mentioned and touched on earlier, it looks fairly well looked up now. I know the backlog might not cover it, but I'm sure -- I assuming your conversations with customers maybe that it's pretty much fully booked up out to 2023. so can you discuss the key bottlenecks that are in your supply chain that could impact your ability to raise your EUV to a capacity beyond the kind of 60 level and how quickly you can get these out to 70 or maybe even more.

Yes. So a first level set on EUV, the demand is, yes, clearly quite strong, as you rightfully noted, with an EUV backlog of around EUR 10.9 billion as of the end of Q2. That covers roughly 80% of our planned output capacity for next year. Again, we're expecting to see this demand -- strong demand continue based on all the drivers, both the logic nodes as well as now increasing demand in memory. And so we're looking to increase our capacity in 2023 greater than 60 systems. So I mentioned that earlier. The key bottlenecks are currently our optics and our source supplier. As the lead time is longer, as I mentioned earlier, we're working with our supply chain partners to determining what is required really in terms of capacity. So we are looking along these 3 areas where there's a short, medium and long-term cycle time reduction, people and equipment and the expanding square meters. So we have to put all that together, and we'll see again what we have by the end of this quarter and communicate more detail regarding our capacity plan. One other thing I think you also have to keep in mind as it relates to EUV, but also DUV, but since we're talking EUV here, is that we are continuing to increase the productivity of our machines, which is effectively delivering additional wafer output capacity to our customers. We're currently in transition from a 3400C to a 3600D D as in David system, which delivers an additional 15% to 20% productivity or you can say it's effective wafer output. And then we have another -- our next model that has higher performance, including productivity targeted for mid-2023. So that piece of it, you'll have to keep in mind, as we increase the productivity, this is effectively increasing the output capacity. So you have to take the 2 together units and productivity into consideration when you think about how this future demand will grow.

Got it. And then one thing we hear from the EUV supply chain is the demands from suppliers are increasing regarding sort of greater derisking of their capacity additions. So would you be willing to enter into any kind of take-or-pay type agreements or prepayments in order to reduce the risk of ongoing bottlenecks going forward? And could you offset this perhaps with more upfront payments from your own customers?

Yes. I think we mentioned this -- I think it was a couple of quarters ago or maybe even towards the end of last year that with the new EUV contracts that we negotiate with customers now, that we're incorporating down payments into this for EUV. So this is already in process.

Got it. So switching gears to High-NA EUV, it sounds like, from what we're hearing, the initial data has been quite promising in terms of throughput even if there was that original 2-year delay on some of the supply issues out of your [indiscernible]. So what level of High-NA throughput in wafers per hour does a multi-patterning step on EUV makes sense to kind of move to High-NA. And what are you seeing in development versus this kind of threshold productivity level?

Yes. So maybe first off, level set the value, the value in High-NA is really in the process simplification, similar to what EUV low NA did in terms of its value proposition as it relates to the replacement of immersion double patterning of process implication for the -- at the starting to 70 nm node, which this translates to lower cost, reduce cycle time and improve device performance or think of it as yield. Without it, you would see more and more use of multi-pattern and immersion before EUV Low-NA came along. And the same would be true on EUV Low-NA multi-pattern and that will obviously increase process complexity if you don't have High-NA coming in the 2025 time frame in terms of volume production to reduce that cross complexity, go back to single exposure. So it's basically an enabler effective scaling into the next decade. So I would say it's very important. In terms of the High-NA program itself, it's making a good progress. So we are still planning for insertion of High-NA and volume manufacturing in 2025. With the planned productivity of the High-NA machines targeted for HBM in this 2025 time frame, there's compelling value in replacing even EUV double patterning with single exposure High-NA. So I think the productivity is there to meet those needs. And just as a reminder, High-NA and Low-NA, they will coexist along with DPV on these future nodes. Obviously, High-NA will be one of the most critical layers to avoid multi-pattern and allow continued scaling. But there will still be a significant number of EUV low-NA and DPV layers as well.

Got it. And can you just remind us about the kind of alpha tools that customers have ordered? Are they basically inside facility in Belgium? Or are they going to be -- are they being developed at customer sites? I guess, it's such a colossal machine. Perhaps they want you to kind of prove it out in a way that you maybe you didn't do with the where you had Alpha tools being shipped to customers earlier?

Yes. No, the way it will work, Rob, as you may recall back in 2018, we received orders for 4 High-NA machines from 3 customers. So those machines will go to customers facility. I think the discussion that we've -- again, we discussed in the past, is in terms of integration, you can integrate the modules at the customer site versus our factory, and we communicated a few quarters back that our plan now is to do the integration at our factory like we do on our EUV Low-NA and then ship the machines to the customer site or a that would mean that 2023 timing to ship those machines. We'll have customer -- we'll have early access at our facility in Veldhoven before then, but we will be shipping these 4 machines to customers.

Got it. And just to wrap up on High-NA. I guess there's a question of potential cannibalization of EUV. I don't know when that might kick in. And then there is a question of -- so how many fabs do you think will be ready to kind of receive High-NA tools given the size of double-decker buses? And how quickly do you think you could get to sort of 10 machines per year or something?

Yes. So first, your first question on EUV High-NA versus Low-NA, again as you go out in time, the target obviously would be High-NA would be target the most critical layers where you could have some EUV Low-NA double patterning without High-NA. So that will be the first place. Yes, there's still a significant number of Low-NA layers out there, but we'll talk a bit more about the layer development on the outer notes here at the end of the month. In terms of planning on High-NA, I think we've been in discussions, as I mentioned earlier, we had [indiscernible] back in 2018. So we've been in discussions for quite some time with customers in terms of planning for High-NA on future nodes and when they expect to bring those in. So they'll plan their fabs accordingly to accommodate High-NA? Yes, they are a bigger machines, yes, they are more weight. And so you need to structurally plan for that as well and customers are doing that. So I think that won't be an issue. Regarding the unit numbers, our -- we plan to talk to -- if you recall in the scenarios that we had in 2025 and the analysis we provided in 2018 that included High-NA. So we'll provide an update on that and that's, again, we talked about the initial start of High-NA and volume manufacturing starting in 2025. So that would be a start of that timing in these scenarios. And so we'll update those scenarios with the latest in terms of our view on how High-NA will play out in that. So again, 2 updates here and 2 more weeks for both of these.

Got it. And there shouldn't be any repeat of the whole revenue recognition discussion presumably with these tools. In the past, you had shipments and you had revenue recognition and they weren't totally lining up. Is that less likely to be an issue this time around?

I think if you look at that -- the rev rec discussion, that came from an accounting requirement that set a new technology, you need to demonstrate that these machines will, in fact, perform to their specifications. And the intent of that, obviously, is so that machines are not returned. I mean they don't work and then you return and you've already count them as revenue. So on new technology, you have to demonstrate through a number of systems that, in fact -- they can, in fact, there's a predictable shipment qualification and acceptance by customers such as that risk is not there. Now clearly, I think the 20-plus years, I asked about I know we've ever seen one come back. So I think the accountants don't know that. So I don't know yet if we'll have to go and what steps we'll have to go through in terms of this proven this new technology before or is it going to be okay to recognize revenue at shipment. But my expectation, I think, is -- our expectation is that it may be revenue after qualification at customer site for the initial machines.

Got it. Let's move gears to this hot topic, 3D DRAM and EUV, which has been well talked about in the market. So how do you see the risk on your EUVs business if 3D DRAM is successfully introduced in 2026? And how much more DUV business could maybe compensate for any reduced usage of EUV and DRAM?

Yes. I think so -- first off, there's been discussion for quite some time about the end of DRAM scaling. So I recall not so long ago that we never get below 20 nanometers. So I think if you look, the industry continues to innovate and our customers find ways to continue to shrink the 3D DRAM. And if you look at the latest communicated public road maps around DRAM, you now see that they have several nodes or 4 more nodes where there's generations of 3D DRAM in front of us. So if you take this into account, that moves you more towards 2030 versus 2026. And in that, it also includes EUV layers and growing EUV layers in time. And I think now all 3 major DRAM customers have EUV in the road map. So I think that's, again, suggest that they're bringing EUV in their fabs. So clearly, they don't -- they wouldn't do that it was for a free node. So I think they have a see a use for that in terms of extendability. In terms of next generation memory. I think customers have been looking for quite some at a number of options for next-generation memory. This -- and it continues to be pushed out in time. As I mentioned not so long ago that discussion was quite hot. So we thought that was the end of 2D DRAM. And so we were looking at all kinds of alternatives, including memory technologies like ReRAM or MRAM or PCRAM. So there a number of those technologies whether being considered that are, you can say, truly 3 dimensional scaling, not just vertically, and that they can shrink in x and y as well as z. And then cases like that, if they scale an x and y, then obviously would need EUV. And as you scale and, z obviously, you need more -- so they can be quite litho-intensive. And again, this alternative is the stacked approach obviously, is being discussed as well as one of the option. So I think the way it stands today, we're continually discussing with our customers about their future road maps and we'll see how this next-generation technology develops in time. But I think right now, it's a bit difficult to make a call if any uncertainty what that next-generation technology will be out in the, call it, somewhere close to 2030 time frame.

Got it. And the other sort of controversial topic has been around scaling, maybe slowing down or some people even saying, I think that it could have even grown into a hole as we move to 3D architectures. And even -- has been talking about this at the ISSCC conference. The Chairman there was talking about the role of scaling kind of diminishing. I was just I'm just wondering how you see the risk to your business if innovation kind of moves in a vertical direction. And is there a kind of scenario where you might wanting to grow the WFE spending picture at some point? Or do you think the innovation will continue to drive higher ASPs and therefore higher line litho intensity versus overall WFE spend?

Yes. I think our customers continue to innovate. Again, the industry as a whole, continue to innovate. And I think many of them would argue Moore's Law is alive and well. And part of that, obviously, is lithography. I think you can always argue the scaling slope in time. But it requires innovation on all fronts, not only lithography, but also device architecture materials, process tricks, packaging. And so we'll continue to do our part by delivering lithography innovation with the -- on all fronts, the DPV, EUV and our application products that will enable cost-effective scaling. And so we'll continue to be scaling in time. Again, you can always -- and we see that, by the way, out beyond 2030 with the introduction of High-NA. So we'll continue to scale. That will be a contributor to Moore's Law. And I think that involves -- if you talk about innovation on all fronts. One of the items in logic that comes up is device architecture like gate all around, and you talk about improving. This is all about improving device performance in electrostatics at these smaller features. So this is good for the industry. It's good for lithography and that it helps ensure our future shrink road map. Just like the transition, we went from planar to FinFET, it took us from, what, say, the 14- or 16-nanometer node down to the 3-nanometer node. Gate all around will enable continued shrink towards the next decade, and as part of that you'll have continuation of the EUV Low-NA. It will bring High-NA to contribute to that. So all those will help drive Moore's Law. And we'll take -- talk about a longer-term view regarding our view on lithography -- future lithography growth during our Investor Day.

Great. And maybe switching gears to the memory customers. You have seen the good spending trends from those guys. And but their spending does tend to fluctuate more than the logic foundry. So what are the latest current demand signals you're seeing from your memory customers given there seems to be some uncertainty on the second half from what some people are saying, although it's a little bit of a mixed picture I think?

Yes, exactly. A mixed bag is probably a good way of saying it -- there's some that argues overblown. Others say, again, there is a concern. So I think maybe our view today is we continue to see strong demand in memory with the second half is going to be significantly stronger than the first half in order to get to the memory growth on the year for the total memory growth on the year of around 60% for the year. So if you look at what we've delivered in the first half and to get to 60%, obviously, the second half is going to be quite strong. Customers continue to say they expect tight supply/demand environment in the second half and even continuation into next year. I think if you look at the DRAM bit demand growth, in particular, is it went from the high teens at the start of the year to 20 and then to over 20%. So customers are clearly indicating that the strength has picked up in time. And I think if you -- and if something additional, maybe to support this, I think if you look at the end market applications, that are driving the strong logic demand. And by that, I'm talking about the 5Gs, the AI, the high-performance compute, the edge computing all these are really fueling this quite strong logic demand. They also need memory. They cannot operate without memory. And the industry is underspent. You could argue on memory of the past couple of years. So it shouldn't be a surprise that the customers need to do the technology transitions and add some capacity to support this demand. On the supply side, I think the other thing you have to take into account is that some people ask and reference back to our 2018 where we had a high memory revenue number, and you said you have to take into account that the spend, the capital spend per wafer today to build a node is higher than it was a few years ago. I mean it clearly makes more -- cost more to make a way for -- there's a lot more bits on a given sort of wafer. So the bit density has increased and so the cost per bit has come down, but cost to build that wafer is higher. So that's one you have to take into account. The other is that if you compare today versus, say, a few years ago. We're also seeing EUV -- more EUV adoption in memory. And some of this will be, obviously, for new technology that won't serve near-term bit demand. And then lastly, you also have some China memory spend that is not you can say yet producing all of it producing qualified bits. So if you look at our 60% year-on-year growth in memory you take into account, some of this more, call it, strategic spend. You get closer to 30% to 40% year-on-year growth regarding revenue on memory. So that's to meet this near-term bit demand. So that seems more reasonable. And does it seem so you could say out of line or shouldn't it be concerning when you take these items into account.

Got it. Great. And I guess the last point, and I'm pleased to do some of your questions, and I see a couple of started to come in. The last question from my side would just be about the installed base business and how that affects your gross margin going forward. Clearly, you've got a very high wafer potential for gross margin to go up considerably given the EUV deals that you're signing, but maybe you just can go into that how you're thinking about that.

Yes. So for installed base, I think if you look at that, this is a reminder, if those aren't fully real familiar with the installed base, we report that as a separate line item, and that includes within that bucket, last year, it was a total of $3.7 billion. This year, we expect it to grow 15%. And in that bucket is a combination of services and upgrades. Services, as you can imagine, is basically maintaining the machines so they -- at customer site so they perform and meet the performance and output to drive the cost-effective solutions that our customers need and that people parts. On the upgrade, its about extending the life of the machine. So its capital efficiency for our customers in terms of value. And so upgrades can be a hardware or software that and roughly we see in the past there is a fairly even split between source and upgrades, I think though if you look at the last year -- past few years the services have been higher than the upgrade business. But on from a margin perspective, the upgrade business has a higher margins or above the installed base line item of margins and the service is below. So I think if you then drill into it say how that work in time, upgrades will continue to stay as margin accretive business. The service though you have the EUV piece which is really the significant growth in terms of contribution going forward was negative as we entered last year and it turned positive over the course of 2020. we expect that EUV service business to not only to grow in terms of revenue, as a reminder it's a bit different in terms of DPV where it's be all about selling you can say parts in labor. On EUV, it's more value based in terms of sharing of output driven models. so the more wafers or customers are able to produce in terms of output, the more we're able to drive the machine output, we sharing that value and so it's a different model in the EUV versus DUV. That initial investment to put that in place was why we were negative at start of 2020 and we weren't generating a lot of revenue. The revenue comes in the form of post warranty as customers start ramping the machines in volume. And that will translate to margins that will move towards corporate gross margins over the next 3 years. So again, I think that installed base has a good trajectory for margin improvement as it relates to that line item, but also a contribution to our corporate in time. And again, that's another item we'll talk to in a bit more detail in 3 weeks.

Got it. And the simple math is, is it 5 million to 6 million per installed total of EUV after the warranty of 2 years has elapsed? Is that how we should think about it?

Yes, roughly, to just to clarify that as 5% to 6% of the ASP what we currently provide in terms of a guide on the revenue piece.

Got it. So we can do the math basically. Okay, great. I just have one question from the audience. And I just want to take it when you just -- I know we're running on the time. But do you expect -- this is a question from the client. Do you expect a more steady profile of High-NA high-volume shipments after initial R&D shipments in 2023 versus what happened with Low-NA? So do you want me to repeat that? Do you expect a more steady profile of High-NA? Okay, you got it?

Yes, I got it. Yes. So the question is what will High-NA profile look like in time. I think we will see the start, and so we'll first communicate what we expect in terms of unit shipments in terms of 2025, which is our timing for high-volume manufacturing. And I think if you look out nodes, and again, we'll talk a bit more detail on future nodes in 3 weeks, but we expect that the number of EUV layers to scale in time both in terms of the High-NA, the critical layers, but both in -- and that in terms of total. And then usually the way it works is that the critical layers will move up as you continue to scale in time. So you see a richer mix of more EUV and immersion in time. But we'll provide a bit more detail on how we see that evolving here in 3 weeks.

Great. I think we just run over. So thank you again, Skip, for this very informative discussion. And thanks, everyone, for joining, and have a great conference.

All right. Thanks, Rob. Everyone, take care. Thank you.