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

Hi, everyone. I'm Alex Duval. I'm the Head of Europe Semis research at Goldman Sachs. Delighted to be here with Peter Wennink, the CEO of ASML. Thank you so much for dialing in today. And really appreciate, Peter, you joining us.

It's a pleasure, Alex.

Great. So maybe we can kick off straight away with some questions. And I guess, first of all, maybe you could set the scene for us a little bit for 2021. How should we be thinking about the key trends as we look at areas like Logic and Memory? How do you think about the near-term drivers in 2021 for those areas across your business?

Well, thanks, Alex. Well, it's a very relevant question because the news we're all reading is about semiconductor and chip shortage. So what does that mean? I think we need to realize that I think there are 2, that the current demand for product for ASML is driven by 2 trends. I think one is I would call an upgrade on the secular trend. The secular trend, of course, we already knew it, that we're in a growing business. But I think what the 2020 -- the year 2020 and the COVID crisis has taught us is that the secular trend, I think we're underestimating it because we're now seeing the benefit of the digital revolution, I would say, the additional transition and we're experiencing ways of communicating and delivering services and products that have accelerated at a very high speed, and it's not going to go away. So I think that will stay. That secular trend is definitely there, and that's a big driver of what we see today. On the other hand, we also see, on top of that, we see the effects of the 2020 concerns that our customers had starting at the end of Q1 and trending into Q2 and Q3, and in the first month of Q4. I think it was only until November-ish time frame that we really saw the demand picking up. And why is that? Because they basically effectively at 8 months or 9 months of concerns about the impact of the GDP growth we saw export control measures being taken, for instance, Huawei, which had an effect on the capacity planning of TSMC. We saw the news out of the United States, coming from big microprocessor maker, Intel. It all had its effect on the toning down, I would say, quite significantly, the capacity indications for 2021. Well, if you do that for 8 or 9 months, and then suddenly you come back and you -- and basically, the estimates, the early estimates on GDP shrink were exaggerated. And finally, in the end, and especially for our industry, it turned out much less worse than people thought. So we underinvested and that comes back. So that is -- on top of that, there is a higher secular trend. That's what we're seeing everywhere. I think memory that we see is specifically illogic, and it's across the Board. It's across the Board. It starts with mature technology, 90 nanometer, 65, 45, 28, it's everywhere, up to and including, of course, the leading edge. So the fear of not being able to fill the hole that was created by Huawei on the leading edge quickly filled up, filled up with all the customers. And now on memory, memory is trending the way that we anticipated. I think memory customers, and especially in DRAM, which is basically a virtual oligopoly, they actually started to manage the supply-demand balance by the idling tools. And the supply-demand that is starting to recover in the course of 2020, and we actually said throughout 2020 that the utilization rates we saw in the installed base in DRAM were going up, and we anticipated by the end of the year that would reach about maximum, which was about right at the end of the last year, early this year. But our memory customers, DRAM customers specifically, they anticipated this and they asked for machines in Q4 last year, which are being sold as we speak, which will generate in our output in this quarter and early Q2. Now if you take 20% bit growth in 2020, then that capacity addition that we shipped last year, which is now being put to work, is not enough. So we definitely need to see more shipments, DRAM shipments throughout the year to make up for that 20% bit growth. So that means that, yes, we have a very strong and not only bullish, but I would say, as a logic industry that is under -- that's in undercapacity, clearly, we have -- as a DRAM industry that is set to order more throughout the rest of the year for shipment throughout 2020. And I think somewhere down the line that 3D NAND will also follow, which is not at the same level of, I would say, utilization as we have seen with DRAM towards the end of the year, early this year. So they still have some, you could say, capacity buffer from a little point of view, that needs to be filled before we see -- we're in same territory as we are with dealer. So that's the picture. And that picture actually shows quite a significant year 2021 for the equipment industry.

That's very clear. So that sounds very exciting, obviously. And I guess, just as we sort of think about quantification, one thing that sort of struck us is obviously some of your peers in recent weeks. They talk about sort of mid-teens kind of growth for wafer fab equipment this year. And obviously, given the EUV story, we think of ASML very much as a share gainer. And I think you've also talked about EUV growing 30% this year, which is actually an upgrade on what you had before. So if the market is growing that much and you're gaining share, how should we think about the sort of 12% roughly that some people think is implied by your guidance? To what degree could there be upsides to that? And what kind of conservatism may or may not be baked into that at the moment?

Well, there is conservatism in there, and that's clear. And that's on 2 fronts. I think it's on logic. I think it's on mature logic, so there's not EUV, it's deep UV. EUV will be limited by our supply chain. It's deep UV, particularly China, but also now popping up in other places. And it has to do with the shortage that you can read about. So it's not only 7 or 5 nanometer shortage, it's across the board. That's definitely upside. That's still continuing in terms of demand. Now China is pretty important there because coming out of 2020, and especially the last couple of months of the former U.S. administration also presented us from time to time with some unexpected regulations coming out. And it's a bit skittish of what does that mean going forward. And we've been, I think, a bit cautious on guiding the Chinese logic market. Now under the assumption, as we said during our call, under the assumption that those export control rules as they are today, yes, we definitely see -- and it stays this way throughout the year, that means we can ship certain level of deep UV tools to China, which would actually be upside to the 10% that we guided. I mean, if you think about 600 million, 700 million on top of that, yes? On top of that, in DRAM, I just mentioned to you. I mean, our DRAM customers are customers that very closely look at the market dynamics. They are not like leading edge lethal or let's say logic maker because that needs a plan 1.5 years ahead in terms of capacity addition. They follow the market more closely. And of course, since it's largely deep UV, it also means that shorter-cycle times, shorter interval lead times for us. So we can actually follow the demand that's popping up now for shipment in the second half of the year, we can actually produce it. So that's another upside. And specifically, in DRAM and towards the end of the year, it might be 3D NAND, but I think this would be too early. So I think there are 2 categories of upside. One is, I think, deep UV mature logic, China, but also now starting to appear outside China and DRAM.

That's very clear.

Both deep UV.

That makes a lot of sense. And obviously, you see customers raising CapEx and so forth. But some investors do ask us, the world feels very good right now. But are there risks of cycle overheating? Or double ordering, these kind of cyclical risks? Is it really just too early to be thinking about that? How should we sort of conceptualize where we are in the cycle right now? Is it just that the structural trends are so strong as well as the cyclical?

Yes. I think the cyclical part is definitely not investing in 2020 creating under capacity, you could call that cyclical. That is just a short-term movement driven by external factors that turn out to be different. And then people say, "Well, oh, we got to grow, we need to do something different." I think the underlying secular trend is very clear. I mean, if we think about the demand for 90 nanometers, 65, 45, 28, it's microcontrollers, it's image sensors, it's image signal processing. It's a lot of stuff that actually doesn't need 7-nanometer and 5-nanometer, but it's fueled by it. If you think about the driver assistant features, it is not only -- it is, yes, it's 7 nanometer, and it's 5 nano, but it's also very much the other stuff. It's the analog translation into digital, which were just not 5 nanometer. So it's kind of the leading edge is the catalyst for the semiconductor demand that's growing in the mature business, and it's not that there's 1 or 2 players that are investing a lot in the mature business, it's many smaller customers, European customers like STMicroelectronics, Infineon, NXP, UFC, they're everywhere, that add a few tools, but many customers adding a few tools and wants some extra capacity adds up. And it's next to the catch-up of the underinvestment. It's really the secular trend. And is driven by the fact that we have 5G that we are rolling out 5 and 7 nanometer, which is fueling or exploring the demand for the mature logic. These are system solutions, which actually affect all the nodes, yes? Not just the leading edge nodes.

That's really clear.

So yes. There's some short-term-ism in there, but I think the underlying trend, which is the most important trend, is more secular. It's more long term.

That's great. And obviously, you're talking about sort of broad-based demand from the more mature nodes up to the sort of leading-edge nodes. If we think about leading edge and sort of technology, obviously, EUV very much plays into that. I wondered if you could talk about the sort of latest improvements on the EUV technology that you're really focused on? And what the sort of feedback from customers is there? And clearly, you have an e-model that will be sort of becoming more and more prominent. Can you talk a little bit to the sort of financial implications of that as well in the context of those EUV technology improvements?

Yes. I think if you look at the cadence of technology improvements, it's about every 2 years there's a new model. And the new model should provide much more value to our customers because why would they need a new model otherwise. So it's about higher productivity, but also very important is not only higher productivity. The uniformity of the critical dimension, which is an optical feature, is also very important. So CD uniformity is also very important with the next node because it keeps shrinking, things are getting small. But also then, if you do that, overlay becomes critical. So the combination of overlay, CD uniformity creates this, what they call, this edge placement era, which is a very important driver for yield and therefore, cost. So this is what every new generation, the specifications are tighter on everything, more productivity, better CD control and better overlay. And that's what you'll find back in every new generation. So the D version, which we're going to ship in the second half of this year, well, in 2 years after the second half of this year, so we talk about the second half of 2023, there will be a new version, it's called the E version, let's call it the E version. And 2 years later, in 2025, you see High-NA. So it's this 2-year cadence, which actually follows the cadence of leading logic makers and not particularly TSMC. So it's all driven up because we create it because the customers demand this. So we come up with these solutions to cater for the requirements that customers need for their next nodes. That's how it works. So it's this 2-year cadence. That's what we are seeing. And we -- what you will see is, of course, productivity advantages, but also overlay, which will create more value. Like you've seen, the ASP going up from about EUR 130 million to EUR 160 million coming from the C to the D version, which is about a 50-50 split this year. But in 2022, you won't see C versions, there's only D versions. And in 2023, you'll probably see D versions split with E versions. And so this is how it is, about a 2-year case.

Got it. That makes total sense. And I'd love to talk a bit more about EUV and the plans there. But before we come to sort of unit shipments, and those kind of topics. You mentioned China and obviously, you talked about the implications for this year. But as we think longer term, obviously, China being keen to sort of build out their capabilities in the semi space. What does that mean for sort of ASML? Should we be thinking leading edge nodes? Should we be thinking logic memory? How do you think about that TAM in terms of those Chinese domestic players in the longer term?

Well, I think, medium term, the next couple of years, I think China, where they are today, where Chinese customers are today, they'll have a significant task in just catching up from, let's say, 28 nanometer to 14 from 14 to 10 to 7, and basically trying to do what others have done in the logic space is just push deep UV technology to the limit, which will take a couple of years. So I think it will be predominantly deep UV. I think on EUV, we all know, that has a big question as to whether governments will allow us to ship EUV to China. But I think where the Chinese customers currently are, I think they need deep UV. Many of those customers, like, Wow, they're not a 10 nanometer, they're not a 14 nanometer. I mean, they are 28, they are 45 and 65, and this is where the demand is. So I think China will probably play a very critical role in these more mature solutions. While we are moving towards the, let's say, high ends of the deep UV applications, that's what will happen. Shipping EUV to China is such a hot political, geopolitical debate. It's almost outside -- it is completely outside our control. Now what does that mean? It actually means that Chinese customers will very likely be able to buy products that are made with 7-nanometer technology or 5 for that matter and just be made somewhere else. So I think from -- if you're a key supplier for that kind of technology for an industry that is global, or just ship it anywhere where it is allowed to make those devices. So I think longer term, it doesn't really matter. Short term, this is why we focus so much on, let's say, standstill on the current export control measures because when suddenly this export control measures are created or are changed, it creates disruption, short-term disruption. Medium term, it will just balance itself out. It just moves somewhere else. But shorter term, customers of Chinese foundry, they need to -- where they can't get their wafers made with the right level of technology, they need to start requalifying with another supplier, which will take time. And so short -- there'll be short-term disruptions. But medium term, chips need to be made, and then it will be made somewhere else. But China, really focused on deep UV.