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

All right. Great. Well, hi, everyone. I'm Alex Duval, I head up the Europe tech hardware team Global Investment Research for Goldman in London. Delighted to be here with Skip Miller, who is Head of Investor Relations worldwide at ASML. So maybe we can kick off just in terms of the outlook for the remainder of the year. If you could give us a bit of color in terms of the various end markets. That would be a great way to kick off in terms of your expectations for the balance of the year.

All right. Thanks, Alex. Yes. So what we basically said in July still remains our view today and maybe talk a bit about what we said at the start of the year, and what we said in July, we had [indiscernible] end market perspective. We said that we would see the memory market grow around 25%, logic around 20, installed base around 10, and the full year, we talked to little over 20%. And that was with EUR 1 billion of delayed revenue due to fast shipments that would end up in 2023. In July, we changed that due to incremental fast shipments. We did plan to do more fast shipments this year, and we can come back maybe and talk a bit about what fast shipments mean, a follow on here. But we basically took an additional EUR 1.8 billion of revenue out of this year and the next for a total of $2.8 billion in the next year. And we changed the view of the growth from 20% to 10% because of that $1.8 billion that went out. And the memory builds around 20%, and logic around 5%, installed base around 10%, because the bulk of that being more of the EUV had a slight skew towards the logic side is where you saw more of an adjustment on that front. But overall, the demand picture still remains, as we mentioned in July. We haven't seen any change in that front. And our focus really is on the supply side, managing through some of the supply chain challenges that we mentioned at the start of the year, we're still seeing and expect to continue to manage through the second half of this year.

Great. And you touched on EUV as part of the demand picture. And obviously, you've given this -- well, you've kicked off a discussion about potentially going to 90 units in terms of capacity. So I wondered if you could talk a bit about what underpins that if you think about the secular drivers, but also we've had some of your customers like TSMC talking about investment on a multiyear basis. So the latest color there would be really helpful.

Yes. So we've been talking about in terms of the secular drivers. We started out with 5G being a carrier for AI. We talked about HPC, high-performance compute. We talk about AR/VR being a piece of that. If you look at the automotive segment in terms of autonomous driving and increased automation in your vehicle, but also the EV movement there. In the industrial side, I think the continued automation, Industry 4.0, what you want to call it, I think all those pieces that are -- that we've been talking about for quite some time are still very much in play. You're starting to see that really take off. So we're kind of in the early stages, I think, from a secular growth perspective. that -- all what I mentioned has an advanced aspect to it. But I think the digital infrastructure around that is where a lot of the more mature technologies. You think about sensors, you think about some of the power management controllers that also -- so you have both an advanced and a mature demand growing on that front. So maybe just to expand a bit beyond the EUV side is that EUV is obviously carrying the advanced piece. With each of these nodes, there's a continuation of lithospend, which means more and more EUV layers. So the number of EUV units continue to grow in time in terms of demand by node. But there's also a significant amount of DUV that stays in there. And that DUV demand is not only needed advance but also in the mature as all these secular drivers continue. That's why we talk about and we can talk about later a bit more that our focus is on expanding capacity, both on the EUV front, but also on the DUV front.

Super helpful context. And I guess, at a conference like this, obviously interested to think about 2023 -- and I wondered to what extent you sort of see risk to the picture into the next year. We've obviously all seen the macro data points out there. We've seen some of the sort of weaker consumer data points, which at some point, presumably do feed into semi-cap demand. So sort of what level of risk you're seeing out there? How are you sort of -- what do intellectual framework around that?

Yes. So we talk about kind of a mixed message on that front first. What do we mean by that? Well, I think everyone can see the headlines and hear the different comments with respect to the consumer piece of it, PCs, smartphones and some of the slowing demand in that arena and even some of the customers, in particular, in the memory space, I have commented about what they're seeing on that front. So that's one piece. On the other side, you also see areas like high-performance compute, data centers, automotive, industrial, where the demand continues to be quite strong. So that's the mixed message that we're seeing today. In terms of what it means for this year, I think it's very clear to us that this year is going to be all about what we can supply. Demand continues to be significantly above the supply, what we can supply today, and that's the case going into 2023 as well. So our focus really is on how to address the supply chain issues that we're seeing today in 2022, make sure we can resolve those by the end of this year, early next, such that we can ship a higher capacity of 60 or more EUV machines in next year and 375 or more machines in Deep-UV next year. And that's kind of where we're focused on now because demand in next year is also still above supply. So it's really about all we can supply get out next year. Of course, that's with some assumption that the recession is more moderate. Obviously, if you have a deep ugly long recession, we would be immune to that. So I think that's the dynamic at play today, but customers haven't made adjustments to date in terms of their demand picture.

That's very helpful. And I think when you're talking about secular drivers, you talked about leading edge and trailing edge is to the periphery. And I think one of your peers was talking about potential risks to sort of trailing edge side of things. So I just wondered in the context of your commentary you've given about 375 DUV units for next year, which is obviously up on low 300s, I believe, the year before. How much risk do you see there? And why would your picture be a bit different to those peers?

Yes. So maybe first comment on the backlog that we have right now. We talked about backlog of over $33 billion. In that backlog, 85% roughly of that backlog is EUV and immersion, which would be for your advanced nodes. Now to build an advanced node, you need EUV and immersion, but you also need dry. So if you bring some of that dry from the backlog to support also the advance as well, then you obviously move above the 85% in terms of the total revenue in support of advanced for mature. So from a revenue perspective, clearly, the majority is towards the advanced nodes. But from a unit perspective, there's more on the mature side. And so I think if you look at our -- in terms of the supply-demand gap, our largest gap is really on not only DPV, but on the DPV dry. And that supports both advanced and mature. In terms of that demand, it comes from this broad digital infrastructure that I talked about. We talked about a number of items that are driving that. We'll talk a bit more detail about that in November at our Investor Day. But I think that today, we see the demand for that technology likely advance to continue to be quite strong and stronger than we had expected some time ago.

Super helpful. And we've talked about leading edge versus trailing edge. I guess another way of looking at things is kind of logic versus memory. And I guess, memory was one of the sort of first areas we've seen sort of CapEx pushouts, et cetera. So I wondered if you could sort of help us with a bit of context there. To what degree are you seeing any change on the memory side in terms of sort of overall demand, thinking holistically. And then secondly, I think there's been talk about memory customers adopting some of your more advanced tools. So if you can talk a bit to the sort of level of confidence you have there as well, that would be really helpful.

Yes. So maybe first start on the second part of your question, then you'll go back the other way. I think if you look at memory, you now see all the memory customers with plans to adopt EUV. There are a number of systems in our backlog from memory customers. You see that both in the low NA and in the high NA. So future technology, EUV not only applies to logic and growing layers in logic, but we're also see EUV growing in terms of layers, adoptions in the memory space, but then also continuation into high NA, which maybe is a bit more of a, you could say, a surprise, a positive surprise in terms of EUV adoption and memory, whereas a few years ago, a lot of question marks there. So I think that piece in memory, you have the technology aspect of that. The strategic nature of that investment as it relates to future technology in DRAM. And the lead times on that technology is quite long. So we haven't seen any changes with respect to our customers talking to us about changes of pushouts in that aspect in terms of both EUV but also DPV. Again, I think the long lead times come into play. They need to understand what is the -- how long is this potential recession going to last. They need to look through that. Keep in mind that when they originally placed orders for these machines, they weren't able to get the full demand that they originally requested. So that limited some of that. So I think that's the dynamic at play. We'll see as we go forward in time. Will they change that view. But as of today, we haven't seen any changes with respect to demand in either the logic or the memory segment.

Great stuff good. And I guess the company has given some comments at the last results saying that key partners are now sharing more granularity on their longer-term plans. Just wondered if you could sort of give a bit more color on your level of order visibility and then another question we get from investors is just around the extent to which customers theoretically could push out or cancel orders or delay orders? What's the broader picture you're seeing there?

Yes. So on the first one, we use many sources to try to understand what the demand is, what's the drivers there. You can use third parties, but I think one of the things that we've seen over the course of this tightened supply environment is customers' willingness to, again, be more open. We always have, we'll say, very detailed engagements with our customers as it relates to their demand plans. But I think another level, deeper engagement is what we've seen over the past few years. I think both Peter and Roger had mentioned that on the calls, that, that helps us to not only stand what they need in terms of wafers, but what's driving the wafers. And so it gives you, again, more clarity on that front. I think that's a piece that helps us in terms of setting this future demand. And I think you'll continue to see that going forward with customers as they keep this openness because they want to make sure we can get this demand picture right.

Very clear. And I guess, no discussion in semis when it would be complete without geopolitics. I guess you've talked about a desire to build out capacity to levels for 2025 that are significantly beyond your current abilities in terms of EUV units, DUV and so on. Some of that secular? Some of that seems to be driven by geopolitical driven demand. So I guess, firstly, it'd be great to get your perspective on the U.S. chipset. That's now moved forward. How significant is that? And what do you expect in other regions. And then secondly, there's obviously been some news flow about the U.S. and China and that obviously impacting the semi cap space. So any latest perspective, if you could give that would be really helpful.

Yes. So maybe the first one China, yes, from that U.S.-China situation, as it relates to export control, no change on that front, meaning we still have in place a restriction around EUV. But from a DUV, we can continue to ship DUV machine. So no change on that front. The second piece as it relates to technological sovereignty, we talked about this in our 2021 Investor Day. We talked about the fact that you had a number of countries, U.S. and Europe, working on plans to secure funding for the support of onshoring, bringing back the manufacturing and technology to the region. And -- but yet there was nothing definite at the time a year ago. Fast forward, I think if you look at today, you not only had a number of announcements that were made over the course of the years from customers that were planning to build fabs in the U.S. and Europe, even in Japan. But you also had closure in terms of passage on the CHIPS Act of the U.S., equivalent CHIPS Act in Europe. And that now in terms of materializes fabs actually out there that is earmarked for. And customers are moving towards that. You can go to Arizona, for example, or on ASML office located and you can see the fabs going up there. So it's real. They're moving. And we will talk in our November at our Investor Day about what that means in terms of the impact has -- it's good for the demand. We're going to see that in terms of now realization of fabs. They need litho equipment. So that means more spending on litho equipment in the U.S. And so that's a good thing. And I think the materialization of that now we'll talk more in terms of what impact that has at our Investor Day.

Great. And I guess many hardware companies and semicon companies have been talking about supply side bottlenecks. So I wondered if you could sort of talk about what the latest ones are, what's the latest situation. We've seen news articles talking about potentially very inexpensive chips that might be needed in lithography machines. Could that be a bottleneck? Are there other bottlenecks out there? And to what degree does that create a risk to your EUV shipment expectations this year and next year?

Yes. So our supply chain challenges that we faced this year and are still working through are pretty broad-based. We can't give it one thing one technical issue that will resolve this it will take care of things. And we did even mention, I think, a quarter ago that chips were even a challenge in terms of one of the things that we're managing through. I think that particular issue is behind us, hopefully, it stays behind us. But there's another number of issues that are popped up, nothing super technical, but broad-based in the sense that there may be second- and third-tier type issues down to cables and hoses and valves. So things that are pretty basic but things you need to manage through it. So we're obviously working with our supply chain partners to help manage through that. But I think there's also an added aspect that not only just the normal challenge that we -- that many industries have faced. But we're also asking our supply chain partners to ramp aggressively ramp the technology this year, pull forward, if you will. And so that's put an additional burden on the supply chain. But we're managing through those. And again, I think we'll still have a plan to ship more machines this year than obviously than past. The supply chain is something that we'll have to continue to manage, we think, for the remainder of this year. And as I said, try to get these issues resolved over the course of this year such that we can ship to higher capacity number next year.

Makes total sense. Then if we think even longer term, obviously, the discussion about 90 EUV units for 2025. What needs to fall into place for that to happen? What are the different moving parts? I think you've talked about your supply chain and so on. So what needs to happen there?

Yes. The -- from a capacity standpoint, just to back up and communicate what we said, we were originally planning for capacity of EUV in 2025 of 70 units. And we were going to now plan for 90 by 2025. That was the goal that we communicated in April. We'll come back in November at our Investor Day, and we'll talk in more detail in terms of what we've confirmed the timing and what's required to get to the 90 in EUV, but also on a DPV front, we're planning to take the DPV capacity to 600 by 2025. And again, we'll talk through what's required and what we've secured, not only from within ASML and in the [ Veldhoven ] facility and our different operations like Wilton here in the U.S., San Diego, but also within our supply chain. What is -- what are they willing to -- what capacity will they have and by when? Because a lot of this involves more than just cycle time reduction or adding people or adding equipment, it's going to be about building buildings. So we'll fill the details in terms of what that means as it relates to the capacity of the 90 and the 600, will we have it in the '25 time frame or about that time frame. What's required in terms of CapEx required there? And then the next logical question is, okay, what does that mean for your scenarios? -- in 2025 that you put out there roughly a year ago. And so we'll provide an update on that and maybe even look beyond 2025. That will all be at our Investor Day here at November 11.

Appreciate that. And if we think again about sort of this year, next year, I think one of the dynamics you've highlighted is fast shipments, what that means in terms of revenue recognition, what that means in terms of cash? And then so any update there would be helpful. And then is that the new normal? How should we be thinking about fast shipments as you look further forward...

All right. So maybe explain maybe first off, what is a fast shipment because I don't -- we've talked about fast shipment style for a number of quarters. So on our standard process flow, we put together a machine in our factory. All the hardware is fully built up. And then at the end of that integration, we run a series of tests in our factory. I call it a factory acceptance test. And once we complete that, we -- and they pass the customer signs a machine, we ship the machine at that point in time at shipment, we trigger revenue recognition. We then ship the machine to our customer site. We reassemble the machine, and we run through a qualification test. And then we run the same set of tests at the factory. Again, at the site. We call that site acceptance test. And then of course, the machines release to customers reduction. In a fast ship sequence, we still put together a full machine so it has a full hardware configuration. But when we run the test, we run a subset of the tests. We don't run all the tests. And by eliminating some of those tests or skipping some of those tests, we save roughly 3 to 4 weeks of cycle time. So that allows us to ship the machine 3 to 4 weeks earlier. But because we don't fully run all the tests and the customer sign off, it doesn't trigger revenue recognition until later when they complete all the site acceptance tests. The cash flow in both those scenarios is the same, but the revenue recognition is delayed roughly 12 to 14 weeks until we complete the final site acceptance test. So that process, why do we do that? We save 3 to 4 weeks of cycle time. That's the positive for both the customer and ASML. And the purpose originally running all the tests are to avoid exporting basically problems into the field. And so now that we've done this for a number of months, we initially went through, we talked about the supply chain challenges we were facing, that created a delay on some of our system starts in order to recover, we went to this fast shipment process. Now that we've tried it for a while, actually, you can say we like it. It seems to be working. We're not finding problems in the field. the time from shipment of our factory to qualification to the customer is roughly the same. And so now we're having discussions with our customers and the accountants to say, can we revenue recognize at shipment from our factory even though we haven't run the full set of originally planned factory acceptance test. And that's in discussion now that we haven't completed that yet. But again, that's our objective. We'd like to get trigger off that revenue recognition. So we don't have to talk about this delayed revenue in time.

Great. I'd like to talk a bit about ASPs, average selling prices. So you've been able to progressively increase these over time, and that's really been in line with productivity improvements. So I wondered if you could just talk a bit about the value-based pricing model that you have. Is there any reduced or increased willingness from customers to pay higher ASPs? And then obviously, there's the upcoming 3800E model. So what's that going to bring in terms of technological capabilities, productivity? And how should we think about the value that ASML connects, right.

Yes. Yes. So maybe first, talk the basics behind value pricing. So the way value pricing works is that we have really 3 major value drivers, one being productivity, how fast the machine runs. Another being imaging, the quality and the size of the lines, the smaller features, obviously, in the higher quality lines or better control those lines translates to another value. And then lastly, overlay or imaging placement, the ability to place one layer as you build this layer of 50, 80-plus layer stack of process technology, keeping everything aligned within a few nanometers layer to layer. Each one of these have value to our customer, and then we share in the value as we move from one product to the next. So to Alex's question, as we do that, obviously, that drives ASPs and that allows us to continue to grow not only the ASPs in time, but also improve our margins. If we're able to bring value to our customers and then share in that, which equals higher ASP for us, but the cost isn't quite scaled to that same level. That's where you see these margin improvements in time. So I think that's -- you'll say something that we'll continue to drive future products. And today, we're shipping the 3600D, which has roughly an ASP around EUR 160 million, which we've talked about and 160 wafers hour productivity. When we go to the E, which is targeted for the second half of next year, that productivity will start out around 195 and eventually move to 220. So that has a, you could say, north of 35% productivity improvement, which is a fairly healthy value gain for our customers. And obviously, we would share in that value. And then that will start in the second half of the year and then through 2024, we'll ship obviously, more of these 3800E. And then on the 2025 time frame, as we communicated at our Investor Day, roughly on this 2-year cadence, we have a plan to go to a 4000F that will provide higher value in the form of imaging, overlay and productivity. And so that will continue to drive this before we then move to high NA also in that time frame in 2025.

Brilliant. And last question for me before we go to questions from the audience. I'd like to talk about gross margins. So obviously, you've been able to make steady improvements in your EUV gross margin. Can you talk a bit about the future evolution of profitability for EUV at a gross margin level. And then perhaps also at a group level, it would be great to get you all later to see that?

Yes. So for EUV, we were talking with these different platforms. We kind of have moved from pretty low margins a few years ago to -- we talked about a 50% margin with the D. And then, of course, we also put out there in our Investor Day that as we move to the E model in the '23, '24 time frame, you should see EUV with DPV-type margins in that time frame. So that's going to be -- obviously, has been a carrier for our margin improvement going forward. I think the same can be said also on the service side of EUV. There's also -- as we continue to scale a number of EUV machines out there, we become more efficient on how we support them and we generate revenue that will also help drive that going forward. So those are, you can say, critical carriers of margin are going forward. If you talk about the margin and the corporate level, I think, first off, maybe at the start of this year, where we are now and what we have planned going forward. At the start of this year, we plan a gross margin around 53% for the full year 2022. In April, due to some of the inflationary items that we talked about, we had to lower that to 52%. And then and primarily, we talked about 3 things. One, we said the inflationary extraordinary costs that we're seeing are a combination of freight, labor and then just the system itself in terms of the components. And then if you then step forward into July, we said that we had 2 things that occurred. One, we took an additional EUR 1.8 billion of revenue due to fast shipments, and this delayed revenue moved out into 2023. That had 2 effects. One, it was the product mix that we were moving out were primarily EUV and immersion. So you could say some of our higher-margin products were moving out of 2022 into 2023. And then obviously, the number more units coming out, but also as we're spending in the ramp, we had reduced fixed cost coverage in the year. So that was the majority of the impact in July that moved us from the 52 number down to the 49 to 50 for the full year. There were still some inflationary effects in there. But again, the majority of the move in July came from the additional delayed revenue into 2023. So that's where we stand today. Obviously, we're in the discussions with our customers as it relates to the inflationary extraordinary costs that we see there in terms of how to fairly share in this 4 to 5x increase in freight, labor growing significantly. And then obviously, the impact on system component level, some this year, but also into the machines that we'll be shipping next year. So we're in the discussions in terms of how to recover that. If you then look out to our longer-term plan of 54% to 56% by [indiscernible]. We still have that as a target and a plan to get back to. And again, we'll talk more in detail on that also in our November Investor Day.

Great. Well, really appreciate that and really appreciate the answers to my question, Skip. I guess we've got 9 minutes left. So I'm very happy to take -- or for Skip to take questions from the audience. Perhaps if you could raise your hand if you'd like to ask a question, and then we can go to you.

No questions.

We cover it quite well. Maybe I'll -- do you have a question over there or I guess I'll perhaps just continue then. I guess one question we get from investors is just around the transition to these new design sort of 3D DRAM, gate-all-around on the logic side as well. How should we think about the impact of that for ASML? Does that reduce or impact in any way the importance of EUV? And then I see you standing there with a microphone so.

So gate-all-around.

Yes.

Yes. Yes. So first off, gate-all-around, we view gate-all-around as positive for the industry and positive for ASML in the sense that it's a -- by the way, gate-all-around for those who don't know, it's a device architecture change. We went, if you may recall back in the 14-nanometer, 16-nanometer node, we made a transition from planar to FinFET. And now we have planned transition. I think all of our customers that are planning this transition have announced the timing between, let's say, this year in '24, '25 time frame, but in that window to go to gate-all-around. Why is it good for the industry? Why is it good for ASML? It allows -- and what you're doing basically is you're changing the control of the gate from one side to 3 sides of FinFET all around, so you get more and more control of the gate. And that's what we're effectively doing with gate all around. And what that allows is that allows us to continue to shrink and basically enable our future shrink road map as we move from today's 4- or 5-nanometer 2-nanometer and sub 2-nanometer down to the angstroms type nodes. And then you look out in time and maybe even something like a C [indiscernible] has been talked about as the next transition. But again, the point of these like the FinFET, it allows this continued shrink. So if you look at our litho, projected litho spend or to intensity for future nodes, it continues to grow in time. Yes, it's a transition to FinFET and gate-all-around, customers may be less likely to make aggressive shrinks at those particular transitory nodes, but then resume shortly after the aggressive shrink and therefore, the continuation of more EUV layers and therefore, increased litho spend. We showed in our Investor Day, this is a logic related question, we show roughly a 30% node-on-node increase in terms of litho spend in this time frame. And again, we'll update that as we talk in November. You also said 3D DRAM. So I guess I can go ahead and touch on that one while you're on it. I think 2 things there. So now we're shifting off logic into memory and in particular, into DRAM. And the question has to do with if you look at DRAM, maybe 10 or so years ago, I think there was a view that we would never go below 20 nanometers. It was just -- we don't know how we're going to do that. that's near at you. And so the question was, what's next? And so customers have been looking at these alternatives or next-generation memory for quite some time. You think MRAM, PCRAM ReRAM, these are all next-generation memory technologies have been investigated for quite some time. But interestingly, as always in this industry, we continue to innovate and find new ways to do things and we keep pushing things further and further. So now you've heard some of our memory customers referenced the fact that they plan over the next 10 years to take DRAM using today's traditional 2D DRAM all the way down to 5 nanometers. And with that, using extensive more use of EUV, both low and A, but also now there's talking and orders in the -- from our memory customers for use of high NA. So when they'll actually need to transition to the next-generation memory, obviously, is pushed closer to 2030. What that next-generation memory will be, I think is still defined today. They're looking at all alternatives. And by continuing to invest in EUV, there's a pretty good chance they'll continue to need EUV to make these devices smaller. So I think both of these paths are all about future innovation and continuation of Moore's Law. And with that, obviously, litho to continue to enable that.

Sorry, we just have time for one last question.

Sorry, can we just go back to the order book. Is there any risk when you look at your order book that there is any double ordering in that as we've seen one of your competitors rephase or readjust their order book. But I would assume in yours that given who the customers are, there's far less risk of any double ordering?

Yes. So may take it separate in 2 ways. So the question was around risk and double ordering. And I think there's -- you could argue double demand. Let's start off there and maybe explain it suddenly between double demand and double ordering. First off, if you look at the advanced technology that from advancing nodes -- that probably comes from, let's say, less than 10 customers. And we have very routine, very deep discussions with these customers. And so we have a very good understanding of what's driving that. And therefore, you can say with a much higher degree of confidence that there's no double demand there. If you take the -- a much longer tail of customers that are producing the more mature nodes, it's not as straightforward to understand that every technology node that they are a wafer that they're building that they may be competing for some of the same application space or some of the end customers. Therefore, that brings in the question, is there a higher risk of double demand on a mature nodes. And I think that, from that perspective, as I just described, there could be. However, keep in mind the difference between a double demand and double order is the fact that we didn't take orders for all of the demand that was out there, meaning we were limited by our supply chain. So therefore, I think if you look from a double demand perspective or double order perspective, there's a difference there in the sense that we were only able to supply today, roughly 60% of what was the demand was. So you already had some reduction in terms of what we took in the order in terms of orders. But it's also probably fair to say that there is a more risk in the double demand side on the mature versus the advanced just because of the fact the number of customers that are out there.

Great. We'll skip on that note. Thank you so much. Really appreciate the discussion, and thank you all for joining.

Thank you all. Thanks, Alex.