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

Hi. Good morning. Good afternoon. Thanks for joining this second session on Day 3 of our U.S. TMT Conference. It's my pleasure to host Peter Wennink, CEO of ASML Holdings. With Peter, we also have Skip Miller, who heads the Investor Relations department at ASML. Thanks a lot for your time this morning, this afternoon.

Peter, maybe if I start with a very high-level question, I guess, if we were to go back about 6 months, there were a lot of concerns, questions about a pace of adoption of EUV at a leading edge. One of your large customers and customer, Huawei, had some serious restrictions from the U.S. government. Now it seems like we fast forward 6 months, and most of that void has already been picked up by other customers. So how should we think about what does that mean for demand, and also the pace of EUV adoption at your foundry customers? How should we see that in going into 2021?

Yes. I think it's -- yes, we're taking up the void. It's been relatively recent and 6 months ago, was also not expected to that extent. So that's positive. Because when we heard about the blacklisting of Huawei and also one of our other major customers shifting out their 7-nanometer node, that was, of course, reason to adjust our 2021 planning for EUV systems. Now I think the fact that this 5-nanometer void has been picked up so quickly, gives us good confidence that it will also happen 4- and 3-. And if you want to do N3, you'd better make sure you have some capacity already starting in 2021 to be shipped, which will be operational in 2022. So yes, I think that is a positive sign. It's a very positive sign.

Just taking that a step further, maybe, Peter, if I look at obviously the industry nodes are getting a bit more complex with every generation. You obviously have been trying to work very, very closely with all your customers. I think one of the things you, at the ASML, are trying to do is bring out a new EUV tool every couple of years, which offers improvement across the board. So how should we think about -- when we look at the D tool, we already are looking at some improvements in the D tool next year, you've talked about it. What are the real big challenges that you're hearing from your customers or areas of improvement that you can still offer improvements? And that's something that customers are actually looking forward to increase the pace of production.

Yes. I think it all comes down to the reason why we are chasing Moore's law like crazy. I mean it's all about cost of function. So they're basically looking at cost, they're looking at being able to create a new function, so the next node. So the issues are do you guys provide us with a patterning tool that will allow us to keep shrinking according to our road map from to 7 to 5 to 3 to 2 to 1, yes? I think we'll be able -- at our current patterning rope and we'll be able to do that, which includes High-NA. And then where they say, "Okay, check," we have the patterning capability. So we have the imaging and the overlay, which is very important. The overlay, the positioning accuracy of the subsequent layers. That's extremely -- and becomes tighter. With resolution coming -- going down, that also becomes tighter. So every new generation has better overlay performance. So imaging overlay, check. That is all about productivity. It's all about cranking out as many wafers per day as you can because this little tool is the gating item of productivity in your fab. Every other piece of equipment you build, some buffer, yes, because you don't want the cheapest tool in the fab to shutdown the most expensive tool. So you basically build by design, as a buffer in your wafer fab equipment. So if you get 1% more productivity out of the most expensive tool, it leverages the rest of the fab. So these guys are all focused on, "Get me more wafers per day." And this is where the subsequent tool generations come in. After the C, we get the D with higher productivity. After the D, you can rest assured, it will be an E with higher productivity, but also with better overlay, yes? So it's the imaging and overlay plus productivity that drives every time our innovation road map. And it's why every 2 years or so, we come out with a better tool. And with that, it creates more value and with the value is a higher price.

So -- and just on that productivity topic, I think one of the questions we keep getting asked from investors is we've obviously come a long way in EUV. I remember like 4, 5 years back, we were talking about 100 wafers in AR. Today, I think your C tool is already doing 170 wafers an hour. You still have your DUV machines doing almost close to 300. So like how should we think about the pace of improvement from here on? And is that gap between EUV and DUV is that gap something that can be closed down or they are like physical limits, which don't allow EUV to go beyond a certain level?

Well, I think it's -- but all these things, it's evolutionary. I mean we definitely have a road map to 200 wafers per hour plus. We don't have a road map to 300 wafers per hour, which we didn't have either for Deep UV, when we were at the same stage of maturity. We never had -- when we started Deep UV at Twinscan, we never had 300 waves per hour on the road map. But the wafer per hour is really a function of the power of the source and the ability of the optical system to deal with that power so we can move the wafer faster. I mean moving the wafer faster is something we know how to do. It is really getting the right level of power on to the wafer so you can move the wafer faster. So the exposure time is shorter. This is what we're focusing on. Now the power road map, where we started with 250 watt, I think it's now over 300 watt to 400 watt. You even have designed -- let's say, design architectures now for 600 watts now. And then we say, "Okay, what does it mean for the optics? What does it mean for the mirrors? What does it mean for the reflectivity potential degradation?" That's all R&D, that's all research that is doing now. So yes, inevitably with that kind of innovation, the productivity will go up. But nothing on our road map that shows a 300 wafers per hour yet. But I'm pretty confident that in 5 years time, it's going to be a high number there, yes?

Yes. I think we are looking for the pace at which we are getting new smartphones on leading-edge, I think, more slaughter than stock here obviously. I guess, Peter, just on that topic of the future improvements that you're talking about, I guess, High-NA, you've already delivered on a lot of promises that you've made on EUV. Now I think all the focus is now shifting towards High-Na we know you have some time lines for R&D tools, and then high-volume manufacturing tools. Can you just help us understand where we are in that process on High-NA? And what are the puts and takes around timing for tool shipments to customers? And then also, you mentioned productivity is a key consideration for customers. How should we think about productivity on High-NA? Would it be similar to EUV? Or would it actually need some time to improve ramp up towards EUV levels?

I think the productivity will be at EUV levels when we introduce the tool. First shipment is planned end of 2022. So installation 23 at our customer site. I think the progress that we've made with High-NA has been good. The biggest change are in the optics. It's in the lens and because the lens is bigger. The High-NA is higher, the lens is bigger. The tool is also bigger. So it's a complete redesign of the tool. It's a very large machine. What your -- what the biggest challenge was our ability to create optics that we could actually measure. You need to realize that these EUV mirrors, which are probably -- one professor once told me, he thinks it's the flattest object in the universe, how you would know that is a riddle to me, but okay. So -- and then -- but for High-NA -- in diameter terms, it's a factor of 3 to 4 higher, bigger. So you need to get that flatness over a much bigger area, and you can measure that flatness not in nanometers but in picometers. And so we had to make and design, together ZEISS, the metrology as a measurement tool to actually measure their flatness, which was really the big question mark, can we do that? That you can -- you also have to test whether it's as flat as you think it is. We've been able to prove that by the end of last year, which was a major technical breakthrough. So I think from a technical risk point of view, there are always risks. But I think we're well underway, yes? It will happen. And I think what was really hampering the introduction of EUV was the source, the EV source, the if EV light. That is not going to be the issue with High-NA because we're going to use the same source, the same architecture, the same source. By that time, we'll be at the level of maturity, which is much higher than it is today, which, by the way, the relative immature level of EUV today has to do with the immaturity of the EUV source, which you're going to fix. So by the time that we're going to introduce High-NA that EUV source, which has been this literally, the source of a lot of issues, will be much more mature. So all in all, the entire effort of High-NA in terms of R&D and investments is about half of what we have for low-NA as low-NA was really starting from scratch. This is more revolutionary -- I would say in optics, is pretty revolutionary, also because these optics have never been made. But this is why this technological breakthrough I talked about is so important.

So on that optics point obviously the supply chain is getting a bit more complex. Are you fairly comfortable that we all have the building blocks in place for the volume ramp-up that the industry may need at that point? Or do you think that there is still pockets where you can actually do a bit of M&A to bolster your presence or expand your presence in the supply chain and control it maybe a bit more than you do today?

Yes. I think we always have some bolt-on acquisitions, the technology types of the Berliner Glas that we did, which are making the EUV wafer claims. And the wafer tables and -- which is pretty high-end stuff, where really, they need to step up, and have the right capacity and do investments in innovation and technology that for a smaller-type company is probably too much. It's a bolt-on acquisitions, which we feel are critical to our innovation knowledge. But apart from that, no, I think whenever we would be able to integrate ZEISS, which -- that's a well-known issue, we would welcome the opportunity. That's not the case. As we all know, ZEISS is an independent company. It's a German foundation. And so it's not a public company. They don't have shares, the foundation is a foundation. So it is what it is. But that's probably the only vertical integration I would ever consider at this moment in time. But we've been talking about this [indiscernible] it's likely not going to happen anywhere soon.

One of the questions, Peter, I've just received from one of our investors is about the guidance for next year. On the last earnings call obviously you talked about 20% growth in your EUV systems revenues. Obviously we can make our own assumptions around ASPs and the number comes out to be anywhere between low 30s to high 30s tools for next year. Obviously there are moving parts in your guidance. And clearly, we don't have visibility on all those things. So can you just help us understand like what were your assumptions around some of these high-level points around foundry, logic and DRAM, that you [indiscernible].

Yes. I think on DRAM, it can be short. I mean DRAM was never going to be a big runner in 2021. It's more like 2022, 2023 because EV is only used for one layer on a relatively small part of the current leading edge DRAM manufacturing capability. So that hasn't changed. I think what has changed from our initial thinking, let's say, a year ago, is 2 things. One, I think it's the delay of one of our major customers on the execution of their logic road map, which is a year later. So that means that you see tools being actually needed later, that had an impact on 2021. And there was, of course, the blacklisting of Huawei, which were our largest foundry customer, basically had to readjust their thinking on how much they would need in 2021 because that customer would fall away in the leading edge node, N3, which is 2022. So those were the big drivers for our adjustment, you could say, in our original planning for 2021. Now what has changed? I think not so much the delay of the road map of the customer. We just have to see how that pans out. I mean they will come out in Q1 with what they want to do. So we'll just wait that. We'll just wait for that. I think what has came to a positive -- in a positive sense is that the void that was created by Huawei for our foundry customer, which also was true for 5-nanometer, which is today and next year, that void has been taken up by other customers much quicker than we anticipated. High-power compute, mobile space. I think it's the hyperscalers, data centers. That development could easily translate into N3 also. So the same customers that have picked up the void or the gap that was left by Huawei are the same customers that would also be looking at the high-power compute advantages of N3. I think that in itself is definitely something that very likely has a positive impact on our 2021 outlook. Now on top of that, there's also the question where the customer that started to push out, will start to in-source some of that production. Will that happen in -- does that mean that our foundry customers would need some extra capacity by the end of 2021 or early 2022? That's a bit of a question mark, but these are all, I would say, developments, that have a positive push on our guidance that you translated into, let's say, mid- or low or high 30s unit numbers. And we would be able to deal with some of that because like I also said at the conference call during our Q3 results, we are buffering in our supply chain for long lead time items. Now not to the full extent that we will use our own production capacity, but definitely over 40 units. So we have a gap, that's -- between, you could say, the translation of our guidance and what potentially could be a higher number. We have modules in the supply chain that we have ordered, and those are long lead time items. I mean, if we would not have done that. And there would be upside for the 2021 number, then yes, there's nothing we can do. It would go to 2022.

No. That's very clear. Peter, I guess coming back, tying that to the gross margin comment, we -- you've always mentioned that your ambition is to get EUV towards DUV-type gross margins over time. And one of the things in that whole process has been obviously ASPs go up with newer generations, but also equally on the supply chain side, you've been trying to reduce the cycle time. And one of the things you've been targeting was 45-tool production or capacity next year. Now obviously the number is going to be -- it's well above what you need potentially for next year and maybe demand starts to recover in 2022 onwards. But does it have any impact that this shift to gross margins? Or is it not that even in how we think about gross margins for EUV as, particularly for next year.

Yes. I think higher numbers would definitely help the gross margin. I mean, you can just get a better coverage of your fixed cost so that's clear. But I don't think it's -- the biggest impact is the mix of the higher productivity tools with a higher sales price. So I think this trajectory that we're on of improving our gross margins to over 50% or close to the DUV margin over the next 2 to 3 years, that trajectory still impact also next year because the biggest impact is going to be the mix. I'd love to sell more, and we get better coverage off of market cost. And if it doesn't happen in '21, happens in '22. But the improvement trajectory is still intact.

Yes. And then on the gross margin, I think the second element of gross margin has always been the scope for improvement in EUV-related services. Obviously I think you've mentioned that now it's going to be linked more towards productivity of these tools. And as productivity increases, you definitely get the economies of scale -- the customers get the economies of scale and you [indiscernible] more. What's the ambition for that EUV services gross margins in the long term? It's still well below what you make on DUV? Is it trying to get towards DUV-type levels? Or is there potential to actually expand that number beyond what you make on DUV services today?

Yes. I think the longer-term target -- because there is still a gap between the service margins and the system margins, but you're absolutely right. I mean the fact that the EUV service margins are more tied to the output to the wafers per day and to productivity gives you an upside on being able to generate more service revenue. Of course, you have some higher cost because it's wear and tear. But I think all in all, it should be a driver to improve our gross margins, which are currently, I would say, [indiscernible], yes? I think we'll probably need until the 2025 time frame to really bring them up to the margin levels that we currently see in the rest of our business, which, by the way, is bridging a quite a significant gap. And I think we can do it. And I think we can do it because of the fact that we will work on higher availability. Well, higher availability means higher output for our customers. That means that we will be able to charge more for service. But the flip side is, if you have higher availability, the deal is not down. It's not down, you don't have the replacement cost. So it's kind of a double-edged sort. That works in your favor, if you want to improve the margin. Also, what I believe is we have to share some of that with the customer. Our customers are crazy. I mean they see risk is going up 2 time up so I say, "Hey, where is the benefit?" So -- but I think it's like sort with Cymer. When a Cymer started to introduce the [indiscernible] model, charging service costs based on number of data [indiscernible] that comes out of the laser. Yet you saw the same effect, sometimes weather because there's a mutual benefit to the majority of the tool, and the cost went down because then the cost -- now what we do is we do break fix. We have cost, we put a margin on top, how we charge it to the customer. So what's the benefit of driving cost down? Because your sales and margin will go down. So it's a kind of role incentive so -- but when you know what your sales is going to rise and the cost is going to be in your pocket, then the cost drive is different. We saw it in the Cymer also. Part of it, we will get back to the customer. But the net benefit will be a growing margin profile. But it's going to cost. It's going to take a couple of years to actually get to the DPV levels from where we are today.

No. That makes perfect sense. If you switch to the memory market, in the last couple of quarters, we've seen some decent recovery in your memory-linked revenues on the [indiscernible] systems business. Can you just maybe help us understand what are the puts and takes as we go into next year? How should we think about this sustainability of memory recovery? And when we think about next year's memory revenues, is a large part coming from predominantly DUV? Or does EUV become a meaningful contributor to that growth in memory in 2021?

Yes. I think it's largely going to be Deep UVs. It's not EUV. I mean, memory is still -- the capacity extensions in memory are still DPV-driven. But you have to split between 3D NAND and between DRAM. I think DRAM -- I was in Korea last week, I had a lot of discussion with our green customers on this. I think we all agree that the memory market -- the DRAM market looks pretty positive. That we will be able to manage the inventories over the next month or month or so. We will ship in Q4, a decent number of deep UV tools the memory customers. They will be put into place early next year and we'll be ramping in Q2. So that pressure that we will see -- because we will reach the maximum utilization of our current installed base by the end of the year, which will coincide with health -- well, healthy inventory levels. You could then see dealer prices going up. However, we will ship some capacity in Q4, which will help to alleviate that pressure. But to support 20% bit growth, which we're now thinking of next year, we need more capacity. The customers need more capacity. So I think DRAM will see a recovery next year. That's what I believe, yes? And will be Deep UV-driven. Now that same pattern of increasing utilization. We've also seen for 3D NAND. But over the last 2 months, it started to level off a bit. So it's not at that you could say, pressure point for DRAM. And that might take a bit longer. I think in all of the discussions we've had with our customers over the last 2 months or so. I think that probably would agree with me. That would probably take a bit longer for -- to come to that utilization point for little, in any case, that would require additional capacity. So it will be a DRAM year.

And maybe, you've just mentioned that you've been to Korea recently. Any thoughts that you can share on how the customer is thinking about a 3-nanometer road map? Obviously there is a divergence. Half of the industry is going towards fintech, and maybe some are thinking about gate all around. And what does that mean? Does the choice of technology, does it mean different things for EUV adoption as we think about 3-nanometer in the future?

No. I don't think so. I think it was pretty -- the, let's say, architectural choices pretty agnostic to us. I mean, it doesn't matter. I mean we -- it's both needs at needs [ indiscernible] little to then -- about the same extent. Yes. I think it's an interesting question because it actually means that logic -- advanced logic development takes place in a couple of places on the plant, as we know. And it's an interesting question to see how the green customer is now looking at the logic market and see how -- what they can do in terms of their foundry aspirations. They have the technology. I think in foundry, is a bit different market. Also when you think about the relationship with the customers, then in the DRAM market. DRAM market is more a commodity market. So basically, you can get your ad plans DRAM through brokers, through different channels. But if you want to outsource your leading edge litho -- etiologic designs to a foundry, that collaboration model is different. Starts earlier, extends over a longer period of time. And once you can check back the technology, it will work. You're going to ask the question as a fabless company, "Where is the capacity?" So it's not only it, it's also capacity. So if you want to play in that area, and you want to be also a leader in the logic area, it's not only having the right technology, but also you have to have the capacity that you serve to the customers. But the leading-edge tools you're having a long lead time. Of course, you need to start thinking about your strategic investment plans. This is not like a Deep UV tool or a KRF tool, which you can get in 4, 5 months. I mean it's not the case with EUV. So this is a completely different way of looking at the world, logic versus DRAM.

Thanks, Peter. I guess with that, we are also running out of time. We could go on for a few more minutes if I wanted, but I'm conscious of that time. So, thanks, again, Peter, for your time this afternoon, and thanks, everyone, for joining us on this ASML session. Take care and thank you.

No, that's my pleasure. Thank you.