Thank you. Welcome everyone to ASML's 2022 Investor Day. I am Skip Miller, Head of Investor Relations. I wanna thank you all who have made the journey over here to Veldhoven. It's so nice to see many of you in person. It's been a long while. For those joining in the webcast, good morning, good afternoon, good evening. Before we go through the agenda, I would like to first maybe walk through some of the housekeeping items. First, for safety, the exit's here and in the back. Regarding restrooms, they're over on this side, on the next floor up. If you would please put your phone on silent or airplane mode now, just a reminder, to avoid any interruptions. Presentations that will be given today will be posted on the website at the end of the day.
Regarding Q&A, we're gonna have two things here. First, for those here in the audience, when we get to Q&A, raise your hand. We'll have people in the aisle who have microphones. You'll come down and get a chance to ask questions. We'll try to mix some in for those joining on the webcast. You should see on your screen an area where you can ask a question during the presentation. Please submit your questions. We will take those, and we will weave those in as part of the Q&A session. If you could, you don't need to, but if you could leave your contact information in case we don't get to your question, we can address it, follow up from the IR team. With that, I would like to move to the agenda.
Today we will first have our CEO, Peter Wennink, who will talk about the end markets, wafer demand, and what we need to do in terms of capacity, followed by our CFO, Roger Dassen, who will talk about the financial model, business model. Peter will wrap up the closing around 3:00 P.M. We'll take a 30-minute break, and then we'll move to the Q&A session. We'll have our members, six members of our ASML management team will join and will be here to address Q&A. We'll have an hour time there. We will wrap up the formal program at 4:30 P.M. As far as those that are in the audience, we will then have a review of the agenda after Q&A that will take us through a dinner at 8:00 P.M. Okay.
Now, before I start, of course, we have to read all this. If you'd read along with me. No, I won't make you do that. Before we begin, I would like to remind everyone that comments made by management during the event will include forward-looking statements within the meaning of the federal securities laws. These forward-looking statements involve material risks and uncertainties. For a discussion of risk factors, I encourage you to review the safe harbor statement contained in our press release and presentations, which will be posted on our website at asml.com. Now, as it's been over four years for many of you, and some of you may not have even been here, but those that were here back in 2018, as you came in, you probably looked and said, "Wow, this place has really grown." You see a lot more people.
Before we start with Peter's presentation, we'd like to do a quick intro video just to show kind of where we've come from, how we've evolved, and how this place has grown as we prepare for the expansion and future growth of this company. Thank you.
ASML is driving the technology of tomorrow. Our customers depend on our products to bring their cutting-edge technology to life. To meet their needs, we invest in the future. We invest in research and development, so our machines can deliver the smallest features and the highest yields. We invest in our factories and facilities around the world, so we can meet our customers' increasing demand for our products. The number of machines we plan to deliver in the coming years continues to grow. We also invest in our workforce, the people who give life to our values: challenge, collaborate, and care. They come from more than 100 countries to work together and advance ASML's mission. Our values push us to invest in being a good neighbor and global citizen. From supporting art preservation to minimizing our environmental impact, our initiatives lay the groundwork for sustainable long-term growth.
To make our vision for the future a reality, we need to collaborate across departments, across sectors, and across continents. Our investors enable the innovation that advances our technology and creates our value. Together, we will lead the semiconductor industry into a sustainable and highly profitable future. Welcome to Investor Day 2022.
Thank you. You know, first of all, I'd like to reiterate what you said, Skip. It's great to see you all. I mean, I've just met a couple of old friends, you know, people that we've been seeing for the last decades, you know, been following ASML pretty closely and have also seen the growth of the company over the last decades. Why is this introduction video important? Because actually it, you know, shows growth.
Growth is particularly important because it will drive what we will do the next decade and beyond. I'm going to talk a bit about what we see and probably mentioning the obvious from time to time, 'cause when I talk about this, these developments that we all know of. We try to somehow translate that into the business that we're in and what ultimately is the demand for wafers, 'cause, you know, we're a semiconductor equipment company, and we process wafers, and ultimately, that's the driver for the size of the company and the volume in growth. I'm going to talk about megatrends, and this is where you say, "Yeah, yeah, Peter, please, Pete, move on.
I'd like to see the numbers." I'm going to give you some numbers in that sense, yeah. Then we'll translate that into wafer, you know, demand and why we need that capacity. Of course, Roger is going to translate that into things that you're really interested in. That is the numbers. Now, clearly, these global trends, it. I cannot repeat that enough. I mean, trends that we are currently seeing, we didn't see a couple of years ago. Now people ask me the question, though, why didn't we see this massive and big demand on our products coming? Because we simply do not connect all the dots.
It's still a challenge today to keep connecting all the dots, but it's the value of Moore's Law, which basically reducing the cost per function, yeah, that will drive our business and will create these building blocks for growth and for solving some of humanity's biggest challenges. We are a strong believer in this, yeah. Moore's Law is alive. It's still alive and kicking, and it's about the cost per function, and we all know if the chip gives you more functionality, more value, then it creates cost. We're going to create applications and solutions, and that's happening. I mean, things are actually happening across the globe, and we'll talk about those trends, but generally, if we go through it, we believe that it will translate into, let's say, a CAGR from 2020- 2030 of around 9%.
That drives the increased demand for wafers, which if you translate that, it's about 6.5%. You can say, "Why isn't it 9%?" Well, because of the value. The value, we think, will rise faster than the cost, which is reflected in the wafers, yeah? For that, wafers will become more valuable. Yeah, in order to do that, just in summary, the 9,620 is the number that we gave to you in April of this year, and we're going to reconfirm that throughout the meeting. First, the megatrends. I talked about this, the connected world. We see a connection stepping up time and time again. It's in smarter cities.
It's in the connecting of billions and billions of things, whether it's 40 billion or 300 billion, it's just rising and growing faster than we anticipate. It creates unprecedented data volumes. Something also happened, and I think it's the focus on climate change and resource scarcity, particularly through the last couple of years. I think one of the areas where we underestimated the growth of semiconductors is in this area. Because, you know, energy prices will go up. The demand for energy will go up. Renewables, they will become more important. We have to focus on waste and, you know, pollution. As I will show later on, there's some very good examples that semiconductors are going to make a big difference. Then, you know, also a very important trend, we'll have more people.
Focus on life sciences and medical and health is going to be paramount. The organization is also accelerating. Now the last two points. You know, I'm from a generation where actually we were given the promise of the global village. I had a discussion where there's some young people that actually said, "You guys screwed up, you know? Your generation, because you left us with all these problems." I said, "Listen, but our generation did create a global village. You know, our generation, my generation, did create the internet, did create global travel." I think this is, and it gave us an enormous push in economic development and in innovation. That is de-globalizing, and I now also see that.
I think we're seeing socioeconomic blocks that will lead to, let's say, geopolitical discussions and the drive for technological sovereignty, which is going to be a very important driver for our business going forward, and we'll talk about that. This is just translating into applications, connected world cloud, artificial intelligence, AI, edge computing, energy transition and electrification. Now, energy transition was not on our radar screen. It's one of those applications, one of those many dots that are created in the semiconductor space that we didn't have a good vision on. Yeah. But it's happening as we speak. It's going to be a significant driver. Of course then there's social and economic shifts, which is the different way of working, healthcare, and the sovereignty drive.
Now, on top of that, we see this drive for AI, for artificial intelligence, and it's not only in our systems, in our networks, it's not only in, let's say, the way that we compute, but it's also very much on device. So it's in the edge, it's in the cloud, it's in our networks, but it's also in the individual device, where whether it's automotive or whether it's the smartphone, whether it's the PC, whether it's drone technology. These are applications that you will find in health and in life sciences, and you will have this application all need this type of innovation. The artificial intelligence drive is a big driver also for our industries. I wanna go into some detail on what I said earlier.
I think the energy transition is something that surprised us, yeah. Just anecdotal evidence. We had a Chinese customer that actually was in our offices at the beginning of the year, and they said basically they were well making very mature you know stuff. Yeah. Was in power in some analog products. Actually said, "Peter, you know, we need your help because we're going to build a 300 mm fab, 90 nms. 90." Yeah. I said, "Why do you need
Wanna build a 300-mm fab for a 90-nm? He said, "Well, it's because of automotive power ICs, smart power ICs, and the energy transition." He started to talk about the energy transition that they needed and the semiconductors they need to make for wind and for solar. He didn't even talk about the energy grid. Basically this is some data that we got from some of our customers and some of other companies, that this one is from Shell. It's an estimate that 1 MW of wind power will take EUR 3,000 worth of semiconductors, and solar EUR 4,000. Now, the question: what was the electricity needed in the world in 2019? It was about 24,000 TWh.
Now, you can then calculate this and take the average of wind and solar. If all the electricity you need in the world, which of course is not going to be the case, but it's going to be wind and solar, you come to significant numbers. Now, the 24,000 was 21,000 three years earlier. It was 2019, the 24,000. Now it's 2022, we're three years, you know, later. That will accelerate because of the consumption of electricity. Now, if you then think about this, if all electricity would be wind and solar, then massive amounts of semiconductors are actually needed.
Now, of course there will be a mix, but it just shows you that there are things that we haven't thought of, and it's particularly in the area of what we call mature or mainstream or specialty semiconductors. It's in power IC, it's in analog, it's microcontrollers, it's sensors, optical sensors, non-optical sensors. We see it on the consumption side, and this is data that Infineon shared also publicly, said, you know, the expectation is by 2030, 70% of the cars will be somehow related to an electric vehicle. Could be hybrid, could be high hybrid, low hybrid, could be full, you know, electric. 70% of 100 million cars is 70 million cars, you know, and it's more than $1,500 per vehicle in this decade.
Could be the end of the decade, but you do just do the math. This is without. This is on top of that you have the ADAS, that is the Advanced Driver Assistance Systems that take massive amounts of, you know, semiconductors. We will come to that later. Yeah. We actually see that back in. This is a slide that I borrowed from a good friend, Lars Reger, CTO of NXP Semiconductors. I saw this slide when he was presenting on the imec Future Technology Conference. You know, the word in Europe was when it came to semiconductors and automotives, that's all 28 nm stuff and older. Well, this is what Lars presented and it's. This is what.
This is across the entire product portfolio. It's basically about sensors and actuators, yeah, that actually pick up the analog data from the sensors in the car that basically compute this real-time close to the sensor, yeah, and then bring it into the communication system of the car so that it can be processed by the main microprocessor, which is 5 nm. Basically, it's what Oliver Zipse told me. He said, "Peter, we are a system integrator." If you're a system integrator, you use the entire product portfolio. This is also what we are seeing if we later on look at some data from TSMC. Automotive will be one of the most advanced users of semiconductors in this decade.
Now, geopolitics, and I've said it before. I mean, some people that know me know that what I'm going to say is 1973 revisited, the oil crisis. Oil was always there until it wasn't, and it was a strategic commodity. It was a strategic problem. Suddenly we fast-forward 2021 or, you know, 2020, and semiconductors have always been there until they weren't. We suddenly start to, you know, realize that there is a dependence, a dependency created in those decades that countries and geopolitical systems don't like. We see this push for technological sovereignty.
I think we've just seen the beginning, 'cause if you think about the growth rate of the industry, and you think about the geographical distribution of chip manufacturing, then if you want to create some level of technological sovereignty, and that doesn't mean that you need to do everything, you just need to stay relevant. Well, with that growth rate, staying relevant, it means that you need to invest, double up your investments, as countries, to make sure that you can stay relevant. Yeah. This is just the first phase. It will. There will be more coming. Having said that, do we need it? Well, and that is a question we sometimes get from politicians when I talk to politicians and say, "Why would we do this?" Why would we in Europe or in the US, you know, have this CHIPS Act?
Because the profitability of the ecosystem, and this is a slide I always like to use, it's the 50 top technology companies in our ecosystem, which starts on the top with the equipment companies, and then we deliver our products to the chip makers, and the chip makers to the smartphone and the hardware makers, which actually ends up in. Basically, it's the you know cloud companies, and it's the Microsofts and the Alphabets of this world. The EBIT in 2021, the combined EBIT of these 50 companies is $688 billion. Compound annual growth rate over the last six years of 19%. The industry has more than sufficient means to pay for this expenditure, so why then these government incentive systems?
Well, you know, at least we tell the politicians that ask that question that it is not about the fact that the industry cannot afford it, but the industry has a choice to build their next manufacturing expansion wherever they want and wherever from their point of view, it's the most efficient and the less risk, and that's where they currently have their ecosystem. If you want to pick up and build an ecosystem, you're basically creating risks for these companies. That means these companies need to be incentivized, you could call it on the non-economic top of that risk, and this is why you need to step up.
If you want to have within your geographical boundaries, you want to have advanced or mature semiconductor manufacturing, you'd better make sure that you provide them an incentive to take away part of that top economical risk. Okay. What does it mean? Do we have enough money? Yes, we do. With that money and with that innovation power, which is not only money, it's also the intellectual power, the roadmap is still very much there. You know, we'll go into some detail, but both in advanced logic, in DRAM and in NAND, we have a very clear roadmap that extends to 2030, and in many cases, beyond. This is what we're discussing with our customers these days. Just to give you an example, this is on, you know, logic.
Actually, this is an imec slide, which, imec, as you know, does a lot of pre-competitive work with the leaders in the industry. This is the roadmap. It actually goes to 2036, and when you look at the metal pitch, it actually scales down to 16-12. You know, the half pitch, you can do the calculation, yeah, is low, is single digit, yeah. That is the pitch. That's not what we would call the lithography or let's say the semiconductor node, and the node is expressed in a different nm term. Even if you go, for instance, from FinFET to CFET, you know, go from gate-all-around, nanosheets to forksheet to CFETs, you basically talk about a sub 1 nm node, as the nomenclature will be called, sub 1 nm.
That is real. This is what the research is currently working on. It's also true for DRAM. Now, this is slides for Kim about a year ago, less than a, you know, year ago. What you're seeing here is that the bit density will go up, and it also means that the design rules will keep shrinking, yeah? Which actually means that the gigabyte density per chip will go up significantly, almost exponentially. This is a roadmap which has been discussed with our customers in detail. This is not just a presentation on an open forum. This is how we and our customers work together to make sure that we understand the requirements. It's not only the technical requirements, but also the economic requirements of what they are doing. Yeah?
This is where this collaboration model that we have with our customers is so critically important, and it's becoming even more important than ever before. It's also true for instance, other parts of the industry, for instance, 3D NAND, it's the same thing. Now, if you take that all together and then you go and say, "So what does that mean for our business? What does that mean for the size of the business of our customers that ultimately are going to come to us to ask for, you know, machines?" Well, if you then look at the industry analysts and the TechInsights or whether it's McKinsey or SEMI, their estimate of the semiconductor industry anywhere between EUR 1 trillion-EUR 1.3 trillion. Yeah? But how does that then
If then we take the data and we split it over seven segments, which is basically, you know, you probably know this if you've followed our capital markets day before. We look at the smartphone market, we look at the personal computer market, consumer electronics, wired and wireless infrastructure, and then the data centers and servers, and then automotive, and then industrial. I'll talk about the industrial a little bit later. You can actually see that on the top row, consumer, personal computing, smartphone, that's single-digit growth. Yeah? It's data that we use from research institute till 2026, 2027, and we do some extra, and we internally do some extrapolation also based on the discussion we have with our customers, based on the roadmaps that we just saw.
You see that there are three areas where we see double-digit growth, and that is in the server and the data center domain. Remember the AI slide that I showed you? It's very much about the cloud and the edge infrastructure that is going to drive this. Automotive. Well, talked about the automotive market and the fact that the electrification is going to see a significant jump up in terms of use of semiconductors. If you look at the drivetrain, you know, an electric vehicle needs twice the number of semiconductors. And you don't talk about the Advanced Driver Assistance Systems. That's all about safety, and that will be on top. If you look at that, it's significant double-digit growth in the automotive market.
It's also what we are seeing today, where the automotive market, as you know, there are still shortages in the automotive market because it grows so fast. The last bucket is industrial electronics. Everything that we cannot put into these six buckets goes into bucket seven, and this is where we have these examples of the energy transition, of life sciences. At the imec conference, for instance, we saw a presentation of a life science company that does DNA sequencing. Actually showed a chart that sequencing one DNA 25 years ago was $90 million. Now it's $600 with that new machine, and by the end of the decade it will be less than $50. Yeah? But that machine generates 20 GB of data per day. Yeah.
I think if you then think about that's in that bucket. We just talked about the electrification, and we talked about the energy transition and what is needed there. We talk about industrial, you know, applications. I think artificial intelligence, you will find very clearly the biggest impact in industrial applications. Now, for what it's worth, it's double-digit growth. I think it's conservative for the simple reason that we don't know. We don't know. We simply cannot connect all the dots. Those dots are being created because the value of Moore's Law is there. We're creating more value than we create cost with every new node. Some intelligent company or a group of intelligent people is going to think of an app.
of an application that we currently don't know of or that we cannot assess in terms of size and impact. It's basically the question you've been asking me for 25 years. "Peter, what's the next killer application?" I say, "Well, same answer. I have no clue." It is also what the CEO of one of our major customers told me when I said, "Why didn't you tell us you needed so much more mature tools?" He said, "Peter, because we have no clue." That's happening. Yeah? That's happening throughout the industry. Why? Because we are continuously creating value. In that bucket, that's where we will probably see the biggest surprises. This is what we take.
This is what we currently know, and we take those growth rates, and we then say, "Okay, what we at ASML, taking all that data, taking those seven buckets, what do we think?" We think if we add it all up, it's only about a CAGR of 9%, and it's, well, pretty close to EUR 1.1 trillion by 2030. This is how we look at the world, at the end markets, in combination with the discussion we have with customers on their roadmaps. Now, what does it do to wafer demand? Now, I'm going to take you back to last year, because one of your questions must be, you know, last year you did the same tricks. You know, you went through it, and you missed a couple of things, and we did.
You know, one thing I told you, the impact of the energy transition, we didn't take into consideration the way we do it today. Yeah. Last year, we actually said, when you look 2020 as the base, that by 2025, on advanced logic, we would probably add 130,000 wafer starts per month per year. That's this 0.13 on the top row. On DRAM, we say it's about 80,000. On NAND it's about 100,000. On mature logic, where we were most wrong, it was 200,000. We said, well, by 2025, probably the industry needs to support the growth of our customers, which was the previous slide. We need about 505,000 wafer starts. Yeah.
Now we're a year later, and we have to adjust a couple of things. They're predominantly in the logic space. They're predominantly in the logic, in advanced logic. That is basically, it's the increase that we're currently seeing because we underestimate a couple of things I'm going to go into some detail later on. For instance, the focus on energy efficient performance in high-end semiconductors. I'll talk about that later. It actually leads to bigger die sizes. Actually, we're now having insights into what these die sizes are going to look like, which is different than what we thought last year. The biggest jump up was in mature logic. That is actually where our biggest shortages are. You know, shortages are in KrF, are in i-Line, are in dry ArF.
That's 15-year-old technology. This is where. Why is that? Because everything I just talked about in terms of automotive, energy transition, is not about 2 nm chips. This is about 28, 45, 65. It's about analog to digital chips. It's about power ICs, very important. Smart power ICs, yeah. Then we have. We're talking about microcontrollers, sensors, optical and non-optical. That's in that bucket. This is where we did much more research and insight together with our customers, and this is where the biggest shortage is. Now, when you then look at what that means for 2020-2030, our compound annual growth rate in terms of wafer demand is then going up to 6.5% with the
Within absolute terms, the biggest growth in the mature market, yeah. It's driven by like I said earlier, energy efficient focus on advanced semiconductors, and it's the mature market basically connecting more dots than we had sight on. That, that's clear, and that's what we're seeing today also. I mean, also for this year, sorry, for 2023, next year, we have a significantly higher demand on our KrF and i-Line and dry ArF output than we can make. There's a significantly higher demand, you know. What are the changes then? I wanna go into a bit of detail there.
I want to start with the CMD of last year, where basically we said we'd need, for 2020 to 2025, 505,000 wafer starts per month additional capacity per year. Of that 505, I think we were wrong on the advanced market simply because of market-driven growth, which is the server market, AR/VR market, which is basically a bigger market than we anticipated, which accounts for about 45,000 wafer starts increase. The mature market, we were mostly wrong, which is basically industrial and automotive, which actually is. You can say again, industrial, what is that? That is this wide array of applications where we simply have not understood the demand of the market. 180,000 wafer starts is our calculation.
Market-driven growth altogether is about 225,000 wafer starts additional from 2020- 2030 in terms of wafer starts per month per year. There's a technology-driven growth, which is another 50,000 wafer starts. That has to do with the larger die sizes that I talked about. I'll give you an example a bit later. The advanced market is growing faster, and I think you clearly see that in the server market, which is Gartner data and the VR market. I think more importantly is probably what TSMC showed. Actually when you look at the smartphone and consumer market, and you look at these little blocks on the left-hand side, and you see, for instance, sensors are actually moving from 0.35 micron now up to 28 nm.
More bands and higher speed in smartphones and on the consumer side also moves up from 65 nm to 6 nm, which is very clearly EUV territory. The same is true for sound and image. Basically what you are seeing is in those products, smartphone and consumer, there is a significant increase in technology innovations that need next nodes. That's a driver that we now have better insights in, and that means that there is a higher demand for advanced semiconductors. The mature market. The mature market I talked about this is a market that's very difficult to understand because it's so wide. The energy transition, I talked about solar, I talked about wind.
I didn't talk about the grid, about the energy grid. Energy grid needs to be upgraded. Needs to be upgraded with microcontrollers, with sensors, basically because when you generate the electrons, you need to offload it to the grid. That's going to be a significant area where mature semiconductors are going to be used. But also on the consumer market, there's also a TSMC slide, and when you later on look into the detail, you see that in all these areas, you see that the number of mature chips used in those applications are sometimes twentyfold higher in terms of units. Yeah. All this has to do with the smart cars, huh?
When you look at the automotive and then look at the increase in low-end chips or mature chips that are used in a car, and you look at those four blocks, you can see that it's 10- 20-fold increases of just mature semiconductors, just for smart cars. This is all data points in here, but what do we see? I think this is an important slide. What we actually see is the mask sets that are being run through our, your systems, anything above 28 nm. 28, you can see it here, it's 40, 65, and 90 to 130 nm, to 0.13 micron. You can see the technology that is needed for it, which is in these green blocks.
You've seen that over the last couple of years, we've seen a 40% increase in the number of mask sets. Every mask set is a product. You see that instead of, you can see in 2012, the dark blue line, which is 90- to 0.13 micron actually went down. From 2016 onwards, you see a significant growth. This is what we're continuously seeing. This, by the way, is also being confirmed by some of the mask makers. There's a E-beam mask maker initiative that actually over a somewhat shorter period, saw an increase of 48%. Actually, we are. That was in this mature market node area. This is a clear proof that there are products being designed and being fed through the manufacturing system that we have no clue of.
We could, of course, seen this, but we didn't understand it. It's happening, it's there. Now, energy efficiency. I think, the point is, if you want to make the leading-edge device more energy efficient, you like to lower the clock speed. You know, you basically like to run them at a lower voltage. If you do that, the performance goes down. How do you then level up to the performance? You include more transistors. If you have more transistors, you need more surface area, which I think is very well depicted by this black part of the slide, which is basically a slide from Apple.
It's probably difficult to read, but if you look at the left-hand side, you see the M1 chip. M1 chip basically has high performance and low energy usage. If you look at the size of those chips and look at the right-hand side, the M1 chip is by far the biggest. Yes, you increase performance, you reduce energy usage, but there's one penalty, and the penalty is size. You need to have bigger die, and to get the same units out, you need more wafers. That's what we're seeing. If you add it all up, this is again the slide, 780,000 wafer starts per month per year.
On top of that, there's the technological sovereignty and the foundry competition as a result of it, because you will see that multiple companies or more companies are now going into the foundry market in the leading edge, and they were going to distribute across the globe. That will lead to some level of inefficiency. We assume that if you build across the industry 10% inefficiency on the installed base, then you need about 150,000 wafer starts additional just to create 10% inefficiency, which is depicted on this slide. To get to an installed wafer capacity, whereby we're able to deal with inefficiency as a result of the geopolitical situation, we need 10% more wafers. That's a 930,000 wafer starts per month per year increase. Can they afford it?
I think yes, they can. That's an earlier slide. These are the announced CapEx spends. If you just add up for our three largest customers and see where it happens across the globe, then this is $330 billion, yeah, of announced fabs. I'm not spilling the beans. I mean, this is just public information. Yeah. That's the global capacity. It's happening, and it's happening in different parts of the world, as you can see. That could be a driver, very likely is going to be a driver from some level of inefficiency. Now, if you believe that 10% inefficiency is too high, you take 5%, which basically you divide, you know, 10 by 5, and you know that you...
By two, and then you know what about the wafer capacity additions will be, if you want to take another number. It's going to add to the wafer capacity that is needed in the industry. Inefficiency is unavoidable. We need to expand capacity. I think if we wanna grow, we need to expand capacity. This is just, we all need to realize that we see this very strong long-term growth. We have to add capacity, which is going to be step functions, like we are going from 375 DUV to 600. We have to prepare for cyclicality because we don't predict cycles. Cycles happen because of all kinds of reasons.
Of course, we want to balance the interest of our stakeholders, which of course includes very close contact with our customers, with our suppliers, with our investors, and with our people, and with society. How do we do that? There are two ways to add wafer capacity, but just selling more machines. If you compare the 600 systems per year with the year 2020, you see a 2.5x increase in DUV and a 3x increase in EUV. That's basically compared with the output 2020. On top of that, we see productivity increase.
With DUV, we think we can have on average a 20% productivity increase, and with EUV, a 70% productivity increase, which actually gives you a wafer capacity increase in DUV of about 3x and Low-NA about 5x. High-NA, this is the first of a kind of a new generation. We need those 20 systems or more than 20 systems in the time frame 2027, 2028. We need to do this, so we're going to expand 65,000 m2 of manufacturing space, which will lead probably to an average CapEx investment of EUR half a billion per year, which over five years translate to about EUR 200 million of depreciation.
You could say if we're all wrong and we're all been dreaming and this growth will not happen, then we will be stuck with EUR 200 million of depreciation because of this expansion. I think that's bearable. I think it's also bearable in the supply chain, which we think D2F has about EUR 2 billion. But for them, of course, the math is about the same. Now, to close off. As again, I hope I was clear on how we see the global trends that fuel semiconductor growth. It's across the entire space. It's in advanced, it's in mature, it's in memory, it's in logic. That means that we believe that our customer's market could grow around 9% compound annual growth from 2020 to 2030, which needs about 6.5% CAGR growth of the wafers.
Which in order to be able to ship those machines, we need to up our capacity in DUV from 365 to 600, and to 90 EUV systems, and at least 20 High-NA systems. Okay, that was my part of the presentation. I talked about us, you know, we should do it, but I know the question will also be, what about your suppliers? Of course, ZEISS is our biggest supplier, biggest partner, and we've asked Andreas Pecher, CEO of ZEISS SMT and part of and a board member of the ZEISS holding company, to just address us and to give us his view on, one, what is ZEISS doing, and two, how committed are they. Thank you.
The semiconductor industry is highly innovative and dynamic, and so are we. In all of the ZEISS semiconductor businesses, our technology roadmap supports the industry's growth and relentless innovation. These businesses comprise our mask solutions, process control solutions, and last but not least, our optical systems for semiconductor manufacturing. 2022 marks the 25th year of the strategic partnership of ASML and ZEISS, and we have been working together since about 40 years. After all these years, our companies together are tackling the challenges, especially in demanding times like today. Looking at the growth of, for example, our headquarters in Oberkochen, Germany, one can already tell our success story. Starting with a few buildings on literally a green field, our current factory evolved. We are now further growing our capacities and ramping up with new buildings, machines, as well as staff.
Additional to our headquarters, we are also ramping up in all our German and international locations. The market demands this growth from us, and we work very closely with ASML and other customers to align our roadmaps on technology and capacity. Our aspiration is not only to be the technology leader, but also the growth leader, all this to support our customers. We need to ramp up in all dimensions, our infrastructures, our equipment, and our supply chain. We also need more educated and qualified employees, and we act fast on this. The development of our complex products takes time. For example, years from start of design to volume manufacturing, looking at High-NA, but also when it comes to DUV.
Speed is important, but at our extreme requirements, mirror precision in the picometer range and integration of highly complex mechatronic systems, it takes significant time and requires top talent from various technical competencies. It's our mission to keep the development lead times at an absolute minimum, and so we are ramping our R&D team as quickly as we can. Let me give you some insights on three of our dimensions. Strengthening our DUV lithography solutions is of critical importance to us, both by expanding our production capacities and together with ASML, further improving our DUV products. Together with ASML, we are the sole providers of EUV lithography systems.
We feel it is our responsibility to the industry and to the overall market to provide the required solutions at the required capacities as quickly as possible, because only with these products, semiconductor manufacturing at the leading edge with highest productivity is possible. With High-NA, we will drive Moore's Law to the next level. The challenges with High-NA EUV mirrors are significantly larger and of higher complexity than conventional EUV mirrors. More complex manufacturing processes must be mastered. To tackle these challenges and to bring this new technology to market, the teams are working at full speed. Our strategic partnership with ASML has a strong foundation. We are confident in facing the challenges together and to meet the market's demands.
Good afternoon, everyone. We just saw the market opportunity. We just saw all the effort that is being put into this, both by ourselves, but also by ZEISS. Now indeed, let's look at what this is going to generate, what is going to generate for us, how are we going to serve our customers, and ultimately, how is that going to result in the numbers that Peter already referred to, which of course is an important topic of conversation today. Before I do that, just a couple of things that we're gonna review before actually going into the euro numbers.
First off, I think it's important to just look backwards just a little bit in terms of, you know, what are the investments that we've made, how does that reflect in the historic performance of the company. I'm gonna talk about our model. How is the model? How do we make these different models in terms of, you know, potential scenarios for 2025 and for 2030? How do we do that? What's the methodology behind that?
Then I'm gonna borrow some of the work that Peter has shown in terms of how do we look at the markets, how do we look at the different technologies, how is the buildup there, et cetera, in order to ultimately get you to what does this mean for the different technologies that we have for DUV, for EUV, Low-NA, and High-NA, for metrology and inspection, and how does that ultimately reflect in our P&L. Then finally, what are we gonna do with the money? You know, the money that is being generated, how is that gonna find its way back into the business, and what, and in what fashion will it be distributed to shareholders?
Obviously, we're also gonna talk about the new share buyback program that we intend to start today. With that, let's go back in time a little bit, and very briefly, this is the investments that we've made within ASML in the past couple of years, starting in 2010. What you see here is three type of investments, if you wanna call it that. Obviously, the R&D expenditure that we have, which is the dark blue element of the bars. We have the CapEx, and you know, a lot of the CapEx you would have seen if you know, if you are here at our premises.
A lot of new buildings in comparison to last time that we were able to see each other here in 2018 in person. Also, quite a few cranes that are an indication for the CapEx that is underway. Then, of course, we did a number of acquisitions that are strategic that we believe were strategic in order to accomplish the execution of our roadmap. The latest one of that being the acquisition of Berliner Glas, now ASML Berlin in 2020. All of those investments and the execution of that roadmap, I think they've generated some pretty good results for the company.
All in all, if you go back again to 2010, you would see a CAGR around our EPS, around earnings per share of 18%, driven by obviously growth, one of the key themes obviously of today, growth in our revenue. You see revenue growth from EUR 4.5 billion in 2010 to, you know, around EUR 21 billion expected for this year. You see an increase in our gross margin, starting at 43% at 2010, going to 50% and up for, you know, for last year and this year. Of course, also with the share buybacks, share buybacks that accounted for, I would say, about 2% out of the 18% in terms of the CAGR.
I think all in all, I think the roadmap execution that we've embarked on, I think has clearly produced results also financially for the company. That's not gone unnoticed in the financial markets either. If you look at the development of the shareholder value for ASML over this period, 2010- 2022, it's been a very substantial generation of shareholder value. 24% CAGR in terms of total shareholder return, which is, you know, north of what the semiconductor index, which the stocks index did, and also substantially north of where the Nasdaq was. You know, some of you thought that the development of this curve was a little bit boring. You know, the only way is up.
Obviously, this year created some interesting dynamics in there, as you can see, obviously, as we've all witnessed. You know, we stopped calculating this by November first. Obviously, the past couple of days and past couple of weeks have been pretty helpful again. It's pretty clear the volatility that is there. No matter what it is, I think it's pretty plain to see that the performance of the company and the good results that the company has generated have also been reflected in the return to our shareholders.
All of that is history, and I think the question obviously is, you know, based on this and based on all the investments that have been made, but also the investments that we're embarking on, what does that mean for continued growth? Again, just to remind everyone how we do this, and the methodology has not changed. The methodology that we use in order to come up with these different scenarios for 2025 and 2030, that, you know, that methodology is unchanged in comparison to what we did last year, and also unchanged in comparison to what we did in 2018 and the capital market days before. Starting point is the end markets, as Peter just demonstrated, you know, segmented into the seven different segments.
We then translate those end markets into wafer demand for all of those segments, and then wafer demand obviously spread for different nodes, right? The different nodes in logic, mature and advanced, different nodes for DRAM, for NAND. We translate that in a very transparent way for you into the litho spend, where we give you the litho spend for those different nodes. That's the way how we come up with the demands in terms of number of tools, obviously recognizing the improvement in productivity that you're seeing, and that we expect to see in these tools.
In addition to that, obviously, we have the installed base business, which is growing with the installed base, both, you know, based on service and based on upgrades. That then gets us to different scenarios. A low market scenario and a high market scenario, both for 2025 and for 2030. For 2030, that is new because you will recall that in the last Capital Markets Day, we only gave you a CAGR for 2030. We now believed it was appropriate also, you know, given the better insights that we have, including better insights on what the technology sovereignty, what that really does, we were able to now also give you scenarios for that year as well. That's the methodology.
I will skip through this because in essence, these slides have been shared by Peter, but they are important just to follow the logic of the model that we have. Here you see the development in the end markets. Here you see the translation of that into the wafer capacity, the wafer starts per month per year, as Peter just laid that out, adding up to the 930, including the 150 from the tech sovereignty and the foundry competition. Here you have the build-up of the technology, if you like, for the different nodes. This is unchanged in comparison to last year, right?
The composition per node in terms of, you know, what do we think is going to be done by High-NA, what's going to be done by 0.33 Low-NA, what's being done by immersion, et cetera, et cetera. That composition has not changed for these. The only thing we did is also in light of the fact that we now have these scenarios for 2030, we added a few nodes on the memory side, so both for DRAM and for NAND. The composition and also the node-on-node growth as it is being projected here, as a CAGR for this period, the 30%, over 20% and over 10% respectively is unchanged.
In light of the importance of the maturity of the mature market, and in light of the growth that we anticipate there, we also thought it was appropriate to give you some insight into the composition of the different nodes on the mature side. Here you see that, you know, the one you see here for mature logic analog and also in sensors, the 45 nm and also 40 nm, you see that, for instance, there is also immersion ArFi nodes are in there. Actually, they're a pretty substantial part of the total cost of lithography for those particular nodes.
With those building blocks, and again, the majority of the building blocks, Peter has spent a lot of time explaining those building blocks. What does that generate? How does that translate into results for the company? Let's start with the installed base business. Installed base in 2020 started at EUR 3.7 billion. That was the starting point that we in the Capital Markets Day of last year. That block, the dark blue block there, or bar is the EUR 3.7 billion. By 2025, we expected in the Capital Markets Day a midpoint for that of EUR 6.5 billion. But of course, you also realize that this year we're already approaching EUR 5.7 billion for the installed base business. That's the expectation.
I think it is fair to assume, you know, growing it from 3.7- 5.7, in essence, in two years' time, I think it's fair to assume that the midpoint of EUR 6.5 billion that we have by 2025 is probably too conservative. On the back of, you know, growth in the installed base and also on the back of our success, I would say, in providing upgrades, we thought it was appropriate to have that midpoint grow to EUR 7.5 billion. Here you see that there is this growth and as a result of that, you see the CAGR that we expected from, you know, 2020 to 2025 in the Capital Markets Day of last year, a CAGR of 12% for that five-year period.
We now expect that CAGR to be 15% at the midpoint, again, growing from EUR 3.7 billion in 2020 to EUR 7.5 billion by 2025. Then the CAGR beyond that point, so the CAGR for the installed base business, starting in 2025 all the way to 2030, we expect that to be at the same level, so at 10%. Obviously off a higher base and as a result of that, obviously, you know, more growth. That's the expectation that we have there. What that ends up doing in terms of CAGR for the entire period, for the entire decade, is that it would take the CAGR of 11% for the installed base business that we discussed with you last time.
We expect that to go to 13%. In terms of systems, what are the expectations for systems? What we present to you here, primarily for advanced logic for DRAM and for NAND, a lot of it is actually consistent with what we had last year, which tells you that, you know, the lion's share of the improvement in the numbers really comes from the increase, if you like, in end market demand. There are a few changes here, and those changes are in the bold lettering in the boxes. What are the key assumptions? First key assumption in terms of market share, that has not changed.
Market share for EUV, obviously 100%, for immersion 90%, for the dry business 65%. Those were the same market share assumptions that we discussed last year. If you look at advanced logic and we look at the reference point of 16 to 14 nm, that was 315 last year. We took that up to 350 for now, because you know, it's clear that you know, our customers keep on adding also in those nodes because there is still a demand in those nodes. You really see nodes becoming bigger and bigger and bigger. The 14- 16 nm node now is estimated at 350 as the starting point.
In terms of the low and the high scenario, in terms of the node and node change, every node after the 16/14 nm, what happens there? Last year in 2021, we talked in the low scenario about -15%, and we really think on the back of what Peter showed, you know, in terms of the increase in wafer start capacity that is needed, we took that down to -5%. Node on node over node, -5%. Still negative in the low scenario, reflecting that obviously you get more bits with every new node. You know, taking that back from -15% to -5%.
In the high scenario, we had it at 0% in the Capital Markets Day of last year. We now put it at 8%+ for advanced logic. Again, reflecting the very significant demand that is clearly there as it relates to the advanced nodes in the high scenario. In terms of EUV exposures, in essence, no real change. That is unchanging compared to what we had last year. In terms of DRAM, more or less the same. Also the bit growth has not in the term that we are looking at here, 2025, 2030, that has not changed. Bit growth low 15%, high 25%.
The only thing we changed here is the assumption on wafer capacity conversion that was at 60%-70% in 2021. We took that to 70%-80% for this time. The background really is what we see our customers do. We really see that our customers are, you know, making those technology transitions. It is affordable for them to go to a new technology transition, you know, to have more EUV, for instance, deployed in there. We really see them see more and more capacity being turned onto the new technology. There we have nine exposures. Used to be eight, but of course, those 9 exposures would also include now in this model, High-NA exposures 1-2.
As it relates to NAND, no real changes in the assumption. The only change there is that we actually took the high down. We took it down from 45%- 35%. The reason there is that we do see, you know, that the technology transition is a bit slower. As a result of that, we think that affordability of that, you know, becomes a little bit compromised. As a result of that, we took that back from 45%- 35%. These are the main assumptions in the model as it relates to the system sales. What does that all translate into for the different technologies and for the different system units that we have?
If we look at the screen here, there's a few conclusions that you can draw from that. As you see on High-NA for 2025, we didn't change anything. The reason we didn't change is it's consistent with what we discussed last year. Of course, there will be more shipments in 2025 than the five that we have here. Given the complexity of the delivery model behind High-NA, and given the uncertainties that we have on revenue recognition, when exactly is the point that all of it is there, that all of it has been tested, and that in fact, you can say that the system has met all the obligations that we have vis-a-vis the customer. We took a pretty conservative stance and only have five systems in here that we recognize as revenue.
That assumption has not changed. If you go down, you see a big jump in the dry business. Both in the high scenario and the low scenario, you see that there is a big jump in the dry business. Of course, that is very much a reflection of what Pieter has been talking about the mature technology and the huge increase in demand that we simply did not foresee to this extent last time that we discussed that. Hence the big jump there. You also see quite a jump on the immersion technology. Immersion technology really benefits from two developments. Of course, on the one hand, on the advanced logic side, immersion there is an important player, right?
The fact that you saw this significant increase on the back of of artificial intelligence, automotive, and what have you, for the demand for advanced logic, obviously that is reflected in the immersion sales. I also showed you that in a number of mature nodes, the 40 nm, 45 nm, you actually saw there as well that immersion plays a significant role, right? Immersion actually benefits from those two developments. There you see that go up. You also see EUV go up. On EUV, you actually see two things. On the one hand, you see it going up from 70- 80 in the high market scenario. Again, I suppose not a surprise because this benefits from a bunch of things.
This benefits from the development on the advanced logic side that we just talked about. This obviously also benefits from the tech sovereignty and the competition in the foundry business. There it's clear that that benefits from those developments. You also see that the variability in EUV goes down, right? If you look at the delta between the high scenario and the low scenario, where you had a gap of 22, right? 70 minus 48, 4.33. You had that in the Capital Markets Day of last year. That is actually shrinking. The reason it is is that it's becoming clear that EUV is, you know, quite resilient also in less favorable market circumstances. That's why the gap there actually decreased from last year to today, on a...
Actually on a higher base. That's the 2025 system developments as you see it. If you then look into what that does in terms of sales, it actually means that the high market scenario that we had last year in the capital markets has now actually become, you know, the low market scenario at EUR 30 billion. You see that we've now modeled a high market scenario there at EUR 40 billion, which is a reflection of the unit numbers, obviously, that you see there on the left-hand side. The metrology and inspection units that get added to that and obviously the installed base business that I talked about in my previous slide. Fast-forward to 2030. What's the expectation for 2030?
If you look at that, actually what you see is, of course, there is a very big increase here on the base of High NA. Again, High NA for being a very significant enabler of our customers' roadmaps in the second half of this decade. There you see High NA in the high market scenario at a level of EUR 30 billion. As you know, for High NA already today, with the tools that we're selling today, north of EUR 350 million, so it's obvious that that is a very significant driver of the top line by 2030. You see Low NA sort of at the same level, both in the high scenario and in the low scenario.
You actually also see that there is still a bit of a dry ArF and immersion, so a bit of a DUV that gets added to the capacity in the second half of this decade. Those are the main changes from 2025- 2030. Again, please bear in mind that the tools that we're gonna sell in 2030 are different tools from the tools that we're gonna sell in 2025. Obviously the productivity and all the specs of the 2030 tools are at a higher level, and of course, that should also translate into a better ASP for those tools in 2030 in comparison to 2025.
If you take the unit increase and the ASP increase into consideration, and of course, you take into consideration the fact that, you know, the installed base continues to grow both for the service business and for the upgrade business, then you get to what you see here on the right-hand side of the screen, the total sales opportunity that you could model on the back of these scenarios. That gets you to the EUR 44 billion-EUR 60 billion opportunity that we see for ASML at that time frame. If you try to analyze it, where does it all come from? First off, the quick analysis from 2024 last year to 2025 this year, this is another cut at it.
You see that, you know, the jump at midpoint EUR 27 billion that we had in the scenarios of last year to the midpoint that we have for the scenarios this year. You bridge EUR 27 billion-EUR 35 billion. You see EUR 4 billion coming out of that from EUV, and that's units in particular. You see non-EUV accounting for EUR 3 billion, and you see the installed base businesses that are referred to jumping from EUR 6.5 billion-EUR 7.5 billion, and that gives you a EUR 1 billion delta there. That's the way to come from the midpoint of last year to the midpoint of this year.
If you go to the 2030 projection, how do you get from the midpoint 35 for 2025 to the midpoint 52 in 2030? EUV is a very significant driver of that in EUR 10 billion. Of course, the lion's share of that's pretty obvious, is the number of High-NA tools, right? That's the key driver there. But there's also an ASP increase because as I mentioned to you earlier on, of course, the EUV tools that we're gonna sell by 2030 have higher productivity, better overlay, et cetera. As a result of that, more value to our customers, and as a result of that, we'll command higher ASPs than the tools that we're gonna sell by 2025. Non-EUV business is gonna chip in EUR 3 billion.
As you saw, there is a unit number increase there as well. Also there you will see the continued dividends, if you like, of the roadmap, where we do produce tools, you know, that have higher performance as a result of that higher value to the customer and higher ASPs. The installed base business at the midpoint gaining EUR 4 billion from 2025- 2030. This is the way it's built up. If you then look at the CAGR that we have there, you actually see that the CAGR for 2020 to 2025 grows from 14% to 20%. Then you see that the CAGR from 2025- 2030, we keep at 8%.
Again, we keep it at 8%, but of a significantly higher base by 2025. Very often we get a question, you know, what does that mean for litho intensity? That's a term that is being used, albeit in very different ways. As you know, ASML is not a fan of WFE because we think it's a lagging indicator. But what we would want to present to you is how do we see the litho sales in comparison to the semi sales? That's the percentage that you see here. It's an interesting journey as you look at it historically, and the people that have been in this industry for a while, you'll probably recognize a few things.
They will probably remember when, you know, the market was very, very tough on the litho builders in 2009 of the financial crisis, where, you know, the sales levels really plummeted in that period of time. Obviously that also, you know, going up very, very quickly as soon as people started to realize, hang on, we've been too stringent here. There is a need for capacity, so new stuff needs to be added. You see the period of multi-patterning. You see the period of the conversion of 2D NAND to 3D NAND. That's the period that you have here. Here you see the periods in fact of EUV, right?
Where EUV, you know, stepped in, and EUV really, you know, led to a situation where, you know, where multi-patterning was being replaced by single patterning EUV. As a result of that, you see the litho sales as a percentage of semi sales go up. You see that at the lowest end of our estimates, you see that that's sort of more or less neutral. I would say more or less a neutral or steady development from where that percentage is today. That's where some analysts, that's the way some analysts are looking at it. You know, that's the gray line that we reference as external sources.
You know, possible, but we think there are quite some reasons why we believe there is potential over that and why we believe there is potential to get to a continued increase of that line. You see those on the left-hand side of the slide. First off, you know, this market is clearly growing. The semi market is clearly growing, and obviously you need to have your capacity in before the market growth, right? You have that acceleration, if you like, of the capacity that needs to be added before it actually returns semi sales. That element of acceleration CapEx needs to be there before you can actually turn out the semi sales.
Of course, that's one reason why we think there is reason for this for a continuation of that upwards, upward trend. The second element we truly believe, and I think that is recognized by many analysts, that the equipment portion within the overall CapEx is going to increase. With High-NA, we believe, you know, within that envelope of equipment sales, we believe there is, you know, reason to believe and reason for positive momentum also for the litho business. For that to be a higher percentage on a go-forward basis, that would be a second reason for to expect that increase.
Thirdly, the tech sovereignty trend leads to lower utilization, so that in and by itself is also a good reason to assume that there will be a continued move upwards of the litho sales in comparison to the semi sales. How is all of that going to reflect? We just looked at the top line, so how is all of that going to reflect into our overall financial model? Well, this is what we presented to you last time, so let's focus on the deltas for 2025. Of course, we did talk about the top line, so I won't spend too much time on that. You see that the gross margin we've kept flat in comparison to last year. Some of you might say, "Well, wait a minute.
If you increase your top line, shouldn't you expect a higher gross margin? Remember that the mix effect of having more dry business in there is having a detrimental effect. You know, on the one hand, of course, with a top-line increase, you would expect an increase in the gross margin, but the mix effect of having more dry business in there probably takes it down a bit. That's why we keep that at a 54%-56% level. You see R&D here at the midpoint, at both midpoints, midpoint R&D divided by midpoint of sales at around 13%, which is a number that you've probably seen before.
You see SG&A at 3.7% of that top line at the midpoint, which is consistent with where we had it last year. Then you see CapEx going up. This is the number that Peter shared with you. We expect every year to have about EUR 500 million increase in CapEx. That's the one you see there. Then you also see the effective tax rate going up a little bit as a result of some of the global tax debates that are going on. In terms of 2030, we talked about the top line.
We do believe in 2030 that there is potential to further increase the value of our tools to customers, and as a result of that, also further increase our gross margin. We believe the bandwidth there to look at would be in the 56%-60% range. You know, why is that? You know, it's because at that stage, we believe we've taken High-NA to such maturity that it will generate the gross margin that we think High-NA has potentially there. As we also mentioned, also for EUV in particular, we do believe there is, you know, the potential to further drive the value of those tools for the customer.
That's why we think a 56%-60% potential for gross margin would be there in that timeframe. R&D midpoint around 12%. SG&A, you know, at a certain point in time, you will get some leverage from there. We think we can take that to 3% by 2030, away from the 4% that we had in 2021. Then you see the rest of the numbers there, essentially unchanged in comparison to 2025. Good question indeed. A good question, you know, what about the uncertainties that we're in today?
Because of course, as Peter already mentioned it in his introduction, it is interesting to talk about growth, but we shouldn't, you know, we shouldn't forget that we're talking about, you know, significant growth opportunities medium term, longer term for ASML. We should also recognize that the global economy is pretty unstable at this stage. Very good questions from some of you, indeed is, you know, how do you deal with the flexibility around that? If for whatever reason that growth doesn't manifest itself, the capacity that you've built, you know, is not being fully utilized, what does it do to you and what are the mechanisms that you have? You know, what are the risks?
You know, look at the forward statement, that gives you all the risks that are out there. I advise you to take a good bottle of wine with you before you read it. You know, if you digest that, then you see all the risks. In essence, if you think about it, of course it's geopolitics. Of course, it's the global economy. Of course, there's a number of things that are under our control. It's the, you know, how effective are we in getting to the ramp? How effective are we in pushing our roadmap, in keeping our costs controlled, et cetera, et cetera?
Those will ultimately be the things that determine our success in the marketplace and ultimately also, you know, the level of sales that we're gonna have. If for whatever reason, we are looking at a down cycle, I think these are important numbers to share. On the one hand, we have the workforce, most of our people are fixed as you can see here, but we do have flexibility measures there. We do have what we call an hour bank that also in the past has given us quite some flexibility in that regard. You see that, more than a quarter of our R&D expense is flexible, so that gives you some leeway.
Very importantly, you know, we have flexibility in our total cost of goods, because as you see here, the total cost of goods, 82% of our cost of goods is actually materials that we source from outside. There, of course, is flexibility. This is not flexibility you can drive to the extreme, right? I mean, at the end of the day, our supply chain is very important to us. There is no doubt in our mind that we need to take measures also in a downtime, you know, to keep our supply chain happy and alive. It is clear that there is quite some flexibility in here.
On the basis of contracts, we have quite some room, given our contracts, you know, to have flexibility with our suppliers. I think this is a very important element for us to drive. As I mentioned, it's important to make sure that all of our stakeholders, even in the more difficult times, that we're able to satisfy their, you know, demands and safeguard their existence. Therefore, financial flexibility is another very important element for us. I think that's a good segue into my final comment here, and that's about our capital return policy.
Of course, in the days that we're going to enter into, you know, the fact that we're growing our capacity, also the fact that we, you know, that we're looking at, you know, potentially slightly more volatile times, of course, it's important that we have a certain financial flexibility there. I think you've seen us build that over the years. We'll continue to do that. We do believe that there will be sufficient and, you know, substantial money available to shareholders, available, you know, both for growing dividends, and that's been our policy. Our policy has not changed in comparison to last year. There will be ample room for giving back money in terms of growing dividends.
There will be ample room, we believe, for share buybacks. Share buyback program that we've announced today is a three-year program of EUR 12 billion, of which 2 million shares would be for employees, and the rest would indeed be repurchased and would be canceled. That's the program that we've announced today and that we will execute on, you know, starting today. Friends, with that, closing remarks. Investments create value. We've made investments, but I think it's also plain to see that those investments have generated significant returns for the company and for its stakeholders.
Substantial growth in the markets, both end markets, but also technology trends that we believe are significant and positive for us. Market opportunity as I laid out, market opportunity for 2025 between EUR 30 billion and EUR 40 billion. For 2030, between EUR 44 billion and EUR 60 billion. All of that, if executed well, will give us ample room to pay back money to our shareholders with growing dividends and with share buybacks. With that, thank you very much, and I'd like to hand it over again to Peter for his closing comments.
Do we need this?
I don't know.
I don't know.
I don't know what you gotta do.
I don't think so I'm just going to give it back. Yeah, closing comments. I mean, I'm not going to repeat what we said, not even in summary. The only thing I would like to actually say is that the first time that we did, you know, analyst days like this or capital markets days like this, I still remember that the company had EUR 3 billion in sales. We came out and we said, "Well, you know, for a certain period of time, we can grow to EUR 5 billion," and people laughed at us. Then we said when we were at EUR 5 billion, there was a certain period of time we will be EUR 10 billion, and people laughed at us.
It's not about and then we did it again and again. I think it's the fourth or the fifth time that we're doing this. Every time, it just turns out that the industry grows faster than we think. That is basically a structural underappreciation of the value of Moore's Law. I said it before, I'm not going to answer any questions. I cannot answer questions. 20 years ago they said, "Peter, what's the next killer application?" Single. Now they say, "What are the next killer applications?" Multiple. I have no clue.
The only thing that we do know, and that's why these roadmap slides are so important, and that's why the collaboration slides with our customers are so important, that we do see a very clear roadmap that this technology can keep reducing the cost per function, and that the functional increase in the minds of our customers is much bigger than the cost increase, which will create next level applications of which we have no knowledge of. I think this is where we are today. I think we are highly confident in our technology roadmap. We didn't spend time on the technology roadmap this time. We also said we weren't going to do it.
I can tell you, and I think it's probably also part of the Q&A that we'll have later, that our confidence in our technology roadmap has gone up, like it should go up with the type of R&D money that we spend, together with our customers. I can also say that the collaboration models with our customers, especially with the leading edge customers, driving the roadmap has been better than ever before and closer than ever before. There's much more, let's say, knowledge sharing, transparency sharing, so we are more aware than ever what the challenges are and how we as industry participants should manage this ecosystem. Having said that, it's very important basic conditions for our expectation that this industry will grow significantly.
Yes, last year, you could say when we did the Capital Markets Day 12 months ago, we started to prepare this 18 months ago, we are finding out every quarter, every year, new types of applications that are not only at a leading edge. The biggest surprise, and I hope that was clear, is also that we're seeing these applications that have to do with sensors, that have to do with actuators, that have to do with power management, have to do with energy transition, with the big challenges and changes that we're currently seeing in society. I think we will play a pivotal role there together with our customers and together with our suppliers. This is why I was very happy with the contribution of Andreas Pecher from Carl Zeiss, because they see the same thing.
We are ready, we will grow, and we'll grow significantly. Most importantly, I think this technology will provide building blocks, very important building blocks, and sometimes cornerstones for the solutions that we need to together create to deal with some of these very large societal challenges. I would like to leave it with that and give the word to you, Skip. Okay, thanks.
All right. Thank you, Peter. Thank you, Roger. We now are gonna take roughly a 30-minute break as we set up the Q&A panel. When we come back, we will have six members of ASML's management team, which we can then turn the questions over to you in the audience and also you online in the webcast. If we could come back here at 3:30 P.M. CET, which is roughly 30 minutes from now, we'll start again, and you on the webcast can take a break away from the screen. Thank you. Welcome back. We have about, I was told about 1,200 people online, so we'll be taking questions from the audience here as well as online.
Joining us for the Q&A, our panel of six of ASML's management team. Peter Wennink, you already have met. For those that are joining maybe from the webcast late, Peter Wennink, our CEO and President. Martin van den Brink, our CTO and President. Roger Dassen, our CFO. We have Christophe Fouquet, who is our Chief Business Officer. We have Frédéric Schneider-Maunoury, who is our Chief Operations Officer. And lastly, Wayne Allan, who is our Chief Strategic Sourcing and Procurement Officer. Welcome them, and let's start with the Q&A from the audience. Maybe right there. Go ahead. Oh, yeah, wait for the mic here.
Thank you. Jérôme Ramel, BNP Paribas Exane. Roger, just wanted to come back to the slide you show during the last Capital Market Day and today, the average between the high scenario and the low scenario of your EUV numbers, because you said that the incremental in value is EUR 4 billion. When I do the math, I find like 15 new EUV tools. That's an average which is, well, in terms of ASP, closer to EUR 300 million and EUR 200 million. Just wanted to double check the number if I got it right.
I said primarily, right? It's primarily driven by number of units. It's also rounded. The 4 before you know it could be like 3.5. I wouldn't adjust my ASPs the way you just said, do the way you just did it in your mind. I think the ASPs that you have in your mind for 2025 are probably good. The EUR 4 billion, I think the lion's share of that comes from unit to unit numbers. You have the right number of units.
I think it was 13, I think on the midpoint. Then you multiply that with something that you probably have over, you know, low 200s somewhere. That gives you a substantial part of that. Then there is a little bit other in terms of ASP, and then you get to what is probably closer to 3.5 than to four. That's the way to get there.
Okay, thank you.
You're welcome.
Maybe just a quick follow-up for Peter. Peter, the slide you show about the geopolitical potential upside, you just show the plus, I would say. Is there any minus from China?
Well, I think this particular one geopolitically has to do largely with the drive of the governments around the globe to bring or to reshore semiconductor manufacturing, which is in first instance what we're seeing today is advanced as leading edges EV related, which got nothing to do with China. I think it is basically. You can just look at the slide where we just added up all the announced fab projects of our three biggest customers, just three customers, yeah. It's more, I think, EUR 30 billion. You see how it's spread around the globe. It's basically. There's a lot of what used to be Southeast Asia is now moving back to Europe and is going, you know, to the US. That doesn't necessarily include China.
If you say, fine, if the geopolitical situation is such that, which I would not expect, but that, for instance, China would absolutely be excluded from any growth, then what I said earlier in my presentation is the demand is there. Yeah. Those fabs will be built then somewhere else, you know. It could mean that there could be a temporary hiccup, yeah. Ultimately, those chips need to be, you know, made. You could argue it is still as inefficient or perhaps even more inefficient and not less inefficient. Chips will be needed, yeah, to the levels that we talked about. Where they're being made is going to be a matter of time. Yeah. So it doesn't change the, let's say, the 2030 picture that much.
Next question. I just want to bounce back and forth up here between the side of the room.
Hi, it's Janardan Menon from Jefferies. Just two questions. One is, on the technology roadmap, you have you obviously got the High-NA coming out, but you know, 0.55 NA is not really high in the broad scheme of things when you look at the historical trends of lithography. So can we assume that you're already working on a higher NA going up to presumably I mean, you can take a non-immersion up to around 0.9. Would that be the direction of travel over the into the first half of the next decade? And associated with that, one thing that you didn't touch on on the model was double patterning on EUV.
Are you completely excluding that possibility, either at the 2-nm gate-all-around in 2025 before your High-NA goes in, or beyond that as you go to 1.4, 1 nm before the end of the decade? Peter, the second question is-
Janardan
What on your model and framework on the things like, you know, increasing die size, and you've got the 50 additional with 50,000 wafers? Where does cost come into this picture?
Yeah.
You know, clearly, to a certain extent, the cost part of Moore's Law hasn't progressed as well as it used to in the decades gone by, and the value of products is increasing. Does that get any pushback either from your direct customers, which I think perhaps you're not, but also from the end consumer because the product cost is also rising? How does that fit into the whole equation?
I think Martin can basically answer your first question, because I think the two questions that you combined it as one question are, you know, related.
Sorry about that.
On the value question, because that's basically a value question. I always say Moore's Law is not only about cost. I mean, it's a big part of economics, and economics has two sides. It's cost and value. To your point, if we keep increasing value, and I gave an example of the M1 Apple chip, you know. Because what's the value there is the energy efficiency of that chip while the performance is still high. When that's the case, then that value, ultimately, we and the end markets need to determine whether we think that that value is good enough to buy the next generation phone or to buy the next generation, compute power. And I think that will drive it.
I think when you then look at this, yes, costs will go up, but it ultimately is the cost more than compensated by the value that we're willing to pay. That's always been the case. You know, Janardan, you and I know each other for more than 20 years. I mean, that's been the driver of the industry. Yeah. Also now what we are seeing today is what this is why this sharing of the customer roadmap going forward, what I mentioned, is so important that for that till 2030 and beyond, there's this transparency and openness that we absolutely need with our customers. They're willing to give, yeah, to make sure that the economics still work. Yeah. Because complexity will go up, and the flip side of complexity is cost, to your point. Yeah.
You have to make sure that that cost per transistor, cost per bit, cost per whatever you wanna, you know, express it in, that that cost can be compensated by more value. This is exactly what we are doing with our customers, and they're doing with their customers. It's a very integral, very integrated job that we have to do as an industry with our customers for the next decade and beyond. That's happening today.
Martin, you want to take it first?
Well, first of all, EUV double patterning happens today, so you shouldn't look at double patterning will it be used then or not be used, it all happens today. In fact, double patterning started 50 years ago on immersion, and it has, as of, let's say 50 years ago, decoupled a bit, the lithography capability of the machine versus the process complexity of the customer. It also is a way for us to measure what affordability is for lithography when you start going into the next generation, because a customer can stay with the existing generation and do double patterning to the extreme, preventing the next generation coming in. This is happening today, we are developing high NA.
High-NA is about to be launched around 2024, 2025, and that will be used for reducing the double patterning EUV. That means the value we can assign to the tool is depending how much cost attractiveness there is to shortcut double patterning. Something similar will happen moving forward. Now, to put your mind to rest on a 0.55 is High-NA because in an EUV system it's pretty high. I think practically impossible to assume that the NA go as high as we used on 193. When we talk to the next system, we not talk about High-NA but Hyper-NA, then the question is what will happen after Hyper-NA. The lenses becomes bigger and bigger and more costly.
Our cautious on the next step is that we have to make sure that whatever the next step is, it's more cost effective than maintaining using the current tools for double patterning. We are way more in a position than we were ever that we have to make sure the complexity and cost of the next generation remains attractive enough for the customer to replace double patterning on the previous generation. That means we're looking now for a possible next generation towards the end of this decade, possibly in the first phase of the next decade. That's the timeline we're looking for. I think Roger has not included in his scenarios anything of that.
The major value proposition when it comes to the next years until 2030 is mainly on improvements we're gonna make on the existing tools, which are also quite extensive. That is, I think, assumed, and you said it even in your presentation, you include productivity, major improvements in cost are rolled out both on 0.33 NA and High-NA even prior to Hyper-NA will play a significant role. Of course, this is also not limited to EUV, it will also go along with the DUV. Not sure it captured the question you asked.
Just a quick follow-up, Roger. On the model that you said, going to what Martin said, is there double patterning included when you say 80 EUV tools and things like that? Is there a double patterning number included in that at all?
No, it's included in there.
It's included.
It's included in there. Yeah.
A customer using it today. That's what I said.
As Martin saying, it's being used today.
Understood. Thanks.
What will happen is. Probably I should add one story is the shrink itself, right? If you look at the historic shrink, the density has been going up, dimensions go down. We see today the density is being a combination of what we print on the wafer versus the process integration, the 3D integration, if you will, customers are able to do. Those two gives you still another decade of shrink. The amount of shrink is reducing. Where it's also important to realize is that how far are we able to work with our customers on the visibility and roadmap around this end of the decade? Will there be still products out there to take it? I think that part becomes clearer as more.
As we start interacting with customers, and it becomes more visible.
Yeah.
Similarly, ten years ago, when we funded the pre-development of High-NA, I didn't know that the DRAM and the logic would make use of it. For now, I think has been expressed by everyone, it's a very solid entry both at memory as well as logic. For the next NA tool, we still have to solidify that with customers, including the affordability and making sure we are able to adjust the complexity of a next generation tool accordingly.
All right. Thanks. Peter?
Hi. Thanks for letting me, Sandeep at J.P. Morgan. Just quick question, Peter, on the ASPs, because, I mean, your growth till recently was, you know, units were coming down and ASPs going up.
Yeah.
Now that model has changed as such, really, given the demand environment as such. How do you expect, to 2023-2030, the ASP trend on the tool? What is going to be driving it? Is it just going to be throughput? Because there is High-NA coming, but High-NA in terms of volumes is not like how in terms of how EUV brought in the volume. I have one follow-up for Roger. This year we saw this big impact of inflation on your margin in the first half of the year. We don't know what inflation is going to do if we are in a new world of continuing inflation. Are you building into your model, and particularly in your contracts with your customers, what will happen if this inflationary environment continues?
Okay. Let me first answer the question on the ASP. Yes, you are right that what we always said in the previous capital market days is that there was a level of cannibalization that we expected when we introduced EUV, that immersion would go down because some of the immersion layers would then double patterning, multiple patterning layers would move to single layer EUV, which actually happened. What we didn't know is that, and we didn't expect that the end markets, which we talked about, grew so fast. It actually means we're shipping more immersion tools than ever. ASP going up and volume is going down is now ASP is going up and volume going up. I think it's a better proposition, by the way. That's that. This is where.
That's driven by value. It's driven by productivity increases, 'cause it will lower the cost per wafer, yeah? It's also driven by what Martin just said, you know, if we can determine the value of the end product together with the customer and the customer customers, and we can control the economics whereby the value increase is higher than the cost increase, we go, yeah? Then there are the building blocks of the technology there that we need to determine with the customer, and especially on a, you know, High-NA tool. That's still there. We're working. This is why we need that close collaboration.
The way we see it today is that the further shrink, the further productivity will lead to lower cost per whatever you want, function, bit, which will then drive the top line of the company even faster than we thought.
Clear.
So I'll take the-
I'll take.
I'll take the question on inflation that you raised. You're right. I mean, we're all taken by surprise, I think, about the size of the impact of the inflation. Obviously what we have in our long-term financial plans is that inflation plays a neutral role. How to do that? That means that you need to not just work with your customers and have in their contracts that you know that they have to pay for whatever inflation is, but also work with your suppliers and see to what extent you can hedge the situation as best as possible.
That's what we're working, not just with customers and the new customer contracts, but also with our suppliers to make sure that the burden, if you like, of inflation is fairly distributed throughout the ecosystem. It's unfair to only have it passed on to us. It's unfair also, I would say, to only pass it on to the customer. We need to have a fair distribution, but, I would say, not the way we're having our negotiations now. I think it's fair to assume that inflation should have a neutral impact on us on a go-forward basis.
All right. Thank you. I mean, let's take one from the webcast here. This, Christophe or Martin maybe. In Roger's presentation, High-NA was projected to be deployed in DRAM. Does this use of High-NA EUV for DRAM mean that 3D DRAM will not be in high volume manufacturing before 2030? Christophe?
Yeah. I think in Roger's presentation, he highlighted the use of High-NA in DRAM as one of the change, in fact, compared to the message we gave last year. I think this is a result of the discussion we have with our DRAM customer. The resulting better understanding of their own roadmap. At this point in time, I think there is a need for High-NA DRAM around 2024, 2025 for insertion. After that, basically, I would say going into volume. We are far more confident today, far more aligned with our DRAM customer that High-NA will be used also for DRAM, which I think is mostly a new element compared to what we shared with you a year ago.
The answer to your question is yes, this really meant that High-NA will be used in DRAM. The timing for volume, I think Martin already said it, we still look at 2024-2025.
Let's see what side. We're over here. Sorry. Yeah, Tammy, if you wanna...
Luckily, we are looking into the light, so we don't see all the.
Sorry. After about the fifth row up, I start losing it, faces.
Thank you. Tammy Qiu from Berenberg. Firstly for Peter, from your node-to-node assumption, which is definitely way more positive than last year, what's the driver of that? It's just purely server, and what's your assumption there for the adoption of chiplet?
I think it's a part of it. I mean, part of our model takes into account the architectural choices and the system integration choices that our customers make. We have pretty detailed model, the translation of the demand and the roadmaps and the node-to-node roadmaps. In the close discussion we have with our customers, we can actually assume a certain percentage of the devices that will ultimately be chiplet-based and others that will be just mono-based, yeah? That's in the model. It's in there. Now, what the split is, I don't have the data exactly, but I don't think we should. It is in there.
Based on the roadmaps of our customers and what we know of our customers, it's there.
Okay. Thank you. Also for Martin, please. Martin, may I ask you, do you believe in High-NA firstly? Also, from my conversation because you definitely quoted from different presses, you have different opinion on High-NA over time.
There has been. I have been criticized by my colleagues who are giving an interview for a local newspaper on what my take on High-NA is, and I was toned it down a lot.
Because the reason is what I just explained, that I think at ASML we are more than before way more cautious about the cost part and the complexity part of the machine. I think it will go with Sheila to the tour half hour from now, you will see things get bigger and bigger and bigger, and we have to be more constrained so that because the NA change will be relatively small, so if something happens, there's enough economic power. Do I believe I will get there? The answer is yes. I always believe in technology, so I will believe we'll get there.
I think it was very useful and I'm still of that opinion, to say to the local environment where the suppliers are, "Be careful, we have to instead of gung ho, Hyper-NA comes whatever it takes," that will not be the case. We have to constrain ourselves and looking to cost-effective solutions in order to make sure that we be competitive to whatever alternative the customer has in terms of processing. That includes double patterning previous generation tools.
Okay. Thank you. Lastly, Martin, what will happen to ASML roadmap after you retire?
Now, first of all, when that happens, I will probably not be there, and others will have to give the answer what will happen after I left.
Thank you.
All right. Sorry.
Right. Right here? Down here. Ready?
Yeah.
Thanks for taking my question, guys, and thanks for putting on such a great event. You guys emphasized very heavily the industry roadmap and how much conviction you have in it. It is a really special thing about your industry, how you know all participants do get aligned and on board on exactly where the whole industry is heading, exactly the role ASML will play in that. For us as the investing community, the public information we have available is more really about process nodes that have been announced by foundries already. For the next few years, it's clear exactly to us how foundries think about that, and you know how processes should develop.
You know, maybe beyond 2025, for the next five, 10 years after that, you know, imec has their roadmap, IRDS has their roadmap, but it's often less clear to us exactly how, well, not only exactly what the processes will be, but also how firm the industry is and exactly what the processes will be beyond just the fact that they'll keep advancing. So we'd love to kinda get your help understanding, how tight do you feel on exactly how new processes will look beyond 2025, both in terms of, like, the process nodes that will advance, but also the litho intensity and litho need, advancement beyond that.
I know that's a bit of an amorphous question, but would love to just get some of your thoughts to help us get our heads around it a bit more.
Well, I guess probably we should share the question a bit.
Yeah, yeah.
Among our team, but let me start off this.
Yeah.
If you remember, I believe, two years ago even, I showed this roadmap, which I copied Mark Liu of TSMC, where he showed for the next 20 years an exponential improvement of what he called energy efficient performance. That is done mainly through system innovation. If you look to people like Intel, they show that as the amount of transistor per chips, per device, integrated device. I think, what that expresses is a confidence of our customers' customers that, the enormous compute power will still have enough, headroom to improve. Either we do that without shrink, and then the square meters will just simply explode.
Explode. Yeah. Yeah.
We do that with as smart shrink as we can. I think as long as our customers and the customers of our customers, being the fabless guys, see sufficient innovations in the system, on the system level, which I think we just start exploring it. I mean, most of the time in chip industry, it was mainly getting things to half and everything comes for free. System innovation is starting, you know, for the last 50 years, but it will continue to grow. As long as we have ideas to improve the system, that will happen.
The value of that in terms of energy use of data centers, energy use of your phone, and I would say the other way around, given the battery power is about the same, you will get more functionality for that same amount of power. That I think will be, in my belief, sufficient to keep this thing going for a while. The only question is how far could we really participate in getting that cost down. That is still an important role we play, and we very comfortably can continue to play for another 10 years.
Yeah. I think basically what Martin is saying is what I said earlier. You know, as long as, yes, it's going to be more costly and it's going to be more complex, but if we can create the value, and what he's basically saying is there going to be an end or a significant slowdown to the currently insatiable demand for compute power? Yeah. If we see an end to compute power, then yes, of course, there's. You know, why would you do this? There's nothing in sight that actually points. On the contrary, it points into the other direction.
You know, when we think about the type of applications that we're currently thinking of and along the horizon, that has to do with AI and all the things that I talked about in my presentation. It needs more compute power, not less. Is that compute power affordable? That's the question. That's the complexity and cost question, which we need to solve together with our customers. Geometrical scaling has always been one of the scaling engines, yeah? It's more important than ever that the integration and the combination of those scaling engines, the system integration, together with our customers and the key players in this industry, drive to indeed creating more value at lower cost per unit. Yeah? There's nothing in sight that we currently see, neither at us.
Technically, I think High-NA will happen. Can we then work it together with the customers? By the way, the customers lately, yeah, have proactively started asking questions about High-NA.
Well, they need some help, I guess.
Yeah, they need, yeah. That's also true, yeah? It's there, yeah? At imec, you know, Robert, I mean, it's a. That's all pre-competitive. It's the industry working together to look at the feasibility, huh?
That's super helpful. Just to round out the discussion, so when you look at the imec roadmap, is there anything outside of what's under ASML's control that is a potential risk to the roadmap? Like, are there any things out there that you would worry about that, you know, are outside of ASML even?
Well, there must be many. They should ask me. We should have a bigger panel then, because there are many elements that get in the way. As a kind of the naive confidence that I have, that an industry that has an earnings power of EUR 688 billion last year with growth at a compound annual growth of 19% for the last six years has one, the money, but also the attractiveness for the brightest minds to find solutions. Just like people told us 15 years ago, "You guys are completely irresponsible by stepping into EUV, yeah, 'cause it will never work." Yeah. Guess what?
Very good. Mehdi.
Thank you. It's Mehdi Hosseini from Susquehanna International Group, and good to see Martin on the panel. I wanna go back to Martin. With 38, NXE:3800E coming out next year, should we expect to see incremental improvement in throughput, like starting from 185, 200- 220, or is E gonna come out with 220?
Well, I'm not sure what's public or not, in terms of specification, but we are on EUV. We have a solid roadmap going to, you know, up to 400 WPH over time, and beyond that, it's still not excluded to go beyond that. We do that by major innovations in the optics, because it has to deal with the amount of heat from the power. It has to do with the CO2 system, that means the whole drive laser. It has to do with the source itself, the plasma and the frequency of the droplets. It has to do with just running faster the stages. We also have plans to improve the transmissions of our system.
We have major, let's say, a rich roadmap ahead in driving the productivity, which effectively drive down also the energy consumption per wafer as well as the cost for the customer.
The public roadmap that is out there has 160, as you know, for the current tool. Has 220 for the ultimate specification of the 3800, but there will be in all likelihood an interim step of 190.
Okay.
Yeah.
Just on that note, given the incremental increase in EUV demand, how should I think about the compute and commercialization of like a pellicle enabling larger ASIC dies coming out? Are we at the inflection point where there is really an inflection point in diversification?
Starting on that one.
I think, I mean, at least so far, you know, the die size doesn't have any impact on the pellicles. I think the two things are completely independent. I think the pellicle has been initially a very difficult topic for EUV because I think for a long time we didn't know how to do it, and then for a long time we knew how to do it, but we are not sure that this will survive, you know, the condition of volume manufacturing. I think at this point of time, the pellicle is available. It's being used or not depending on the layer, depending on the yield risk on the layer. We are basically continuing to improve the performance of the pellicle.
I think that the biggest disadvantage of the pellicle was a loss of productivity, which we have been also improving over time. We continue to work on that. We have a few challenges. One of it is to continue to increase the transmission to reduce the impact on productivity, and the other one is to make sure that, as we increase the power of the system, the pellicle can deal with that. I would say at this point of time, this is more work we have to do than anything else. We don't have major concern about, you know, the pellicle. To be honest, we're also working very hard with our customer to try to reduce the need for it as much as possible, because it has an obvious productivity benefit.
Can I squeeze in one more for Peter?
Yeah, sure.
A quick one.
You have the microphone, so that's easy.
Your presentation on end market trend, they all had a dip in 2023 assumption and lower forecast for 2023 compared to Capital Markets Day 2021. I'm not asking for a 2023 guide, but if I were to take the updated slide, should I assume that there is some sort of an air pocket with your customers, but they're gonna continue with the investment? If your customers are gonna see a lower revenue, they're just gonna have to find a way to continue the investment. They take a gross margin hit, but they won't change their investment plan. At least on the litho.
Yeah. I think you'll always have these unpredictable events called, you know, cycles. The root cause is always different. In this particular case, we can probably have a long discussion about what the major root causes are. But it is what it is. Also going forward, I mean, especially for 2023, we've of course included that because that looks like it's the reality of today. Now of course, people say, "Well, you guys seem to be pretty optimistic on the fact that your business for next year is still pretty good, you know, while some of your peers say, oh, it's pretty bad." You know? If.
Effectively, what it comes down to is that we have to conclude that the average customer or our customers' idea about the length, the potential length of the recession is shorter than the lead time of our tools. 'Cause that's what it is, you know? 'Cause if it's different, if they think that the recession will last much longer than the lead time of our tools, then they can start pushing back. Now you have the risk, if they don't believe that and they think it might be shorter than the lead time of our tools, and we're pushing back, guess what? When upturn comes, you're in the back of the row. Everybody knows, 'cause we've been pretty clear about this, that the current demand that we have is higher than our build capacity.
This is the situation of today. You have, well, okay, do our customers then have the money to do it? Well, some do, some don't. This is the situation where we are today. The fact that if some of the machines are being pushed back in a cycle, there are other customers that will pick those machines up gladly. Yeah? This is where we are today, and this is why 2023, without giving any guidance to 2023, but that's the situation where we are. It's really, I think it's driven by, obviously, I cannot conclude anything else than the idea that our customers have about the length of a potential recession. Yep. All right, let me take a couple here from online, 'cause we're queuing up quite a few here.
For Wayne, maybe this is more background from your past, but as your EUV install base grows over the coming years, can you help us understand how customers would trade off improving productivity by upgrading system versus purchasing new? Is it similar to how we should think about DPU, you know, how DPU has evolved? I think they're really just trying to understand how customers think. Both from your CS and your prior customer experience, how do you view?
Yeah, I think obviously our customers would prefer to upgrade and then not have to use up more fab space. If we can provide an upgrade at a cost that allows the throughput to go up cheaper than buying a new tool, that's what our customers will prefer. I think at some point, if we're not able to provide an upgrade that is cheaper than buying a new tool, then the customers will continue to just add capacity.
Frédéric, this one is on cycle time for EUV and DUV. Comments or questions on anything you can comment on the current cycle time and/or some of the key challenges both for DUV and EUV with respect to manufacturing cycle time?
Well, to freeze the capacity that we need, we have all the engines working. One, of course, is permanent improvement on our cycle time, which has several factors. A factor which is to make sure that we have the parts on time with a high level of quality, and this is all the work that we do with our suppliers. Second, mainly with the engineering folks of the company to improve permanently the sequence. Year after year, we have close to a double-digit objective to reduce cycle time on all platforms. Of course, it's flattening a little with the machine that are more mature. For machine like the new machine EUV or tomorrow High-NA, we are far more than double digit objective of reduction of cycle time year after year.
All right. Thank you. Let's go back to the top there.
Good afternoon. Amit Harchandani from Citi, and two questions, if I may. My first question is with regards to capacity expansion. As you go from doing less than EUR 20 billion in sales to potentially up to EUR 60 billion by 2030, clearly, so far, Veldhoven has been the epicenter of your operations. But as you look to grow the company over the next 10 years, how are you thinking strategically about adding capacity? Would be looking to locations beyond Veldhoven to emerge as system integration sites, particularly given this whole talk about reducing the risk profile within the supply chain and the broader sort of diversification, if I could call it, of chip manufacturing locations. That would be my first question.
Secondly, if I may, if I look at the lithography market longer term, you're clearly gone well past 90% market share, 95, potentially even higher. How are you thinking at the opportunities beyond lithography? For example, we haven't talked a lot about metrology and inspection, but you're sitting on a lot of data that's coming out of your EUV machines. Strategically, how are you thinking at the world beyond lithography? Do you see more opportunities in process control, for example? Any thoughts on that would be appreciated. Thank you.
I think Martin's or Christophe can take. Chris, why don't you take the second question? I'll. On the first question, I think Roger showed it. I think 82% of our system cost is outsourced. This question of. We're a system integrator, and when we talk about the 9,620 plan from a system integration point of view, if, like Frédéric says, we're reducing the cycle time, we could probably do this here. Now, there's another element why this site grows. I mean, you've been here before, and this is like a mushroom in the fall, you know? It just keeps growing, yeah. Why is that? Because we also need the R&D to it.
Even with Deep UV, we're spending more R&D on Deep UV than ever before. It's still new product introduction, what we call NPI. The biggest number of functional people here are the R&D people on this side, and they work very closely together with the integration people, huh? And not with the people that actually make the modules. There's 82% or 85%, it depends on what type of machine is in the supply chain. Your question really relates in our mind to the supply chain. Where should we get our parts from, yeah? The integration with the reduction of cycle time and the necessary closeness of the R&D people to continuously integrate together with the operations people, new product introduction. Yes. Let's take a Deep UV KrF tool.
KrF, 20-25 years old technology. We put it on NXT. Very high productivity tool. It's a new product, yeah? It needs this closeness, this close vicinity of R&D and operations people to make that happen. Most of what we do is still NPI, new product introduction. From a system integration point of view, I don't see that quickly. I don't rule it out for the lithography tools, but more importantly, where do we source?
I think this is also why, you know, Wayne will actually start looking at this, as our strategic sourcing and procurement officer to say, "Okay, how do we manage that?" I think Frédéric said it, the output is not only cycle time reduction, it's also getting the parts on time at the right quality and the right volume, yeah? We need to look at that when we look at diversification, yeah? Is that? Is our supply chain able to follow us? This is the kind of work that we're doing today.
Maybe we could add, Peter, but for the other factories where we produce modules, in Wilton or in San Diego, in Taiwan, they're increasing at the same speed as we increase.
At the same speed.
Veldhoven in terms of capacity and infrastructure.
Yeah.
I think we're also expanding where we are doing our sourcing. We do a lot of the sourcing where we're doing manufacturing in the U.S. We are looking for local suppliers in the U.S. When we do manufacturing in Asia, we're looking for sourcing in Asia, more parts in Asia as well.
Yeah.
Amit, if you look at the unit numbers that we presented for 2030, even in the high case scenario, you would see that those unit numbers are still within the envelope of the 9,600 that we're building, you know.
Right
here today. The only one that would exceed the medium-term ambition would be on High-NA, right? Where we've talked about the medium-term ambition of 20, and you would have seen in the high market scenario, that's higher than that. There, you know, increased capacity would have to be built. The, you know, for all the rest, the unit numbers fit within the envelope that we're currently building here.
Chris, you want to answer the question on.
Yeah. I think on the second question, I will maybe answer it a bit from a general point of view and then give you maybe a specific example. I'd like to connect it to one of the previous questions. There was a question about, you know, what else the customer needs to get the roadmap going. Metrology inspection is absolutely critical to also advance in order for us to continue with litho. I think this is the reason why. I think it's more than 10 years ago, I guess, Martin, we started to talk about holistic lithography.
At that point, we already believed that without metrology, without the data metrology will create and the use of this data on the scanner, there was no way we could deliver the process performance the customer was asking. If we were trying to do that with a scanner only, this will end up making extremely expensive machine. I think we do think a lot about metrology and inspection. If I had to give you maybe one specific example, I would take the one of inspection. Inspection is needed to find defect that are related to the process customer are going to drive. Typically, you need to find defect that are 25% of the size of the pitch.
Optical inspection today can find defects of about 12-10 nm if you push it, which means that the limitation of this will be around 40-45 nm pitch. Today we are beyond that. This is why, as you know, we have acquired HMI a few years ago. We see a lot of, I will say, demand from our customers to use Ebeam. We have a good, what we call single Ebeam business today. Customers are using those tools to start to do defect inspection at very, very slow speed. Also metrology.
As you know, I think we have also announced that a few months ago, we started to work with our customer on multi-beam. We have our first five by five tool at customer. This is very difficult, so this is another major endeavor we have taken upon in ASML. But the reason we do it is because we believe that we have to bring the productivity to eBeam in order to be able, basically, to find the defect customer will look at when they reach 20 nm pitch, for example, which is the kind of pitch High-NA is going to enable. Not only will we look at it, but I think we invest also quite some time, some energy, I will say. I see the results of that.
If you look at the business, I think Roger talked about this is still about more than 20% growth year-on-year in ASML. That's also significant from a business point of view. Yeah.
I'd like to add one element. With everything you said is fine. The issue is that we're not just competing as a metrology company to another metrology company.
Yeah.
It's very important that you capture that. It's metrology with a purpose, with a prime purpose in getting the whole litho optimized, which we need next to good metrology, also the litho solutions we talked about a lot, but also the one not named is the computational part. The complexity in which we have to connect the dots with this increased metrology. First are the additional knobs on the steppers. We are exploring that. On DUV, for instance, this year, we're adding another innovation where we're able to modify the image, so we can just increase the potential yield on the edge. You have that knob, and now you have to have the measurement and the modeling to make sure you make it right.
It's a whole connection between the three, which we continue to believe is enabling for our customers to get a cost-effective patterning.
All right. Thank you. Roger, maybe this. We have a couple on here on CapEx. If you can, I'll try to cover it in one question. But first was a clarifying question, was EUR 1.5 billion in CapEx per year a reasonable forecast going forward? And then of the CapEx required for the capacity expansion that Pieter had in his material, is there a way you can put any color on how that breaks out in terms of DUV, EUV, and the significance there in terms of that investment?
Sure. The 1.5 billion, indeed, is the CapEx number that we're looking at for the next couple of years. Of course, it will be erratic, right? I mean, it's clear that one year it could be near 2 billion, the other year it could be below 1.5 billion. We think on average for the years to come, we believe EUR 1.5 billion in CapEx is probably the right number. That's 500 million more than we said last year. Last year, we were talking about a billion, and that's also the calculation that Peter made when he talked about a 2 billion extra. That's 4 years of 500 million incremental CapEx. That's the whole background of that.
In 2025, 2026, you get to the capacity level that we're envisaging. That's, I think, the background on that number. In terms of distribution over DUV versus EUV, I think the lion's share is EUV. Lion's share of the investment is EUV. The CapEx needed to get to the DUV capacity at 600 is fairly moderate in comparison to EUV.
All right. Very good. Then maybe one, just while we're on, Roger, on gross margin, will EUV be at the same level like non-EUV business or even above in 2030? What are the ASP inflation assumptions in your projections?
Yeah, I think the inflation question we already answered, right?
Yeah.
Inflation, we have to model in such a way that you know that we distribute it equally in the ecosystem, and ultimately it shouldn't touch us. Therefore, I think inflation should be a neutral factor in our model. In terms of
The margin on what?
Yeah, the margins.
The EUV margins.
Yeah, EUV margins. Yeah, I think we said at a certain point in time, we think that EUV margins and DUV margins are gonna cross. I think we said that we probably are going to achieve that around the 2024-2025 timeframe. I think that's where we said that would happen. I think it's fair to assume that would be sustained, that line, if you like, would probably be sustained until you know, throughout the second half of this decade. What that means is that, you know, if indeed you are going to see improvement of the gross margin in the second half of this decade, which is what our model would suggest is possible, I don't think that would come from DUV.
I think the lion's share from that should come from High-NA getting better gross margin, because obviously in the first couple of years, the gross margin there will be under a bit of pressure. EUV Low-NA obtaining better gross margins, and the service margin continuing to make the improvement that we've seen.
All right. Thanks. Let's go over here somewhere.
Hi, good afternoon. Lee Simpson, Morgan Stanley. I was struggling a little bit with the rationale for where the value growth was really coming from a sort of top-down perspective. It was interesting you put up the 3% of sales as a litho intensity metric, but then the delta seemed to go to 5%. It wasn't until Martin mentioned system innovation and Peter had said energy efficiency. It almost feels as though for you guys, the delta in value is coming from delivering things into the systems integration or systems innovation field. With doing so, can you help us understand who you're taking that value from? I mean, whose lunch are you eating there?
Well, at least historically, I think you can look at that, right? Historically, I think we demonstrated that with the introduction of EUV, we were eating someone else's lunch, and I think it's not too difficult to see. I mean, one reason why litho intensity went up there is because we reduced the process complexity for the customer, because the customer could have, you know, multi-patterning, which obviously has a number of iterations in terms of deposition, etch, and what have you, have that replaced by one step. And I think that's the contribution that EUV has given and continues to give. I think that's the contribution that High-NA is going to give. It's the avoidance of process complexity that I think has increased that percentage for us.
You know, my submission would be that I think also for the second half of this decade, that potential is still there.
Maybe just a quick one on margins. Installed Base Management puzzles me a little bit that it doesn't seem to demonstrably grow. You know, there is a service element to this. There's a large installed base, and you would have thought at scale, the margin structures would improve. Is there anything you can talk about how that could improve? Could we get to 60% margin in Installed Base Management by 2025?
Well, I think the installed base margin has grown substantially in the past couple of years. I mean, I've been with the company for four and a half years now, and I recall that in the first year, we were looking at a gross margin of around 30% on the Installed Base business. I think if you look at our semi-annual report today, you would see mid-40s% by now. I think there is a significant improvement.
One of the key engines to drive that, of course, it's the upgrade business. I think there on the upgrade business, both on the software side and also, you know, some hardware upgrades. I think that's driven the percentage up, and there is reason to believe that, you know, that potential will continue to be there. Also on the service side, I think the gross margin has gone up. Will it be north of the corporate gross margin?
That's not what I would expect, because there is always a you know, break-fix type element in the service business where I think it's hard to argue that you know, that should be at the corporate gross margin. I think having it north of the corporate gross margin, I think, is a bit of a pipe dream. Driving your overall installed base margin to approximately that level of the corporate gross margin, I think is a good ambition to have. That means that in comparison to the gross margin that we showed in the first half, you know, I think there is still some ambition left there.
At the top there.
Thank you. Thanks for taking the question. CJ Muse with Evercore ISI. A couple gross margin questions. You talked briefly about it, but I would love to hit a little more in detail. High-NA margins into 2024, 2025, evolutionary product, but obviously the optics a lot more challenging. How should we think about gross margins entering the model in 2024, 2025? Then the aspirational 60% in 2030, you talked about bringing Low-NA and High-NA margins to comparable to DUV. Are there other drivers that you need to get to that 60%? I guess could you speak to kind of the rank order of importance, whether it's you know, lower cost of ownership, higher throughput, you know, reduced depreciation, whatever is important? Thanks.
First conceptual answer on High-NA in the 2024-2025 timeframe, and then I'll be a bit more precise as to what you could expect. Conceptually, I think it's important to recognize that we do not expect the type of development in gross margins that we had for Low-NA. The reason there is that while High-NA is clearly not a walk in the park, and I think we're clearly gonna see that in the video as well. I think it's also clear that a lot of the technology of High-NA is already leveraging technology in Low-NA. I mean, it's. There is not a different light source.
You will see quite some commonality between parts between Low-NA and High-NA. From that vantage point, I think the maturity of the product, I think, is probably at a higher level than what you saw at Low-NA. I think that's an important point to recognize. That said, if you look at the sheer numbers that we're looking at in the 2024/2025 timeframe, and if you recognize the preparation work that needs to be done in the factory in order to really prepare for, you know, for a significant rollout of High-NA going forward, you have to hire a whole bunch of people, you have to have the tools in place, et cetera, et cetera. Then distributing that over a fairly short base. Of course, don't hold your breath for the gross margin in 2024/2025.
As a matter of fact, you know, you could actually see that that could be detrimental in that timeframe to the gross margin, not because of issues in the maturity, but just because you have a lot of preparation work to do because you're driving that far more aggressive roadmap in the second half of the decade. That's important to recognize. You already see a little bit of that, right? Once in a while in the quarterly calls, I do tell you there is about a 1.5% gross margin depression as a result of High-NA preparatory work. I would submit that that would only go up in the years to come until you really see higher volumes of systems being recognized in revenue, let's say, from 2026 onwards.
In terms of what needs to happen in order to get the gross margin to the 56%-60% bandwidth that I indicated as an opportunity in by 2030, I think I just gave the key elements, right? High-NA clearly is one, and simply getting that to the corporate gross margin level. As I mentioned, I do think EUV still provides us with a lot of value to the customer as a result of that ASP and gross margin potential. That should be a key driver there. For DUV, my expectation will be that we continue to have that at the current level.
As I mentioned, I think there is still some potential in the service gross margin. That's our. Those are the puts and takes, if you like, on gross margin, and that's the basis for the 56%-60%, CJ.
Very good. We have one last question here. François, if you can make it quick. Two minutes.
Quick one. Thank you very much. Martin, since you're here and since we know that you are. You believe in High-NA, that was very interesting. Can you talk about the readiness in more detail on what kind of things you need to improve? What are the bottlenecks in terms of cost that, you know, you have to improve and the different roadmaps? Just a bit more color on the readiness. If we look at, you know, High-NA and EUV, the products were introduced like five, 10 years before, you know, the actual mass production. If we think about the imec roadmap, it's like early 2030s. Should we expect, you know, like prepayment kind of or commitment from your customers, time would be great. The second question is, Peter, you...
We asked you before this question about the capacity significant increase and about the supply chain. Having suppliers like ZEISS is quite important. What about your partners like deposition, etching? You know, do you see any bottleneck possible in a way that the others can't follow or because they have the different lead times of visibility than you, but not in terms of new fabs, it's the same lead times, it might create an issue on the upside. Just. That's the two questions.
I think for our peers, I mean, if the industry grows, I mean, they're perfectly happy to grow with us, and I think generally their lead times are shorter. They also need to prepare the, you know, capacity, you know, out there. I think. Really the issue is if you want to build out capacity and you think about the supply chain, as you just mentioned. It is, do the suppliers have enough financial capability or capacity to actually build out the 9,620. I think generally, yeah. Yes. There will be some which have to really step up, like for instance, ZEISS, which we're going to do together, which we did in the past.
You know, sometimes these very expensive clean room facilities, as you saw in the you know, video, they just go beyond what is reasonable for a supplier, in this case, ZEISS, because they have the most expensive module in our system, and then we do this together. But I think on the ecosystem, I'm not concerned about the ecosystem. I think the capacity lead times there will be manageable, that will happen. On the other question was on, I think High-NA and on prepayments and on what are the. I'll leave it to Martin to say, okay, what are the technical challenges that we have overcome? You know, what is still out there?
I think on prepayments, I mean, you see it in High-NA. You know, we have High-NA, it's not shipping. It's shipping 2024. 2024, 2025. We have the first prepayments and the down payments there because it's a massive investment. You know, it's just their customers' contribution.
The oldest prepayment is from 2018.
Yeah. The oldest prepayment is 2018. Yeah.
Four years ago.
Yeah, yeah. I think it's which is something which I think is these technological developments are not only a matter of complexity, they're a matter of cost also. This is something that you share, and this is what we are seeing today. This is not new. It will be the same with the whole High-NA, I suppose.
A few comments. We'll not be able to cover the whole thing. If you look to our EUV development, it has been an extremely rocky road from the initial EUV to what we have today. When we start talking about High-NA, we thought we have to improve a number of things fundamentally. I mean, we start building EUV machines as long ago as 20 years ago, we offer demo tool and start making derivatives until we made it stick in 2018. For High-NA, we had we built a whole experience and created infrastructure and technology, which in fact prepared us to make a big step in optics design and optics manufacturing, and metrology to get us ready for the future.
What we try to do with High-NA is staying within the infrastructure we have, which is my cost comment. Now, what you have to understand, if you build an optics with an NA or if you go with the photolithography guys who are sitting with the balance sheet with an F-number. If you go to an NA of one in air, the optics need to be bigger than the Earth. This is the fight against High-NA. You need to get designs which are compact enough so you can manufacture them and put them into the existing infrastructure. This is our chance.
Today the fight is just with ourselves, but we're getting gradually into a situation where we're gonna have to discuss the customer discussions as well and say, "Well, this is what we could do, and how would it fit, and how would you plan this?" Something we did also with High-NA back-
Yeah.
Around, if you remember, the co-investment program in 2012 was a major theme. We're sitting there with Intel, like people like Bill Holt and Mark Bohr discussing what it will be in 10 years' time and what to do and how much, what's the affordability. Those things are likely to come in the coming years.
I think our confidence level with High-NA from a technology point of view has gone up. Why? Because to Martin's point, we've learned so much in the development of High-NA optics that actually gives us more confidence that we can also, from a technical point of view, create optical systems that can actually, you know, touch on the NA target that we have in mind, which basically is the result of the discussion we have with customers. What do customers need to make this economically feasible? This is why this collaboration is so critically important, and that's why it takes sometimes a decade, yeah, before you get things done. You know, when we talk about High-NA now will be introduced by the end of the decade, beginning of the next decade.
It's quite normal, that length.
All right. We're up against time, so I want to thank first off all the ASML panel members here, the management team. Thank you for your questions, both from the webcast and the audience here. I'd like to formally close the Q&A session and today's investor event. If you're unable to get the question answered, we can follow up through the investor relations team. Now, on behalf of ASML, we wanna thank you all for joining us today.
Thank you.