Christophe Fouquet (CF): EUV (extreme ultraviolet) has been a very important machine for AI. I would say without EUV, you don’t have a three-nanometre node, you don’t have a two-nanometre node, and therefore you don’t have AI.
Lawrence Burns (LB): A lot of academics would widely see EUV machines as being the world’s most complicated machines. They have more than 700,000 components. These are not easy machines to design or build.
CF: The transistor density has to double every two years. And if you look at what has been done in the last 40 years, if you look at what is being done today with AI, not only is this still happening, but this is in fact accelerating.
LB: Whatever happens at the application or model layer, whether it’s OpenAI, Anthropic or Google, ASML wins. They’re still going to have to require lithography machines to build the chips. And I think that’s the core attraction, that it’s effectively an agnostic royalty on AI demand, and more broadly, compute demand.
Claire Shaw (CS): Hello, and welcome to Invest in Progress, the Scottish Mortgage podcast. I’m Claire Shaw, portfolio director. In this podcast, we take you behind the scenes to hear conversations between our investment managers and leaders of the world’s most exceptional growth companies. As a UK investment trust, we can only market Scottish Mortgage to certain audiences and geographies. So check out the podcast description to ensure this episode is suitable for you. And as with any investment, your capital is at risk.
Sixty thousand times a second, a laser inside ASML’s newest machines strikes a pollen-sized droplet of tin with three successive pulses, producing a kind of light normally only found in outer space. That light is how computer chips will keep getting more powerful and energy efficient for years to come. ASML builds the machines that semiconductor manufacturers use to pattern billions of transistors onto each chip, the tiny on and off switches that make computing possible. AI datacentres, the smartphone in your pocket, cars on our streets, they all depend on this Dutch company’s engineering.
Today’s guest, Christophe Fouquet, took charge of ASML two years ago. And under his leadership, it’s become the most valuable company in European history and has begun shipping its next-generation extreme ultraviolet, or EUV, system, a tool that weighs as much as a blue whale and took a decade to create. So let’s hear from investment manager Lawrence Burns and Christophe Fouquet.
LB: Hi, Christophe. Thank you so much for joining us from your headquarters in Veldhoven. I have fond memories of the dinner we had near your headquarters just before you actually became CEO, also of your predecessor, Peter. That seems a long time ago, and a lot has happened since then. Perhaps we could start with the same opening question we ask all our guests, which is, what does ASML do, and what problem is the company trying to solve?
CF: Yes, so ASML is doing what we call lithography. And when I say that, I think that doesn’t help you at all. So I would try to start from what most people know. Everyone, of course, has heard about AI nowadays, right? It’s becoming very, very important. I understand that AI is making use of very powerful semiconductor chips. And those chips are very, very powerful because the number of very small transistors on them is what is creating basically the power of compute either for logic chips or memory chips. And if you want to be able to put a lot of transistors basically on the chips, someone needs to have a machine that will print the image of the transistor in very high quantities on those chips. And this has been the job of ASML for many, many years, more than 40 years now. Our mission has to be basically to allow our customers to print very, very small features in order to create the most powerful chips.
You’ve, for sure, heard about Moore’s law. Moore’s law says that every two years, transistor density has to double in order to continue to create basically a roadmap for this industry. And the ability to print smaller and smaller transistors over time has been a key component of this model. So that’s a bit of the job of ASML. And of course, as AI becomes so important, you know, our role is also extremely critical nowadays.
LB: Can you tell us a bit about ASML’s origins?
CF: This goes back more than 40 years ago and ASML is a company that was created out of Philips. I think you’re very much aware that Philips was an extremely important company in technology in the Netherlands, but I would say far beyond the Netherlands. And in fact, here in Eindhoven, Philips has been the source of many spin-offs. ASML has been one of them. I think you know ASMI, our colleagues here in the Netherlands. So that’s where we come from, basically.
LB: And can you give us a feel for the scale of the company and over different time periods, how it’s evolved on its journey to where it is today?
CF: We started in a very small, I would say, shack somewhere, you know, in Eindhoven with a few people. And for many, many years, I mean, the company was really the underdog in this industry, trying to capture market share basically against some of our Japanese competition. And this lasted for, I would say, a good 20 years, all the way to the early 2000s, maybe even 2004 or 2005, where even many times the company got almost bankrupt because money was very, very tight. I think the first really big success of ASML was with what we call TWINSCAN. So the idea to have a lithography system with two stages, one that we could use to expose, one we could use to measure the wafer. And this really created a major step in the performance of lithography machines. So this was a very first big step, but this was early 2000.
The second big one was the immersion technology. So the crazy idea to put water between a lens and a wafer. Between 2005 and 2010, we got a lot of traction on this technology because this was the most advanced technology. And by the way, we’re pretty much still the only company providing this type of system today.
And of course, later around 2018 or 2020, the success of EUV, I would say, has really finished transforming the company. So the size of the company pretty much doubled between before and after EUV, because suddenly we were not only the provider of a DUV lithography system, but also EUV, and both required equal attention. So, I would say, EUV brought us to pretty much double the size of the company, both in people and in footprint.
LB: You became chief executive two years ago, after 16 years at ASML. One of your previous roles involved leading a core technical department at ASML. And that must have been no small feat, given that ASML is widely seen as making the world’s most complex and advanced machines of 100,000 parts. And you once told us that you actually slept at TSMC’s chip-making facilities in its fabs in Taiwan. You also have told us about, during the pandemic, the efforts, the incredible efforts, you went to to go and visit and meet with them face-to-face in Taiwan under those conditions. Could you take us through your relationship with customers? Because it’s quite clear that the relationship goes far beyond just delivering the equipment itself. How closely do you work together?
CF: Yeah, I think it’s a great question, and I think that you bring back memories to me on some of those examples. But I think I picked this example of, you know, it’s a long time ago, of course, I used to be an apps engineer, and we had to fix the system at TSMC. And indeed, in order to fix the system, we had to be there and do whatever it took. And that included spending the night in the fab. But the reason we did that is because this industry is really relentless. You know, everyone wants to deliver. And I think this is absolutely unique. And in order to do that, we all have to innovate together. And I think that together is very, very important. You already said it, we develop a very complex machine, a very expensive machine, and therefore the relationship with our customers starts even before we design the next tool.
We have this long-term relationship where we try together to always dream up what could happen in the next five, 10, 15 years, because this is how far we need to be able to look to support the growth of this industry. So we have, I would say, a very, very intense relationship where we are strategically looking at the future together, and we stay very much aligned on that. And what’s also interesting is we have a similar relationship with our suppliers, and I think that our customers nowadays have a similar relationship with their own customers. So you have a very, very tight, what I call, you know, chain of trust and chain of knowledge and chain of support, basically, between, you know, the different players in the industry. And if you take anyone out, then, you know, you take the risk that the whole industry stops. So this is a bit the level of, you know, connection dependency we have, but that’s also the reason why we have been able for so many years to create so much innovation and keep basically the industry going so strong.
LB: It’s an incredible shared human endeavour of unprecedented scale and complexity that all comes together across the globe. I want to come back to ASML’s technology later, but if we could take a step back. For four decades, your company has played a key role in keeping Moore’s law going, as you referred to the law just earlier. There are some that say Moore’s law is now dead. Your website states it’s alive and well. Could you help us untangle and understand these different perspectives?
CF: Well, I think you have many different versions of Moore’s law, and we all tend to pick the one we like the most. And depending on the version, I think you will find out that the law is either dead or very much alive. So I think there are two main versions of it. The first one is the one that says that the cost of a chip has to go down by a factor of two every two years. Well, that one is dead for many, many years because the cost of technology has been increasing, already, for a while. The version I picked before is the version that said that the transistor density has to double every two years. And if you look at what has been done in the last 40 years, if you look at what is being done today with AI, not only is this still happening, but this is in fact accelerating. So AI has taken us off this exponential growth of the transistor density and basically is requiring even more transistors, you know, added every couple of years.
So if you look at the most advanced chips or products from Jensen, you don’t look at a factor of two every two years, but a factor of 16 every two years. So there you could say, well, Moore’s law is not dead. In fact, you could say NVIDIA has put it on steroids. And I think that’s what we see happening.
Now, the last thing to say about Moore’s law, which also links a bit to lithography, you know, initially when Moore’s law started, I think lithography was by far the engine to get the density, because most of the density was achieved by shrink. But over time, of course, there’s a need for a lot more technology in order to get the transistor density. But if you go back to the way Dr Moore himself talked about his law many, many years ago, he had already anticipated that. And he said Moore’s law will be about shrink, and it’s also going to be about integration. So from the very beginning, there was the idea that, you know, there would be a need to both drive more transistors per unit of area, which is basically shrinking, and also more integration.
And I think today Moore’s law is really being supported by the combination of those two, and the two would be very important moving forward.
LB: I’m pretty sure most consumers don’t feel the speed of technology and computing power is slowing down anytime soon, thanks to the efforts of ASML and NVIDIA. We actually have holdings in several of the leading artificial intelligence providers, Anthropic, xAI, now via SpaceX, Meta. And so it’s quite clear from our vantage point that the appetite for compute to train models and provide inference is currently insatiable. You’ve cited this as one reason that you’ve recently increased your sales forecast to between €36bn and €40bn this year. But I’d be interested to understand how you think about the long-term structural demand that you’re seeing for compute and how you plan for that in the context of what has historically at least been a cyclical industry.
CF: Yeah, I think it’s a great question and I think it’s a question that we have certainly not fully answered. I think a lot of my colleagues in the industry try to really understand what AI means in the long-term demand of chips. And you summarised it yourself. I think we are getting all very bullish on what AI will drive in terms of chips demand, both logic and memory. And therefore, you know, we’re looking most probably as a need for more capacity. How much capacity exactly? I think that’s still a question we all try to answer. The jury is still out because what we see happening with AI today is major investments in infrastructure. Investment that, by the way, will have to continue over time because the infrastructure will have to evolve.
You know, we mentioned NVIDIA before. They’re already working on the next two or three generations of products. So this means basically that whatever infrastructure we build today, we have a chance to upgrade it, you know, in two, three years from now and then again in five and six, etc. The other thing we still don’t know is also the demand of chips coming from what I would call the end-user use of AI.
So once the models have done the job, once the models allow really to use AI on more products, either, you know, a consumer product or industrial product, what does it mean for the demand of chips? So that I would say the jury’s still out. What you see is all of us are anticipating more need for chips. And therefore, we are all, I would say, ramping up our capacity.That’s true for us, that’s true for our peers, that’s true for our customers, of course. And I think that in the next few months, we’ll try to understand where this goes in the long term.
When it comes to cyclicality, I think this, well, most probably it’s going to stay because, you know, we are an industry where people like to move very fast. So when they see a huge opportunity, they all go in. And there is always a moment where they say, well, you know, maybe we need to pause a bit, and then everyone gets out. So I think this has been the reason why we have seen cycles across the history of this industry. I still believe there will be cycles. You know, it could be because of the industry itself. It could be because of all the excitement we have around us when it comes to geopolitics.
I think this is almost extraordinary that, you know, so far we are so insensitive to some of that, at least when it comes to this industry. But I think most probably we’ll still be looking at cycles.
LB: I want to focus on your own technology now, and I think it’d be useful to delve into the different types of lithography you provide to help the non-specialists in our audience make sense of the opportunities ahead. So very broadly speaking, you make two types of machines. One is using deep ultraviolet light, known as DUV, to create shape patterns on a wafer. The other is more advanced and more costly: extreme ultraviolet. In simple terms, can you help us understand the difference between those two key products ASML delivers?
CF: Yeah and maybe the best way to do that is to go back to what our customers are doing. So when they build up chips, advanced chips or non-advanced chips, they do that by putting on top of each other a lot of different layers with different functions. Then there will be a lot of interconnection in order to create basically the integrated circuit. So in order to create this stack, basically, which becomes a chip, customers are going to use several dozens, sometimes several hundreds of different lithography steps, meaning different exposures. And not all those exposures require the very tight resolution I described before. So not all of them are at nanometre level.
A few layers, the most critical ones, are going to be at nanometre level. And they will be using the most advanced lithography machine, low NA EUV today, high NA EUV tomorrow. But a lot of the other layers will be in fact using deep UV machines, so machines with lower resolution.
EUV in volume, in terms of units, is in fact the smallest number of machines, so we ship a lot more deep UV machines than we ship EUV machines. We tend to always talk about EUV because, of course, those are the most complex and therefore the enabling technology today for our customers. While deep UV, you could look at it more a bit like as the workhorse of lithography. And what’s interesting is the tool in ASML we designed 40 years ago, we shipped 40 years ago, most of them are still being used today by our customers because they still serve some of the applications that require this type of resolution.
LB: That’s extremely helpful. And within EUV, you’ve had the complete market to yourselves now for nearly a decade. How do you think about your ability to sustain that lead?
CF: Yes, I think, you know, there are two ways we look at it. So I go back first to our customers because our customers want to sustain their lead. And it starts there. So it means to sustain their lead, they ask us to continue to innovate heavily on EUV. And, you know, today we are, of course, having a tool that is working very well. But when it comes to R&D, we spend more money on EUV today than we did even 10 years ago when we were bringing the technology to life. So we continue to invest heavily to build on that technology and to continue to drive basically our customers’ roadmap. And of course, as you do that, you keep your lead, right?
Because, you know, people typically can start to dream, catching up on what you do if you stop moving. But if you continue to move and you continue to move very fast, potentially even move faster, then I think you keep the lead. And when it comes to EUV, the technology is very, very complex, as you know, and just catching up to the very first tool we even shipped to our customers now 10 years ago or more is already very difficult. And now, on top of that, there is a need to catch up on whatever improvement we have done in the last 10 years. So that’s a very difficult mission, of course, for anyone.
LB: And it took ASML 20 years of research and development to create your EUV light source. Many thought it would never be commercially viable. Canon and Nikon both abandoned their efforts. And looking back on our notes from around that time in 2013 at Scottish Mortgage, your first-generation machine’s wafer throughput was well below target. There was still uncertainty about the endeavour. What made ASML take such a bold and costly bet?
CF: It’s a very good question, and you have to go back in time. I was not in ASML back then, so I heard the story from Martin Van den Brink, who used to be our CTO and, as you know, has done so much for ASML. But he was, of course, the person to bring EUV as the technology of the future for this industry. And when he did that, there was a choice basically between three different technologies: e-beam lithography, nanoimprint and EUV. And all of them were difficult technologies because they had to be developed, you know, to be economically viable.
Each one of them had specific challenges, but EUV was the one technology that if the challenges were to be addressed, could scale. So the challenges with EUV back then were to design an optic that was capable basically to reflect enough light to bring enough light to a wafer and therefore print the feature you need, and in such a way that the productivity of the tool could be good enough. So the scalability of EUV is now being realised, which means that yes, we spent 10 years of very hard work to get the technology going, but now we have a technology that can be used for lithography for the next 10, 15 years.
And if you look at our roadmap, we are in fact projecting ourselves as far as 2035 and 2040 when I look at our own internal plan. So Martin did definitely the right choice more than 20 years ago.
LB: Some foresight and vision. You’re also launching a newer system, high NA EUV, high NA – referring to the numerical aperture – a measure of how much light the optical system can gather and focus. That’s also been an immense engineering undertaking. I’d love to give our listeners a sense of the scale of complexity involved, given that all of that has to happen in an area often a little bigger than a fingernail.
CF: So, typically, nowadays we talk about nanometre-size features. So you know a nanometre, to help you, is 20,000 times smaller than a human hair. So we’re looking at extremely small dimensions. And we need to be able to print those dimensions basically on the silicon wafer. So we have to develop very, very complex systems. First, with the imaging capability, the resolution down to nanometre, and this requires an extremely sophisticated piece of optics. You know, when it comes to optics, usually people point to, I don’t know, the Hubble telescope as one of the examples of very sophisticated optics. I can tell you that our optics is 1,000 times more sophisticated than that when it comes to EUV. And it also requires basically the ability to do things very, very fast, move very, very fast, also at nanometre accuracy. And to give you an idea of how fast we move.
For example, on our NA system, the critical stage, the stage that holds the reticle where the chip information is, has an acceleration of up to 32g. If you look at the fighter jet, you look at an acceleration of 4-5g. So if anyone, any human being was to sit on one of our reticle stages of our NA system, he will not survive, he will be killed by the acceleration. This is how high the acceleration is. And we have this acceleration at the same time, we have also a nanometre accuracy, meaning we are able to move, stop, start the stage at nanometre level. So that’s the kind of technology ASML has been developing over the years, basically.
LB: What do you think it is that gives you the cultural ability to make a pivot like that to a new way of doing something where there’s so much sunk capital in the old way?
CF: Yeah, I think it’s, you know, we talk about company, you cannot not talk about culture. And I think that you most probably noticed that in ASML, the belief that you have to deliver what your customers will need, and you have to anticipate that is absolutely critical. And over time, of course, that very strong belief has made our success. It came with some risk. I think that, you know, you mentioned before that there was a point in time in the industry where a lot of people thought EUV would not work. I still remember a day in 2014, I think it was SPIE, and there was this nice article saying “EUV is dead”. It was not about Moore’s law, it was about EUV. EUV was also dead. And this was just, you know, the sign of how much difficulty we had to get the technology to work.
At the same time, we were convinced that this was absolutely important for the industry. And therefore we kept going. I don’t think we kept going with having in mind, you know, the benefit that the company would get if this were successful. I think it was more almost a pure mission that we had to make the technology work. So, this was really driven by an engineer mindset that this was needed by the industry. And that if we kept, you know, looking at the problem from a different angle, put the best people on it, we could solve it. I think this is very, very strong in ASML.
And, you know, it’s another story I didn’t tell you, but the reason I joined ASML 18 years ago was because one day I visited Albany factory. You know, they have this very nice corridor in the factory where you can see all the machines. And all the machines look the same, you know, pretty much in the factory.
But when I made this visit, there was this very, very strange machine. It was extremely big, with a lot of pipes. It looks like, I don’t know, a spaceship or a nuclear fusion reactor, whatever you want to call it. And I asked the team, what was that? And they told me, well, this is the ASML EUV Alpha tool. And I said to myself, what kind of company is that that has the guts to do something like this? I didn’t know ASML very much back then. And then the day after, I was sending my resume just by seeing the type of technology the company was willing to make. And I think that’s still true today.
Of course, the company has evolved. You know, we have our customers relying on us. We also have to be very responsible in delivering quality, capacity, etc. But that mindset is still with us, and it’s with us with lithography. But I would say it’s also with us on anything we can maybe help our customers with moving forward. So, I think that’s really, really strong in the company.
LB: Thank you for sharing that. Every time your colleagues are kind enough to let us go around ASML, we’re always incredibly impressed going into the cleanrooms and seeing the sort of construction of EUV and just how impressive and what an incredible engineering undertaking it is. Mainland China was your biggest market last year, accounting for 33 percent of your revenues. And that was despite the Dutch government blocking you from selling EUV machines there at Washington’s request. In April, a bipartisan group of American lawmakers proposed a new law that would restrict you from selling some of your more advanced DUV machines, immersion DUV, as well to Chinese clients.
I was just curious, how do you wrestle with being simultaneously an incredibly important commercial supplier to your customers, but also having this role as one of the most important strategic geopolitical assets in the world? How do you manage to plan the business when those geopolitical fault lines are moving, and they’re moving outside of your control very often?
CF: Well, I have to tell you that we had to spend a bit more time on those questions in the last, you know, three, four years than we ever did before. And what’s happened in the last three, four years is a bit the realisation of, you know, what we discussed at the beginning of our talk, which, you know, semiconductors become extremely important. AI becomes extremely important. It is becoming strategic for almost any country in the world. And lithography, being a critical element of that, is also becoming strategic. So I think the whole world realised a few years ago that, you know, chips, AI, lithography was becoming the new oil or the new gold or whatever you call it. And I think everyone started to look at it this way. Now, what we see now is a lot of discussion and a lot of different opinions.
And I think you have the whole spectrum on what to do with export control, and that’s a discussion that is ongoing, really ongoing. And I think at this point in time, we have not yet reached a certain balance. We have not yet reached a place where people all agree about, you know: what is the technological gap? Maybe we want China to be behind, and how do we implement that?
So I think it’s still a bit messy. If you ask me, there’s a lot of discussion on that. There are a lot of different opinions. This is, of course, a challenge when you’re a business, because you know that business likes and thrives on stability. So, I think that’s also something we explain to our government.
And I think what we need to do over time is find that balance, because, of course, you have the impact on the long term, which can be what we can or cannot achieve. But you also have the impact on the longer term, which is, what does it mean when you block a country from having access to a certain technology? And what it means most of the time is that this country will double-up, triple-up on its effort in order to develop its own version of the technology. So, what I’m trying to say is that, you know, if you are too aggressive in restricting, you are extremely inviting in creating competition somewhere else.
And I think that debate about what is the right balance is a difficult one because a lot of people have a different opinion about that. That can be, of course, we talked about geopolitics, so that can be also very political. It’s not always, you know, engineering-driven. But I think that our wish will be that we see some time, some type of stability and balance to answer some of the, you know, the technical concerns, which mostly are relevant, but also some other concerns, such as what do we really create for the long term in terms of competitiveness for, you know, this part of the world.
LB: Thank you. And the other thing we mentioned earlier was culture and the importance of that in what ASML has been able to achieve. And I know it’s something you’ve been focused on. But ASML, for good reason, is experiencing a lot of change. Headcount has risen by nearly 80 percent since the start of the decade. You’ve broken ground on a massive new campus in Eindhoven that could accommodate a further 20,000 new employees. You’ve had a handover in management. You alluded to Martin Van den Brink and the important role that he’s had for so long within ASML. As CEO, how do you think about the challenge of sustaining ASML’s culture throughout so much change?
CF: Well, I think, you know, I always say, and I explain that to our whole team all the time, I don’t want to sustain ASML culture, I want to enrich it. Because in the world you describe, the recipe of the past will not be the solution of the future. And, you know, we talked about AI, we are in an industry that is moving extremely fast. We are in a world where geopolitics creates new constraints, new requirements, and the idea that we should not change and continue to be successful is a very dangerous one. So, like with any culture, we have to keep the good element of it.
But we also have to realise that some of it has to change and that we need to continue to develop products that will enable our customers. We need to continue to be leading, as you asked before. We have to stay on top of our game.
We need to continue to be very competitive against any potential newcomer. So a lot of things are new to our business, which require us to continue to build up a very, very strong culture. The one thing I would add to that is, you know, the employees we hire today have very different expectations than the ones we were hiring 20 years ago. I had a meeting yesterday with some of our young employees, and we were talking about culture. And I told them, well, I think you would hate the culture we had 18 years ago when I joined ASML. So I think the idea that you don’t change the culture is wrong. I think you have to enrich the culture. It’s a very important part of the company. But it has to be a living culture. It should not be something we set in stone.
I don’t like people who say, you know, it was better five, 10 years, 20 years ago. I think that’s completely wrong. And that never solved any problem. You know, going back to what it used to be to solve a problem, I think is a disaster.
LB: Now, that makes a lot of sense. The culture has to match the mission of today, not the mission of yesterday. Christophe, thank you so much for talking with us about ASML, about its history, about its incredible products, and the role it’s playing in driving technological progress. We always close with the same question. What does the world look like if ASML succeeds in its mission?
CF: Yeah, and you know, the answer, that’s been the beauty of what we do, the answer is we don’t know. And I think that’s the best possible answer because you create with what we do, with what the entire industry does. You create opportunity no one ever saw coming. And we have many examples of that in, you know, the last few decades. And today I can tell you we’re working on machines, machines that will enable new revolution, technological revolution – I think AI is the last one.
We never know what would come. So EUV has been a very important machine for AI. I would say without EUV, you don’t have a 3nm node, you don’t have a 2nm node, and therefore you don’t have AI. But I can assure you that when we were struggling with the power of our EUV source, we had no clue that that machine would help create that opportunity. So what happens in five, 10 years if we do our job right? I don’t know, but I’m pretty much sure it’s going to be great.
LB: Great. Well, thank you so much for your time and for joining us. I’d also like to say thank you for the huge amount of value that ASML has added for Scottish Mortgage shareholders and that you yourself have been a part of, that Martin has been a part of, and all of ASML, and also for the incredible role the company has had in continuing the technology revolution and giving us many of the devices and capabilities we use today and often take for granted. So it’s been a pleasure, and thank you so much.
CF: Thank you very much. We’ll continue to do our best for you.
CS: Lawrence, what a fascinating conversation, and it’s remarkable how ASML’s expertise at printing the tiniest of details onto silicon makes so much change possible. And I thought Christophe’s suggestion that Moore’s law had been put on steroids was one of the standouts. But as ever, we finish by hearing the investment case from your perspective. So let’s start with what convinced Scottish Mortgage to invest in ASML when it was unclear its EUV technology was going to be a success.
LB: I think that’s an important point. We didn’t know that EUV technology was going to work, and we had to accept that uncertainty when making the original case. I remember we went through and we tried to talk to people in the semiconductor industry, we tried to look up academic journals, but our ability to analyse that was pretty limited. This was working out if one of the, at the time, greatest human technical endeavours would be successful or not. And as investment managers, we have to be conscious of what our limitations are, and that was certainly one. I think what gave us confidence still was a combination of factors. The first was that ASML’s success in EUV was critical for Moore’s law continuing. And in some ways, that was critical for a lot of other holdings. So we wanted to be invested and close to ASML in part to understand its progress because it had such profound implications for the world and for our portfolio.
A second one, though I think it was the clearest signal of confidence, was in 2012 when their customers Intel, TSMC and Samsung did an investment scheme into ASML. They gave them €4bn. They committed to funding the R&D. And that was really the biggest players in the industry saying this is the only way to solve this bottleneck. And we have to cooperate, and we’re going to fund them, and we’re going to give whatever it takes to ensure this is achieved. And so, as I said, limitations of investment managers, but I think you could put a lot of weight on how much Intel, TSMC and Samsung knew about the odds of success here.
And if there was a final reason, it would just be asymmetry. If they were successful, there was going to be significant long-term upside that would come from this, and, therefore, we would be more than compensated for taking the risk. But I think if you take a step back from that original investment, it’s just that reminder that all of our truly great investments for shareholders have had an element of risk and they’ve had an element of uncertainty. Sometimes they don’t work out, sometimes they do. But a bit like ASML, when these special companies truly do work out, the returns can be fantastic.
CS: And so, Lawrence, just picking up on that there. You know, ASML has been described as Europe’s most indispensable company, the only one capable of making the machines needed to manufacture the most complex computer chips. And Christophe made the case in the podcast why that lead is durable. But I’m interested in your view, as an investor looking out over the next decade, what’s giving you the confidence that ASML can keep that position?
LB: I think it’s a number of factors. The first is that a lot of academics would widely see EUV machines as being the world’s most complicated machines. They have more than 700,000 components. These are not easy machines to design or build. And ASML spent the better part of a decade doing it, and they spent the better part of the next decade making them a lot better and improving them. So it’s not a static target that you’re going after. Different people in the industry have previously said that if you were to give an EUV machine to a competitor or even a nation state, it would take probably 10 to 15 years to reverse engineer it. And you can imagine that ASML would make some progress in the intervening period. To give you another example of that, the optics are done by a company that ASML itself has a stake in. So it’s also control over the supply chain that ASML has succeeded in.
The mirror in the latest generations, if you were to stretch it over so it was the size of a country, the largest imperfections would have to be materially smaller than a millimetre for it to be successful as a piece of kit. So that’s just to give the idea of the technical difficulties here.
Now, at the same time, are there efforts to try to replicate EUV technology? Of course there are. It’s a device, it’s a tool, it’s a machine of genuine geopolitical importance, and that’s why some people have referred to China’s efforts to develop it as being equivalent to the Manhattan Project. I wouldn’t want to indefinitely bet against Chinese innovation, but again, catching up with an existing technology versus being able to move at the same speed as ASML as they continue to improve is difficult.
And I think the other element that’s really important here is that if China did succeed with its own EUV machine, it’s probably likely to be banned for use in a lot of fabs, TSMC, Samsung and Intel, particularly by pressure from the US government, which you’ve seen on the semiconductor industry. So even if they did succeed, the impact wouldn’t actually be to the vast majority of their revenues, it would be to their China revenue base, which itself is already declining and people already see at risk. So even in the state where someone did copy it and was able to get an EUV machine, it’s unlikely to be the sort of existential threat that you might expect.
CS: So Lawrence, it all sounds very positive, but we wouldn’t be doing our job if we didn’t consider, you know, what can go wrong. And we touched on this, you know, during the episode that the chip industry has been highly cyclical. You know, there have been demand slumps, which have tipped ASML into losses at points in time. And Christophe himself even acknowledged that the jury is sort of still out on how enduring those AI-related sales will be, and that there’s probably further cycles to come. So how do you factor in those risks around things like the cyclicality, shall we say?
LB: I think the first way of factoring in is just being both honest and realistic that this is a cyclical industry. Is it going through an incredible period of structural demand? Absolutely. But that doesn’t mean it’s no longer a cyclical industry. And I think Christophe was still clear on that himself. I mean, it’s interesting if you go back to that first investment that you mentioned in ASML about 13, 14 years ago. One of the things I think that held us back investing in ASML earlier was that we kept trying to view it as a cyclical business, and we kept trying to time the cycle. And there are two problems with that.
The first problem is that it sort of misses the wood for the trees a bit, that there’s a huge structural long-term demand for compute. Is that going to come through in cycles? Yes. But on any long-term view, you could probably reasonably assume that the demand for compute is going to be a lot higher.
And the second element is, if you’re going to start treating it as a cyclical company and trading constantly on that basis, you introduce the risk of getting that cycle quite wrong. And so I think one of the liberating factors that has enabled our long and successful ownership of ASML was going, yes, this is cyclical, yes, we shouldn’t get carried away, try to take a bit of money out at the top of the cycle, try to add a bit more at the bottom, but let’s focus on the big long-term structural demand driver. And right now, we just happen to be in the biggest structural demand driver of all, driven by AI demand.
CS: So, let’s dive into the AI angle a bit more, because ASML is far from being the only artificial intelligence opportunity that we have in the portfolio. You know, Scottish Mortgage invests in AI model developers, other semiconductor businesses, you know, companies putting AI to use in ecommerce, you know, autonomous vehicles, to name a few. So how do you think ASML’s position in the kind of AI value chain compares to some of those other players, Lawrence?
LB: So we start by framing it as three layers. Applications. So an application might be an AI-native company that’s doing autonomous cars, or an AI company that’s using AI to do healthcare better, or one of many things that you can imagine. Second layer is the frontier models or just the model companies, so Anthropic, OpenAI, MiniMax and then the other layer is the supply chain. I think the advantage of investing in the supply chain is that it’s very difficult to know which applications are going to be successful. You’ve seen that the market goes, this is a beneficiary of AI, this is a victim of AI, very quickly on different companies. That gets a little bit more clear as you go to the model layer, because there are fewer players, particularly at the frontier, that are able to keep up. You’re probably down to three with Anthropic, OpenAI and Gemini, part of Google. But you don’t know if they’re all going to win. You don’t know if it’s a one-player market, a two-player, and whether each of them is individually successful.
If you go down to the supply chain layer, you happen to have these bottlenecks that have incredibly strong competitive positions, and that’s reflected in very high margins, very strong free cash flow. Whatever happens to the application or model layer, whether it’s OpenAI, Anthropic or Google, ASML wins. They’re still going to require lithography machines to build the chips. And I think that’s the core attraction, that it’s effectively an agnostic royalty on AI demand, and more broadly, compute demand. If you go very specifically into ASML, a lot of these companies have bottlenecks in semiconductors. ASML has a complete monopoly in EUV, and that does make it stand out.
CS: Lawrence, I love that phrase, a royalty on AI demand. That feels like a great place to leave it. Thank you, as always, for all your analysis. And, Obviously, many thanks to ASML’s Christophe Fouquet for making the time to speak to us.
A reminder, we’re releasing 10-minute cutdowns of each of this season’s interviews between the full episodes for those times when you just want the highlights. And we’ve explained any jargon that crops up in the show notes. If you haven’t already, please subscribe to know whenever new episodes become available. And you can learn more about Scottish Mortgage by visiting our website at scottishmortgage.com. You’ve been listening to Invest in Progress. Thank you very much for joining us.
