ASML, the Obscure Powerhouse at the Cutting Edge of Chip Technology

You say, there's one thing that we have yet to do, but I have a feeling like this is the Endless Semiconductor series, and as soon as we finished this episode, we're going to discover some other hidden component of the semi conductor supply chain and that's going to lead to

another episode. But yes, you're right, um, there is one sort of big elephant, big semiconductor thing in the room, and that is a company called a s m L. Not to be confused with a s m R, which I always seem to do a s mL but hopefully for like a certain kind of person, listening to an hour of people talking about chips is a type of for them SMR. So maybe killed two birds with one stone, but yes, s m L, you know, one of the things that we established in thinking about how these sort

of the chip ecosystem works. Maybe part of our characterization is we twan semiconductor the biggest contract fabrication company in the world. I think we sort of think of as like they're the final boss right in chims, Like in the end, they're like the central bank of chips. Their capacity kind of almost dictates chip capacity overall. There's some other companies that make chips, including Intel and Global Boundaries,

but t SMC is the big one. But TSMC has to buy equipment from others to you know not no one is completely self sufficient in this industry, and t SMC is a huge client or a huge purchaser of

equipment made by this company. It's a Dutch company a SML. Yeah, so you mentioned that it's Dutch, And this is the other thing I mean, in addition to not really understanding what this company does or the type of equipment it's actually making for semiconductor manufacturers like t SMC, the other thing I don't really get about it is why is

it a Dutch company? Because the one thing I know about it is it has its origins in the US UM I think in the you know, like nineteen eighties, it sort of came out of the collapse of a bunch of US lithography firms or something like that. And yet now it's a Dutch company that has this enormous role in the global supply chain. It's squarely like kind of crowning, like it's sort of like a kingmaker for semiconductor technology or expertise. And I don't know, I just

have so many questions already. I know we haven't even started. Yeah, I know, I have no questions too. We'll get started just a second. But you know, you mentioned the oddity of it big Dutch. There's another element here, and I I don't you know, I feel like reluctant to like talking like cliches or stereotypes, but I don't really think. No, it is true of like Northern Europe or Europe in general as being like this like cutting edge high tech

hotbed for anything. When I think about tech, I think about Silicon Valley, maybe more of the consumer and but also like you know, obviously a long history. It means Silicon Valley for a reason. And I think about various parts of East Asia, and when I think about the

engineering prowess in Europe. Hey, I don't think about tech in Europe that much, and when I do think about engineering prowess in Europe, it's typically I'm thinking more on these sort of like bigger industrial engineering, so a company like Siemens or companies that are really good at public public works or trains or whatever. And I don't think of Europe as being a hotbed of say, semiconductor innovation,

and there's probably countless like counter examples. I'm just sort of thinking, like it doesn't fit into my mental models of this stuff. So it is interesting that it's Dutch. Well also, just when you think about the European market, like you start thinking about the biggest companies there, and yeah, sure,

stuff like Semens, LVMH like luxury good makers. But mL is absolutely massive, and like just looking at the share price chart, it has had a huge, huge run up over the past year or so, I mean basically since the global pandemic, much like a lot of other semiconductor stocks, but I mean amazing run up, huge market cap, and yet it's sort of like this company that outside of the semiconductor sphere, it doesn't seem to get that much attention. Yeah, I mean it's A. It's like a it's like a

third over three billion dollar market cap. It is. It's it's one of the biggest companies in the world. Um, but not many people know about it, far from my household name. Okay, so we have a million questions. So we got to get right into this discussion. And we have the perfect guest to tell us a out this company.

We're gonna be speaking with Chris Miller. He's an assistant professor of international history at the Fletcher School at Toughs University, and he is the author of a forthcoming book that will be out next year entitled Chip War, The Struggle for the World's Most Critical Technology, And he can answer all of our questions about a SML. Chris, thank you so much for joining us. Thanks for having me. Chris,

what is lithography? You know? I think that this is one of these questions that's like the word gets It's probably come up on like every episode. And I pride myself on never being too embarrassed to like ask a question, but I think I actually was too embarrassed to ask this on some of the other episodes. I'm like, hm, phothography,

what phithography? So if you want to make a semi connected device, you take a slab of silicon, you cover it with chemicals called photo resists, which are chemicals that react with light, and then you shoot light raise or now extreme all tra violet light rays at the silicon wafer UH, and the shapes that you shoot at it will form the transistors. So that's that's the simplest version. Now today, if you buy a new iPhone, the most

events process around it will have ten billion transistors. So you've got to shoot extraordinarily narrow wavelengths of light through masks that create these shapes on the silicon wafer, and the masks need to be able to project all of these shapes onto the wafer. So making this possible at the scale of ten billion transistors per chip is is what a SML does. Wait, how many per chip? What do you say? Ten billion per chip? That's right, a new Apple processor in your iPhone will have ten billion

transistors per chip. Some chips that go into data centers will have more than that. But the scale of transistors that we produce at any given year is is more than the scale of all goods produced by all companies and all other industries and all of world history. It's a tremendous number. So could you maybe describe where a SML sits in the sort of ecosystem of the semiconductor industry. So I gather it doesn't seem to have much competition, but like who does it actually supply? And also who

does it not supply? Like are there people out there who try to do this on their own. In the early days of the chip industry, companies built lithography machines in house, so Texas Instruments would have had its own Lithography Machine Division IBM. But today the machines are so complex and expensive that there's just a couple of companies that make lithography machines in general, and just one company, a SML, as able to make a u V lithography machines,

which are the most advanced type. Anyone who operates a chip fab sility where chips are made has to buy lithography equipment, and so for the most cutting edge chips, you've got no choice but to buy from a s mL. This is fasting. So whether we're talking about UNTIL doing its own chips or TSMC or any one else. And we've talked with other people who talked to Stacy Raskin of Burnstein and about the nanometer wars and all of them. If you're doing cutting edge manufacturing, you are a customer

of a a s mL. That's right, that's right. For the most cutting edge lithography machines, a SML is the only supplier for slightly less cutting edge machinery. Uh Night kind of Japan is also a competitor of s mls. They have a duopoly for anything that's not the most cutting edge. So what is it about the technology that makes it, I guess so proprietary to a s mL, Like, how did they get into a position where they basically control it? And what is it that they've been able

to do that others haven't. So the the challenge with EUV lithography in particular and lithography in general is that you've got to uh manage a supply chain that is extraordinary complex. A SML has got around four thousand suppliers, and many of these suppliers are producing equipment that only

they can produce. So, just to give you a couple of examples them, the mirrors within s MLS lithography machines, the eu V machines are the flattest structure that humans have ever made, the flattest man made structure in the universe. And when you go through the list of materials and components that you need to produce an e V lithography machine, there are multiple parts of the system that are the most this or the most that, and managing that is

an extraordinary complex business. If you talk to people at a sm L, they'll say, our biggest engineering challenge is not actually engineering any particular part, engineering the supply chain, making sure that all of our suppliers are producing things so that they all fit together, they all work together, they arrive on time. And it's hard enough to do that with basic machinery, but when you're trying to manipulate uh, individual atoms, which is what SML is able to do,

it's even more complex. Tracy, I already loved this episode so much. I don't know, like how many things like I've learned already in five minutes. And then the fact that like it's it's also a supply Like Okay, obviously there's a chip supply chain, but then the idea that the most advanced technology within the chip is actually itself

a supply chain. I'm just like, I'm already obsessed with this, but where do they So Okay, you mentioned that for sort of like okay, for the very cutting edge, there's just a SML for slightly less cutting edge. Uh, Nikon, did you say, Nikon? That's right, Nikon. In Japan, they're also in the game. Do other players aspire to be cutting edge or is there some barrier that just basically makes it so that no one else is really trying to get be at that level in the which is

when investment in EUV began. Nikon made a choice not to try to commercialize UV technology. The first physics papers on the V actually came out of a Japanese university, so there's plenty of optics expertise in Japan that a SML was the only company that was willing to bet on EUV from the forward and capable of raising the funds and assembling the expertise. So right now, if someone wanted to replicate what U vs what a SML has done with UV, it would take them a decade and

billions and billions of dollars in investment. And because the suppliers that work with a SML have exclusivity agreements with a SML. SML has invested in some of his key suppliers. It's just basically impossible for anyone to break into this without replicating their entire separate supply chain. It would take a decade to do. So, maybe this is a good place to start talking about the history of the company

and where it actually came from. And I think that will help us like understand some of these dynamics how it built up a competitive edge versus it's you know, non existent or very few competitors. But my understanding is the sort of like sprung up out of US military technology of some sort. Can you start like at the beginning? How like? I guess this goes back to Joe's question, what is lithography? Why is the U. S Army interested in it? And how did it play into a s

ML's creation story. When the transistor was first invented in the late nineteen forties in Bell Labs in New Jersey, it was predominantly used in military devices for its its first commercialization, and there was a scientist in a US Army lab named J. Lathrop in the nineteen fifties who was trying to find out how to minim miniature eye transistors produced them smaller and smaller so they could be

put in smaller devices. One day, he and his his assistant realized that they could use photo resist chemicals, these chemicals that react with lights to create shapes on on the silicon and germanium that they were working with, and they turned their microscope that they were using the lab upside down. So normally a microscope lets you see something small and it expands the image for your eye. They

did the opposite. They had a shape that was large and were able to project that in a smaller version by using an upside down microscope, And with that they filed the first patent or photo lithography and coined the phrase in the nineteen fifties. Over the next couple of decades, photo lithography was used both by chipmakers who were making their own machines and house and then eventually a couple of specialized photo lithography companies emerged in Connecticut and Massachusetts.

They were defense contractors primarily, but realized they could use their specialized optics things that they had honed in spy satellites and military equipment like that for the semiconductor industry, and so until the mid nineteen eighties, the center of the photo lithography industry was in New England, but those companies faced hard times in the nineteen eighties. They were poorly managed, and the nineteen eighties were a time when

the Japanese ship industry in general was rising. A Nikon as well as can and then the two camera companies began investing in photo lithography. For a time in the eighties and nineties, they were the dominant companies in the industry,

which the US was quite worried about. Worried about being too reliant on Japan at a time of commercial and also geopolitical tension, and so US chip makers began turning to a s mL, both to diversify their supplier base, but also because a SML was able to produce very high quality equipment as well. In the nineteen nineties in the mid as well, was the only company willing to take the gamble on EUV, and since that point it's become the dominant firm in the industry. We just have

extreme ultra violet. I don't know if we just want to make sure we've established what UV stands for? Can I just ask so, can you explain again? What's the difference between extreme ultra violet and I guess non extreme ultra violet. So over the past couple of decades is we've tried to make ever smaller device is in ever smaller features on silicon wafers. We've begun to use different and smaller wavelengths of light, and so extreme multi violet

has a wavelength of thirteen point five nanometers. Is the smallest wavelength of light that we've been able to use in in mass productions. So if you rewind several decades ago, we were using larger wavelengths of light that we're uncapable of producing the small feature sizes on silicon wafers that we demand today. One of the reasons why I think the chip episodes, well, why we've done so many chip episodes, and why they're why they keep resonating. I think there's

a few things. I mean, one is there's the chip shortage, and it should be noted that the shortage is actually wore at the is not really at the advanced level. It's a lot of cheap chips, etcetera. But we're starting to the chip shortage that relates to automobiles, etcetera. Has sort of brought people a lot of awareness about lack of domestic US manufacturing capacity. I think another reason people care about chips is obviously just the general explosion of

chip demand. Even where there isn't an acute shortage, there's chips h and everything. And then I think the other thing that makes it sort of an interesting story right now is that, at least in the US and probably elsewhere around the world, there is a rethinking about the role of state capacity in and um state investment into

certain space. And of course, as we all know, the chip industry and as you just talked about, the chip industry overall really was sort of born out of defense, so like sort of the ultimate in UH state investing and government spending, and at various times throughout US history at least we seem to go in waves of how much the government wants to get in to protect the chip sector, to invest in the chip sector, to build

and bolster a homegrown chip sector. You mentioned the sort of stress and tension with the Japanese or reliance on Japanese companies in the eighties and nineties that seemed to produce a wave of um sort of defensive and use. Perhaps it could be characterized. Talk to us about how a SML fits into that in terms of you know, when we talked about to say, the history of t SMC, that was clearly in part it was a very like

public sort of private venture. There was the government backed it up under the condition that it could raise private foreign money as well. Talk to us about the role of like public money in the creation of a SML. So a sm L emerged first as a division of of Phillips, the Dutch electronic company UM and it was spun out in at a time when the Europeanship industry was relatively small as a player on the world stage.

It was the U S and Japan at the time that were the biggest players, and there were a variety of Dutch and European Union programs to support R and

D at s mL. But for a SML in particular, actually the most important government support was from the U S government because in in the nineteen nineties, when the investments and EUV were being made, Intel, which at the time was the the industry leader in hip making, decided to take a big bet on eu V being the next lithography technology and established a consortium of a number of private ship firms and a number of US national labs, uh the Lawrence Livermore for example, that would work together

to produce prototype UV machines. And so the technology that we use today an SML Systems really comes from this work with US national labs. It was largely funded by industry but using the scientists there. And at the time there was some interest in trying to turn the technology over to a US company to producing commercialized on the grounds that it came largely out of US national labs, but there was no US lithography firm at the time that was seen as a credible candidate to commercialize. If

the options were NICON or a SML. Given the tensions with Japan, a SML was seen as the least risky option, and also they had a long track record of producing quality machines. And so we've got this strange situation now where a lot of the core technology in this machinery that's assembled in the Netherlands actually comes out of California. And indeed SML has actually bought a number of companies over the course of the past couple of decades in

California as well. So there's a lot of US technology in a SML Systems partially funded by the US government. Could you imagine something like that happening today? Like, I just think the environment is so different, and the idea of like the U S government funding a technology and then deciding like, well, okay, I guess the best company to actually make this stuff is over in the Netherlands, so we'll just let them do it and give up like a key component of a highly competitive supply chain.

It just seems so so unlikely in the current environment. Yeah, it's it's an interesting question. On the one hand, you do hear a lot of conversation in Washington, d C. About joint R and D project with Allied countries, and in some ways this is a perfect example of this. I think. The other thing is that a SML is a Dutch company, But if you look at the components of their EUV machines, for example, they're sourcing from all

around Europe, around the US and really worldwide. So to describe them as as a Dutch company misses the fact that you can't produce an EUV system was for example, the light source, which is produced by an S and L subsidiary in San Diego, so that there are a Dutch company S, but they're really a global supply chain

that's focused on on the U S and Europe. So this is interesting because you mentioned that, Okay, at the time that the technology was sort of they decided a SML would be the most credible entity to commercialize this sort of US funded technology. There was this view that, Okay, it was better them than a Japanese player, in part because we already had anxiety about our reliance on Japanese chips at the time for other other chips, including DAM.

How much are the same dynamics essentially in play when people think about the geopolitics of chips. Obviously one of the things that you know, we talked about anxiety about how my we rely on Taiwan. There's perhaps some anxiety about the domestic homegrown chip industry. Although China seems to be a several years behind in terms of mainland chip technology. How much does it still sort of benefit everyone this idea that this crucial component player is not part of

either in US or Asia. That's an interesting question. I think certainly if you're a Chinese customer of SML, you're pleased that it's not a U S company, But the reality is that if if the US wanted to use actually poor controls to constrict SML sales to China, that wouldn't be very difficult to do. A SML already doesn't send it's eu V machines to China. In theory, that's because of Dutch restrictions. In reality, it's because of US

pressure on the Netherlands to impose these restrictions. And there's discussion um in in Washington and Japan elsewhere about whether there ought to be stricter limits um the type of lithography machines you can sell to China, and legally there's there's nothing that would really stop the US from posting

those restrictions unilaterally. Is the concerning that if those machines were shipped to China, that they would be able to UH that would accelerate China's semiconductor capabilities, or that literally having them in Chinese hands would then maybe allow them to be more easily sort of deconstructed and reverse engineered and that would be a big knowledge transfer. No, it's the former. I think if you just receive an SMaL machine,

you have no idea how to produce it. Okay, Okay, it's that the more advanced cocoric machines you have, the more advanced shipmaker you have, the stronger the Chinese ecosystem is. So how big of an impediment is not having access to a sml S eu V technology to Beijing's like overall semiconductor development drive, Like, is it such an essential piece of technology that it basically means they're on a

completely different footing to something like TSMC. That's right. For now, there's there's no viable of producing the most advanced chips with the smallest features without using UV. There are some some scientists who think there might theoretically be ways to get around it, but for the next decade there's there's just no choice but to use a SMLS machinery if you want to produce the smallest chips. So what are you know, let's talk a little bit more about a

s MLS constraints. Everyone this year is becoming aware of, like, you know, constraints, and there's only so much boundary capacity in the world at any given time. The entities that wanted to buy cheap chips that go into cars sort of got shut out because they canceled their order for a while. And now they're scrambling and it might be years before they could catch up again. So we know that that's constraints. How strained is a s mls own capacity to grow and where do they face Is it

just in the complexity of the supply chain? Is it in raw materials? Like what are their constraints? It's mostly in the supply chain complexity. So SML last year's shipped thirty one EUV machine sans, so we're talking getting one or two more machines out of the out of their production process is something that's hard to do because each

of their suppliers is similarly constrained. And the ability to ramp up manufacturing, you know, this isn't high volume manufacturing when you're producing thirty one machines a year, and because their supply chain has so many specialized parts solely for their machines, their suppliers are producing thirty one or so of the components needed each year, and so there's just no way to ramp How many three hundred billion dollar companies in the world make produce thirty one make thirty

one units a year, so we're talking like each one is like half a goon or something. There's thirty one of the of EUV machines. They will also sell some of the older equipment, but they sold units in total last year, so it's still a tiny number of units. It's still not very much. How much is it if you or I wanted to pull together and by a EUV machine? Like what are they? What are they retail for? Average average revenue per UV machine last year was around

a hundred forty million euros. Got it? Okay, So this is something that comes up a lot in our supply chain discussions. But like how does ordering actually work and is there preference given to certain customers over others? Like, you know, if one company wants to buy an EUV machine, I imagine there there are plenty of other companies that want to do the same thing. For a limited supply How does a SML actually make the decisions about who gets allocated what? Also, how long does it take? Like

what's the waiting time to actually get one of these things? Yeah, we don't really know the details as to how SML decides to allocate the number of potential customers for a hundred fifty million dollar machine. Is is limited? It's a half dozen potential customers in the world, you be realistically

looking to buy one. But if you look at the main customers, it's t SMC, which has around half of operating the UV machines in its own fabs, Samsung, Intel, a handful of others, and there's almost certainly a premium that t SMC is paid for getting so many machines valuable. If if if a new company came online and wanted to buy machines, they'd face a weight of at least a couple of years because capacity has been purchased and advanced.

Intel has said it's going to be the first customer of the next generation EUV machine, which would be online around Presumably it's paid something for the right to get the first generation, but we don't know any of the details. Yeah, speaking of Intel, and I want to back up to something we talked about earlier. Why was this never part

of Intel's own ambitions? Because you know, over the years, I guess the degree to which Intel has wanted to be an i P first company or a manufacturing company at Waxes and wind, so at one point it was a manufacturing powerhouse, then has sort of scaled. Back then it was more of an I P company, and that's sally the anxiety these days now they seem to want to get back into being uh manufacturing, and the new CEO has made a point of like, we're not going

to give up on being a manufacturing powerhouse. Why was lithography or advanced lithography never part of the Intel strategy? Well, when INTIL was founded, it was. It was founded at a time where you could already buy lithography equipment on the open market, so they always decided they were going to produce ships, but by lithography machines from suppliers. Over their fifty years, they've been one of the biggest buyers of lithography equipment in the industry, and the development of

e V actually wouldn't have happened without Intel. When Andy Grove was still the Intel CEO in the early nine nineties, he made the first big bet on the development of EV lithography, putting up two or million dollars in nearly nines to begin to develop this on the grounds not that Intel was ever going to produce lithography equipment, but that it would eventually need EUV to produce the most advanced ships. And even as as recently as when a SML needed to raise more capital to to fund its

development of UV. It went to Intel, t SMC, and Samsung, and Intel was the biggest investor in a SML at the time, putting in a several billion dollars to help fund SML's development. So until quite recently, it seemed like Intel would be the biggest user of e V lithography machines. It's only in the past couple of years that Intel decided, in what looks to be a mistake in hindsight, that EV wouldn't be ready by around now, where t SMC maybe opposite that that EUV would be ready for high

value manufacturing. And t SMC was proved right, which is why it's done so well the past couple of years, and Intel has proven wrong, which I think most observers can explain some of the manufacturing problems it's had in recent years by not using EUV and trying to use older versions of lithography to produce its most advanced ships. Now Intel is changing it's it's it's plans, it's it says it's investing very heavily in UV, but it's gonna take a couple of years for them to learn how

to actually use the UV and high volume manufacturing. Wait, so I have a slightly related question, although maybe it's sort of reversed, I guess, but like, given a SML competitive edge in producing a key technological component for semiconductors, could they ever have just gone into making semiconductors themselves? Like if they have a monopoly on this technology, no one can do it as well as them, Like why

not just start making the finished product yourself. To make a chip, you not only need lithography, which is one of the key steps, but there are other steps as well. You need to be able to deposit films of materials um with with atomic level precision. There's there are different companies Applied Materials, for example, in California that make that type of equipment. You need measurement equipment that can measure individual atomic level errors in your final chips to make

sure you understand what errors you have. That's a different set of companies that makes that equipment. UM. So there's a lot of different specialized equipment that you need to

make chips. The cost of a new fab that for example, t SMC will build more than half of that cost is the equipment that goes in it and a SML and as ptography machines are a critical part of that, but that's far from enough to make chips, and SML specialty is really solely in lithography UH and not at all in deposition or matching or the other types of

equipment that you need to actually make finished chips. This at this point is so fascinating to me, like to think about, like, okay, a s mL among the many

extraordinary things. They also lay claim to having the flattest service in the world, and presumably in order to get the flattest service of the world, there's some technology, as you sort of just mentioned, that has to be able to measure flatness and actually measure if the services not flat and it sort of speaks to like you know, we think about like in the US these days, and there is UH. There's a bill in d C that's designed to invest in UH, designed to bolster US capacity.

And again that's part of why you keep having these discussions, because there's this effort on the way. It's kind of bipartisan. The bill might pass, it might not pass, but there's

this sort of bipartisan interest to bolster domestic capacity. But I don't even know what that means sometimes because obviously, as you describe, the internationalization and complexity of the chip supply chain is so extreme, and as we've talked about with other guests, chips across borders a million times before they arrive in your xbox or your iPhone or whatever.

It is, Like, what does it even mean in your view just to think about like this question of like expanding domestic capacity in an industry that but that just is so extremely fragmented and international. Yeah, I think the first thing is you've got to be specific as to what type of capacity you wanted to spand expand domestically.

Certainly the US could expand production of chips domestically if we wanted to, but that wouldn't have any effect on the reality that there's no way to buy lithography equipment, for example, except from foreign suppliers, either Nicon or a SML. I think domestic production is a is a great thing to support UM, but thesis that we're going to have

a entirely domestically produced supply chain is a fantasy. The only reason that we're able to produce ships with such small features is because we're able to take advantage of expertise from companies in many different countries around the world, and there's no one in the industry who thinks there's any conceivable future even if you're to spend a trillion dollars over a decade where you'd get a domestically produced supply chain as anywhere near as efficient. Is what we've

got now, you know. I think the supply chain risk discussion is often takes place at a thirty foot level, when what you really need to look at is what of the specific components you're worried about, Are there specific suppliers you're worried about, and how can you mitigate those specific risks? But just talking about domestic versus foreign production is not nearly specific enough to have any sort of real meaning. So we kind of mentioned this in the intro.

But again, one of the themes that comes up repeatedly on these episodes is the idea of supply chains all the way down. So if there's a bottleneck in one thing, there's probably a bottleneck um and a component and even

smaller component that leads into that one thing. So if there's a bottleneck in lithography equipment that's impacting semiconductors, I guess my question is is there a bottleneck in something that goes into the lithography machines or the eu V machines that is preventing a s m L from expanding production. It certainly could be there's not enough public information about SML supply chain to know, and it's very plausible that a SML, despite being real experts at management supply chain,

doesn't always uh No. They report having around four thousands suppliers, and all of their suppliers who you speak to will say they ask lots of questions about their suppliers suppliers, But the reality is that there are you know, multiple orders of magnitude more suppliers of their suppliers and so tracing them all down the chain is is basically impossible. So what a SML tries to do is understand what

other biggest risks. They've even gone so far as to purchase some of their suppliers to gain more detailed control over managing those risks. But they simply can't know every ultimate component that goes into all of all their suppliers systems, and so that's that's where the supply chain management just becomes extraordinarily difficult. Now what they've been really good at I think better than their competitors over the past couple of decades is managing that so it hasn't generally caused

any sort of serious delays. And one of the things that they're known for with their customers is being able to deliver mostly on time when they promise a machine and managing this, which is something that no one else has been able to manage. I think. The other thing to note is that you know, when you've got this equipment that is is manipulating individual atoms or trying to trying to control the movement of light with extraordinary precision. It's it's one thing to have a machine that will

do this once or twice or sporadically. It's another thing to have a machine that will do this day and day out, operating twenty four hours a day. And that's that's the other thing that a SML has done very successfully over the past couple of years. It was clear as early as the nine nineties that it was possible to make a chip with evy lithography. What's been difficult is making millions of ships with the euvy lithography and

doing in a cost effective way. And that's what a smlls really stood out is that their machines are rarely broken, always functioning um They need less um, less tuning, less cleaning than their competitors. That's what a sml has done quite well. So it's not simply managing the physics, which is very hard, but it's also making sure that you've got this extraordinary precise physics that's always operating in these act way you expected to operate. Yeah, I'm thinking back

to one of our discussions with HBS professor. Really, she and I don't remember the math exactly, but if like chip making is like a seven thousand step process, then even you know, ninety nine point nine nine percent execution in each step is insufficient in many cases because by the time you're down to the seventh seven thousand, you've

like basically lost all your chips. I don't remember the exact math, but it is interesting to think about like building up that comp that competent now just in can the machine make the chip, but can make it over and over and over again without without many errors. If you look at a s mls revenue statements, what you'll find is they've got a growing share of their revenue coming from services which is servicing the machines that they operate.

They've got staff in t SMCS, facilities in in Samsung, etcetera, making sure that not only the machines are operating, but they're operating exactly according to plan, they're not breaking down. The other thing that SML is is doing more of

is managing the software for their machines. And the way that lithography works at at the scale that we're talking about is that if you want to print a certain pictures, so you want to print an AX, you don't reflect the shape of the ax on your silicon because the way that light reflects if you print and actually get something different, so you actually learn over time all of the unexpected errors and light refraction and errors in the way the chemicals react, and you print the error version

and then it will give you an act. And so there's extradinally complex software that now tries to understand how all these different effects work. And you can actually look at the images that smls producing to produce a straight line and it looks nothing like a straight line. And so that that software as well as something that SML has been focusing on. So one of the things that we've talked about. You know, it's like the sort of

the nanometer wars and the obvious. You know, people talking about Moore's law and whether it's t SMC or Intel or anyone else, so they're always or a MG. Maybe they're always bragging about like making smaller and smaller chips. And one of the things that we learned is that actually the chip companies all defined these measures a little

bit differently, so to some extent it's made up. But how much are the chip manufacturers themselves as they advertised, like Okay, we're gonna be able to make a seven NIM chip or maybe a five NIM chip or whatever. How much are there timelines dependent on a s m l's I guess I would say a smls own learning curve. And what are as a sort of monopoly provider, I don't want to say monopoly, but is the sole provider of the most cutting edge technology. What are the forces

that drive technological gains for a sm L itself. If you look at how a SML has begun to roll out at CV machines in into high volume manufacturing, which has mostly been at t SMC, the learning has happened collectively with a sm L and t SMC, so there's been lots of SMA personnel that spend tons of time in Taiwan and vice versa. So you really wouldn't have the rollout of e V over the past couple of years had you not have this collective effort between t

s MC and a SML. And that's that puts other companies at a disadvantage because t SMC knows better than anyone how a s MLS machines actually work in practice, and and the highlight manufacturing is really crucial to understanding how how these machines work, because you don't really know

until you've got them in production. And once you've got them in production, you've got thousands and thousands and thousands of iterations every single day where you can understand what the errors are, what the idiosyncrasies are of a given machine, and begin to correct for them. We talked about technological progress and that that's important, but in some ways the

real challenge here is actually manufacturing progress. Understanding what the what what the ADO syncrasies are at the manufacturing stage and then learning to correct for them. And that's that's what t SMC has done extraordinarly well ad and it's been hand in hand with with a SML. So I

have a sort of market oriented question. But you know, we think of semiconductors as this highly cyclical industry that usually moves in line with whatever is going on with GDP and economic growth, and that hasn't really been the case since the pandemic because we've had, you know, this big boom in demand for electronic goods and it's been a struggle to keep up. But I imagine for a company like a s mL, it has also traditionally been considered cyclical and its fortunes are sort of tied to

what's going on with the actual semiconductor manufacturers. But just looking at SML's most recent results, they're forecasting basically like a boom in revenue for the next decade, something they expect to last for ten years. Is there anything that could sort of knock that revenue cycle at this point in time or is this business like, is there such a steep moat around the eu V technology that is just going to be impossible for anything to um to

hit it. There's there's definitely at mote around SMILS technology, and they won't be overtaken into UV in a decade. I should say I'm mixing my metaphors and steep and impossible to overcome mote. There's no real competition that a small faces when it comes to UV. That the question is what is going to be demand for UV machines, and that's a question ultimately of final demand for the

most advanced chips. A SMILS projections are are based on the assumption that we've reached a point where there's a secular increase in demand for chips, as we have more demand for data center capacity, um as we have the five G rollout and new devices that are taking advantage of five G networks. Their bat is that we're going to have more chips per g d P and therefore more demand for a SML. That's a bat that you know,

it's not clear what that's going to play out. What is clear is that anyone who's producing advanced ships will have no choice but to turn to sm L and so in some ways they're perfectly exposed to the fluctuations in the semiconductor industry. The more chips that are produced, the more machines you need. But the opposite is also true. You kind of hinted at this early on. But the idea that at uh seven nnimms in below at cutting edge,

maybe EUV in theory isn't the only technology. So it's like, okay, if if if EUV is the only technology that can perform the test, then there's no one who can attack as well. Are there other theoretical approaches for accomplishing the same thing besides the UV, I feel like you hinted at in the beginning, there are It's really not a

question of science but of of manufacturing efficiency. So if you if you take the previous generation of lithography machines, which rather than thirteen point five nanometer light, we're working with one. It became it was possible over time to produce ever smaller feature sizes on silicon wafers by using

a number of tricks. So, for example, you can shoot the light through water, and if you think back to high school physics, when light refracts differently through water, that same principle lets you shoot lithography machines through water and and carve more specific shapes. You can also use multiple steps of lithography to carve specific shapes that are more detailed.

The challenges just can you do this efficiently? Um So, every step of lithography you need adds to the time it takes to produce a way for ads to your costs. And so there's no doubt that if you wanted to produce an equivalent of one of Apple's new iPhone ships using older generation lithography, you could do it in a lab and do one of them. The question is can you do a million of them as skin? And that

seems pretty implausible right now. There's no really credible pathway of of how you could do that efficiently today, and especially when you project forward five or ten years, we're expecting to be making ever smaller transistors with more complex shapes on them, and it seems really implausible that you'll be able to do that using anything besides UV. Chris, is there anything any other sort of least key things

do you think we've missed? And I mean, I'm sure there's a million things, but other key ideas that I think we need to get across. I think if you're interested in US China dynamics, obviously it's a big. One of the key reasons why t SMC cut off huaweih In was because the US could restrict t SMCS access to machinery, of which lithography machinery was a It was

a key example. So when when the Chinese ship industry looks out and says we're we're gonna get the tools that we need, the impact of the US support controls on companies like a SML, even foreign companies that nevertheless use US technology in their systems, is a pretty fundamental roadblock that China faces, and that the big concern that SML has right now is that the US is going to expand its restrictions on what you can send to China in terms of lithography machinery, and China has been

a big growth market the past couple of years for older generational lithography machines, and so SML does face a risk that the US it expands these restrictions. What inspired you to write a book all about a s m L because it is like it's not something that ms up necessarily in daily conversation. UM, so I'm just wondering how it sort of came on your radar and what is it that piqued your interest? We like the two

of you. I spent the past couple of years realizing that semiconductors were vastly more important than the average person, including myself realized, and also far cooler. The technology needed to make them as extraordinary. The fact that we're able to manipulate individual atoms in some cases is extraordinary, and to do it at the scale of trillions and trillions and transistors, I thought was was was really just wild

in terms of what what what was possible? Uh? And it seemed to me that I took my iPhone for granted, I took my computer for granted. I took the cloud, which is just a bunch of silicon in big data centers in IO. I took all that for granted without thinking through how complex it was too actually make these tools work. And I think for for a long time we thought of the Internet is something out there. We thought of data processing is something that happens somewhere else.

But it's all actually very physical, all things being carved onto silicon by shooting light at them and depositing uh,

layers of atoms and using different chemicals. And the reality that our entire digital world is in fact existing on millions and millions of silicon wafers is something I don't think we think enough about and we're just having to come directon with that with the some innunctory shortage right now that you can't just imagine an increase in computing power and increase in memory, You've actually got to carve it onto silicon in billions and billions of tiny transistors.

You know, it's interesting, I mean, your assistant professor at the Fletchers School, which I associate with diplomacy and government, and that seems like another sort of like fascinating dynamic here, which is like maybe it's it feels again, or maybe it goes in cycles, but there's appreciation. And you sort of said it for your last point about yours trying to like that this particular industry is sort of inseparable from thinking about how governments relate to each other. That's right.

It's it's it's crucial for for military systems, for example, it's crucial for controlling computing power in the future. And it's been a prominent tool of of of geopolitics for the past three quarters of a century. And I think

we're seeing that more to the four today. But in fact, when you look at the history of lithography and of semiconductor is more generally you find that it's constantly been something that governments have thought about in in political terms as well as in economic terms, and constantly been an area of dispute between different governments as they tried to vy for a bigger chunk of the semiconductor ecosystem. Chris Miller,

thank you so much for coming on. That was the the a s m L episode We needed to do it, and you are the perfect guest for it, and I just learned a lot, So thank you so much for coming out on Oblock Bain's invitation. Thanks Chris, that was so interesting. Obviously, I love that conversation, and you know, I sortaty interrupted like seven minutes in because like this idea of like thinking of like a component as in

itself a supply chain story. Like the idea that really the breakthrough is how do you coordinate four thousand different suppliers of them of highly specific raw materials and machines into one thing that forms a cohesive whole. The idea that that is what the thing is is pretty fascinating to me. Here's the important question, which is, what, like what idea did you get out of that conversation For the next Semiconductor episode, so I'm sure there is one. What is the next one? Now we we we got

to the end. Now it would be like really interesting actually, okay, so like it all seriousness, like I would like to learn more about that process, like the actual like the coordination the heart. It's almost like you think of like a conductor of an orchestra. Is sort of like the mental model I use for a company that has to like have four thousand parts all coming together to form

thirty one units or units or whatever. It is like thinking about like how do you do that from like a management perspective even beyond this sort of tech perspective is like a super fascinating thing to explore, especially at its especially right now. I mean this is almost verging on state secrets. But we've got to get you know, we have to try to get the a s m L supply chain manager on all thoughts, so you know, a s m L hit us up. We're interested in how you're doing it. Um, and we have to keep

the semiconductor series going. Yeah, No, that was fascinating, and um, Chris was the Chris was the perfect guest for that one. Yeah, definitely, shall we leave it there, Let's leave it there alright. This has been another episode of the All Thoughts podcast. I'm Tracy Alloway. You can follow me on Twitter at Tracy Alloway. And I'm Joe Wisenthal. You can follow me on Twitter at the Stalwart. Follow our guest Chris Miller. He has a book coming out next year on the

chip industry. Assistant professor at the Fletcher School, he is at cr Miller One. Follow our producer Laura Carlson. She's at Laura M. Carlson. Follow the Bloomberg head of podcast, Francesca Levi at Francesca Today, and check out all of our podcasts at Bloomberg under the handle at podcasts. Thanks for listening.