Episode 271 - Semiconductors, Lithography, and Moore's Law with Adam Kane of ASML
Max interviews Adam Kane of ASML about the industrial processes which make hardware innovation possible. They cover extreme UV lithography, the law of accelerating returns, and implications for software engineers.
Links
Adam Kane: LinkedIn
ASML (Advanced Semiconductor Materials Lithography):
Website | Facebook | Instagram | LinkedIn | Twitter | YouTube | Glassdoor
News 12 Connecticut: Made in Connecticut: ASML in Wilton is one of the largest semiconductor companies in the world
Wikipedia: Moore’s Law
Kurzweil: The Law of Accelerating Returns
The Local Maximum: Email
Transcript
Max Sklar: You're listening to the Local Maximum episode 271.
Narration: Time to expand your perspective. Welcome to the Local Maximum. Now here's your host, Max Sklar.
Max Sklar: Welcome everyone, welcome! You have reached another Local Maximum.
I have a brief introduction today before we get to our guest, which is Adam Kane from ASML, a topic that is breaking new ground today at the Local Maximum. I am speaking to you today, for the last time, from the Local Maximum studio here in Salem, New Hampshire.
By the time you hear this, I will be back in Connecticut. In 2020 or 2021, New York City, to me, went from being the best possible place to live to the worst possible place to live. But now I need to get back, at least for now, closer to where my network is both personal and professional. That said, I don't think I'm finished with New Hampshire and I expect to be back here quite frequently to visit, and in the long term, who knows. I know a lot of people who I know here in New Hampshire are going to be disappointed.
But speaking of Connecticut, this gives me a chance to get closer to where I grew up, back to my roots even. It's fitting that we do this interview today, which is about a fascinating and a very important company right there in Wilton, Connecticut. Their work on hardware on the semiconductor space has global implications for the world of software. In fact, you'll see at the end of this interview, I tried to wrap my head around all these implications and I just started doing it. I just was, for the lack of a better term, scratching the surface and I feel strongly that we need to learn more.
I guess the question that I'm asking is, why is it that for the duration of our whole lives, not only that but all of modern history, we have devices that seem to improve year over year? From the actual hardware of the device to the software that's written on top of this device, this occurs. And yet historically, this state of affairs is rather unprecedented.
So someone is doing the work, the research work at every level, from the hardware level and the software level, and it starts at the atomic level of the semiconductor. And that's what we're here to learn about today.
You'll learn that far from being just some standard process to produce these critical components, you have a dynamic team of scientists and engineers that are constantly pushing the envelope. Including one of the recent paradigms, which is extreme UV lithography, I know it’s a big word but we'll get into it in a minute, to etch these things out on a scale that is very hard to imagine.
So just to get a taste of it today, we're going to talk to the director of software and electrical engineering at ASML. Adam Kane, you've reached the Local Maximum. Welcome to the show.
Adam: Thank you very much. I appreciate it.
Max: All right. So this is a very different kind of a topic that I usually. I'm always talking about technology but I actually never really, in the five years I've been doing the show, never really got into hardware, never really got into semiconductor ecosystem. So I want to hear more about that. So maybe you could tell us a little bit about what is your role at ASML and what is ASML?
Adam: Sure. So I am the software and Electrical Engineering Director here at the Wilton, Connecticut site for ASML. We have sites all over the world and Wilton is one of the biggest development site in the US but our main site is in the Veldhoven, Netherlands. Worldwide, I think we have about 37 or 40,000 employees. In Connecticut, we have about somewhere between 253,000.
What we do is, we have a saying we're the most important tech company you've never heard of. We actually make the equipment that all of the big semiconductor companies like Intel, which is probably the biggest household name that you have heard. Intel, TSMC, Samsung are the three biggest ones. Other ones include Toshiba, Sony, Hynix, companies like that. They use our technology to make microchips.
Pretty much everything you buy today has a microchip in it. There are very few things that don't have chips in it. Clothes are probably the most common thing that doesn't, but that will probably change in the future. And if a product has microchips in it, chances are they were made using our machines.
Max: Wow. It sounds like the companies you listed like Intel, Samsung, it sounds like just about everybody. I almost want to ask, maybe this is not appropriate to ask, who doesn't use your machines?
Adam: I don't think anyone to be honest with you. So we have a pretty big market share, that's why I say we're the most important company you've never heard of.
There are two main types of technology that drive semiconductors today. One is called deep UV, and one is called EUV, stands for extreme ultraviolet. That has to do with the wavelength of light that we use in order to create the microchips. Basically, for EUV, which is the more advanced technology, we are the only company in the world that has that technology in our systems. So anything that is on the higher end of the microchip scale is made by us. There is no other company in the world that has this technology right now.
Max: So what what is EUV? Why are we talking about ultraviolet light when it comes to making semiconductors? What's the connection there?
Adam: Yeah, that's a good question. The way our tools work is it's almost like a big printing press. If you go back to your lessons when you were a child. The way a printing press works, you've got the metal plates, and then you've got ink, and then you've got paper that presses against it. That's how you recreate patterns over and over and over again.
For semiconductor equipment, you replace those things. Instead of a steel metal mask, you have a piece of glass, and that's called a radical or a mask. Instead of paper, we etch onto silicon wafers. Then the most important part of it is, instead of ink, we use laser light to etch the pattern from the glass to the silicon wafer.
The difference between the DUV and EUV is DUV, I don't know, off the top of my head, remember how big it is. But basically, you can shoot DUV light through air. It's got to be very, very clean air, but it still will go through air and it'll go through glass. The big evolution in the industry, which we made that step, it started probably 20 years ago, but really to our customers closer to 10 or 15 years ago, was the EUV technology. This light wave is so small and thin, it does not go through air. Anything that it hits, it bounces off and you lose it. So you have to do all of these things in a vacuum.
Our newer systems are basically really big vacuum chambers which creates new and interesting technology challenges to overcome but it allows us to print thinner and thinner and thinner and thinner lines. The thinner the lines, the more you can fit on a chip, the less power the chips can consume, which makes the chips have today. Which is why they're so much more powerful, both from a battery usage standpoint and from how much you can fit on it and how fast they are than they are every year.
Max: Awesome. So how long has this EUV technology been in production?
Adam: That's a good question. Again I don't know the exact details but I know that It's been in research for decades. But to get to the point where we could create a affordable system for our customers to use in their production process, that really happened I want to think back in 2014, or 2015 was when the first, and maybe even a few years before that, was when the first EUV systems went to our first customers.
Since then, it's just been a series of industrialization and upgrades to make them more powerful, more throughput, which is how many wafers come out of them per hour, and more reliable for our customers.
Max: Awesome. Do you have a sense of, and I know the answer might just be No. It might just be like, hey, people demand that semiconductor has been made smaller, faster, more reliable, all of that. But do you have a sense of when you push an innovation like this, and when you make changes over the years, are there any applications that become possible with this technology? Like, products or anything like that weren't possible before?
Adam: That's a good question. That really goes to the end customer, probably our second or third-tier customer. But it's really, I would say your imagination is the limit here. What our systems do, and a lot of people, if you know anything about computers, or chips, or anything, you might have heard of something called Moore's law.
Moore's Law was basically a paper that was written by, I think his first name was Gordon Moore who worked for Intel back in the 60s. He said something to the effect of for every two years, the number of transistors on a chip in the same amount of size will double every two years. Something like that.
What that does is it allowed every sort of evolution of technology to take place from the beginning of time, in a way, up until today. You can imagine that, back in the 60s and 70s, everyone worked on mainframes because these big computers took up huge rooms. Then we kind of got the evolution into the desktop market where everyone had a personal computer in the 80s and 90s. And then when you transition to the 2000s, in the early 2010s, everyone started having mobile devices.
Now we're going through a transition where people don't have one computer and one phone. You've got tablets, you've got mobile devices, or phones and your headphones and everything. Now everything that you buy has a chip in it and it wants to have access to your Wi-Fi so that everything can be connected. This is what we call the sort of the internet of things, where every device is talking to every other device.
All of these things are only made possible because of Moore's law. Our company really drives that with the ability for our customers to pack more into every chip that they make every year. Not only that, but the power usage goes down. So we're living in a world where obviously ESG is very important, being green is very important. One of the very, very good side effects of what we do is that it takes less to do a lot more when it comes to power with each generation of these chips that our equipment generates.
Max: Awesome. So there's Moore's law and then there's also this broader law, like Law of Accelerating Returns of Information Technology. Ray Kurzweil talks about a lot where it's like even before Moore's law, you had punch cards where a new punch card technology would come out, and then it would be twice as powerful the next year.
My other question was what you guys do related to Moore's law. Just answered it, totally integral. You often hear people say that oh, Moore's law is ending, or I think Moore's law is ending in 10 years, or maybe there's a broader statement about Moore's law versus Law of Accelerating Returns. Do you have any thoughts on that? Do you guys have a more front-row seat? Because I feel like a lot of people who say this don't really know what they're talking.
Adam: It's an excellent question. I'll tell you, I remember when I was, I think, maybe I was in middle school, or early high school. I remember they would say these types of processors, the 286 or 386, that it was the end. It couldn't get any better. But what happens is, and we can get really technical, I won't get too technical because also I don't understand all of it.
Every few years we make technological advances in the art of the semiconductor equipment. So you hit a wall with what you know and then, for example, I'm gonna say probably 20 years ago, something called Emerging Technology got introduced. What that did was on the DUV line, we were able to figure out a way to put a drop of water in between the laser and the wafer. That gave us more resolution and more throughput and so on and so forth. So this immersion step brought us into the next realm.
Then we thought we were going to hit another wall and then EUV happened. And EUV happened because we figured out a way to corral this extreme ultraviolet, which no one thought we could figure out a way to generate, not only generate but get usable in a tool because it had to happen in vacuum. Then there's the next step because at some point you get smaller than at the atomic level.
So is there an end in sight? Theoretically, yes. But I know the company itself, we're always looking. Our roadmaps in our company, we don't look in the one to two years. We're looking 5, 10, 15 years down the road constantly. So we have roadmaps that go out that far. So I can tell you that we will continue to drive the market, at least for the next 10 or 15 years with what we know today.
But this company also relies on forging ahead with technology that we don't know. We said we need to get there. It's not a question of if, it's just a matter of how do we do it, and we just try and figure it out. It might not be a traditional solution. It might be something that no one's ever thought of. That's why working at SNL is kind of fascinating and fun as it is.
Max: Yeah, it's crazy to think you guys, it sounds like already have ideas on how to keep the innovation going out to the next decade. And then it's like, well, what are the chances that nobody thinks of any ideas from there to for the next 10 years? Like almost zero, right? So okay, this continues for quite a bit. The rate of change could always change but it's sort of hard to predict.
I have some questions here about… that I'm going to try out. We'll see if we have anything to say about this. A lot of my listeners are people who are in software, software engineers, a lot of people who work with data, like machine learning engineers, like myself. We think about hardware just a bit more in terms of like, okay, I have this algorithm that I run that's kind of hungry so how much do I have to pay for it? But we don't really think about it in terms of the technology underneath.
So do you have anything to say specifically to software and data people that you think might be of interest to them who don't think about hardware that much? Or even you guys are even, like several steps below the hardware.
Adam: Well, yes, and no. I mean, we have, I think 3000, if not more, 3000 software engineers worldwide. Specifically in Wilton, we have about 240 software engineers working here today. We certainly write a lot of software and we kind of split it up. There's what we would call embedded software engineers which, by trade, that's what I was, that's what I went to school for and that’s what I did. I started at ASML as a software engineer, 15 years ago.
These are the engineers that you're working on a board, whether it's a Raspberry Pi, or it's an old 8086 board or whatever it is, you get really excited when you make an LED light up or run at a one-hertz frequency because that's what you were trying to do. Then you take that to the next level and now you're not driving an LED, you're driving a robot to move a radical around or move away for around. That's really what the the embedded software engineers do.
Then you have the layer above that which are more the application engineers. Those are really the people that look at what's coming out. What is the end product of this machine? We're spitting out wafers as fast as we can but we're trying to also do that with a certain quality. Overlay, things like our KPIs, which are overlay and throughput, we want to make sure that every single line that we hit, or try to write, it happens at exactly the same place because the layers just get printed right on top of each other. When you're talking about nanometers, which we haven't talked about yet, but a nanometer is really small.
Max: Is that a billionth of a meter? Am I right here?
Adam: Billionth of a meter or something.
Max: Billionth of a meter, okay. So a millionth of a millimeter…
Max: It’s meter, then millimeter, then micrometer, then nanometer. Count the zeros.
Max: Is that close to the atomic level yet? Or is that more like… I'm trying to get a sense.
Adam: We’re getting close to the atomic level of silicon, yes.
We have closed-loop systems that read out how well we're printing out. Then in the system, there's probably a million machine constants that get twiddled as these things go to make sure that the next wafer is even better at the nanometer level.
Then you've got data science and machine learning, which are kind of an abstracted layer. I would say, this is a fairly new, commercially new science. For ASML, we generate tons of data, every second of every hour of every day when these things are running. The data scientists are taking in the data and again, doing things like predictive maintenance or more applications using the data to look at patterns over time.
We do sort of all the layers of software at ASML. Really, what you're interested in, there's a place. Whether you want to be down at the world level, really getting sensors and actuators to work, all the way up to you've got gigs and gigs and gigs of data that don't mean anything to anyone. How do I make sense of it?
Max: I'm almost imagining there's innovations on the hardware side, on the almost down to the atomic level, chemistry. Is this considered chemistry, light? I'm trying to figure out what the right engineering discipline would be for the EUV technology. You almost have that the hardware layer up to the soft layer, and then the people thinking about very abstractly computation, machine learning data. Then that goes back and feeds into innovation on the bottom layer.
I'm picturing a self-perpetuating machine that the whole world of software and the whole world of engineers are unwittingly involved in. Some giant cycle of, what do they call it? The… there's a name for it in business, the cycle of… shoot, I forgot what it's called.
Adam: I don’t know but in life, you would call it a self-fulfilling prophecy.
Max: Yeah. Okay. I like that one.
Well, this has been fascinating. We got a lot of great information today and I feel like this is a lot of stuff that maybe my listeners haven't been as exposed to yet. I think we'll get a lot of people thinking, so thanks a lot.
Are there any last thoughts that you have for our listeners today about our discussion? And where can they go if they want to learn more? I think there's a lot of things that people might want to learn more on both ASML, you, and the technology that you described.
Adam: Sure, from ASML perspective our website, everyone's got a website, right? We're no different, asml.com. They have a lot of information there.
I would say from an engineering perspective, you sort of touched on it, you talked about chemistry. Every type of engineer is required to work to get these things to run. We've got mechanical engineers, electrical engineers, software engineers, those are obviously we got mechatronic engineers, a large number of physicists, chemists, specialists, and material scientists. It's really, some people call it a sort of science and engineering playground because it requires almost every discipline.
On top of that, we're a high-tech company, but obviously, we have all different sectors. We have a very extensive supply chain. Tons of customers and vendors and so on and so forth. Really, there's opportunities for almost anyone at ASML.
And as for me, I'll share my information with you. If anyone ever wants to get in touch with me directly I'd be happy to answer any questions that come in.
Max: Awesome. We'll make sure to put it all up on the show notes page when this goes out. Adam, thank you so much for coming on the show.
Adam: Sure. Thank you very much. I enjoyed it.
Max: All right, let me know what you thought about that one. If any of you in the audience have any thoughts about where we go from here, from this interview, because we usually take what we learned on Local Maximum and follow it up with other interviews, other topics.
Next week, I think next week unless I'm gonna slip another solo show in there. It'll either be next week or the week after, I'm going to talk to the chief data scientist at the New York Times, Chris Wiggins, and historian, Matthew Jones, about a new book that tracks the history of data science and inference from, let's say, the 19th century to today. I think you're gonna be really surprised about how much you like these guys, even if you also enjoy my occasionally harsh criticisms of the New York Times. Chris Wiggins, I've interacted with quite a bit in the New York data scene and he's always a pleasure to talk to and to learn from. So look out for that one. Don't miss that. Have a great week, everyone.
Narrator: That's the show. To support the Local Maximum, sign up for exclusive content and our online community at maximum.locals.com. The Local Maximum is available wherever podcasts are found. If you want to keep up, remember to subscribe on your podcast app. Also, check out the website with show notes and additional materials at localmaxradio.com. If you want to contact me, the host, send an email to localmaxradio@gmail.com. Have a great week.
