Barron's Streetwise - The Dutch Company Powering the U.S. Stock Market

Episode Date: January 10, 2025

Without ASML’s machines, the A.I. spending boom would fizzle. So why are its shares slumping? Learn more about your ad choices. Visit megaphone.fm/adchoices

Calling all sellers, Salesforce is hiring account executives to join us on the cutting edge of technology. Here, innovation isn't a buzzword. It's a way of life. You'll be solving customer challenges faster with agents, winning with purpose, and showing the world what AI was meant to be. Let's create the agent-first future together. Head to salesforce.com slash careers to learn more. Just as a reminder, you know, an EV machine, you've got a thin droplet at the top of the chamber. It drops about 40,000 to 50,000 droplets per second.

And then the CO2 laser at the bottom needs to hit those droplets twice. 40,000 droplets per second. You can't miss a single one, right? In 24-7, 365 days a week. Yeah. Hello, and welcome to the Barron Streetwise podcast. I'm Jack Howe, and the voice you just heard is Didier Simima. He's a French-born, UK-based analyst for an American firm, B of A Securities, and he covers a Dutch company called ASML. And without its machines, NVIDIA and other AI firms would not be pushing the value of the U.S. stock market to new highs. Only ASML stock itself has started slumping and we want to figure out why. Which is why we'll be talking about firing lasers into tiny drops of molten tin.

My favorite part of that clip is when Didier says, so just as a reminder, as though I knew this sort of thing all along. I did not. But he does and soon you will too. Listening in is our audio producer, Jackson. Hi, Jackson. Hi, Jack. We have mentioned in the past that you live in Los Angeles. You're spending this week in New York. Los Angeles is in the news for wildfires. There are wildfires surrounding the city.

You are not here in New York because you're fleeing wildfires, although they're pretty close to your home, correct? Yeah, they're about a couple miles away from the west side of LA where we live, and a couple miles away from where my parents live near the fire that's in Altadena right now. And it's really scary. All my family and friends are safe, though, and thanks for thinking of us. And my heart really goes out to everyone who's affected. We wish you and them the best. Now we want to say hello to Oscar from Naperville, Illinois. Oscar was good enough to send us a recording of a question. And Oscar, this episode is for you. Jackson, let's play his question. We can skip the part where, you know, he tells us how wonderful we are. That's not a

requirement for getting your question used, although we're always happy to hear it. Thank you. Okay, will do. All right. So my question is about a company called ASML. This is a Dutch company. Given ASML's critical role in semiconductor manufacturing and the surging demand for AI-driven technologies. How do you view the growth potential for ASML? Could its position as a key supplier to companies like TSMC and VDM make it a good investment? Looking forward to your insights. Thank you very much. Thank you, Oscar. You mentioned ASML's critical role, and you're not kidding. The U.S. stock

market would not be doing this well without the chip designer NVIDIA, which is at the center, really, of a flood of spending on artificial intelligence. NVIDIA just designs the chips. It needs the company that runs the most advanced factories that make the chips, and that company is called Taiwan Semiconductor Manufacturing. And Taiwan Semi is dependent on the chip making machines. And there's none more sophisticated than the lithography machines. Those use lasers and mirrors to print these impossibly dense patterns on silicon wafers. And that is where ASML comes in.

Now, all of these stocks have done very well for investors over the past decade. NVIDIA has done the best. It has returned 29,000%. Taiwan Semi over the past decade, if you're using the American depository receipts, TSM, those returned 1100%. And ASML, that's also the ticker on the ADRs there, those have returned over 800%. But there's a difference. NVIDIA and Taiwan Semi are trading close to their highs, but ASML slashed its revenue guidance in October, and that stock is down nearly a third from its peak last summer. So that raises a couple of questions. What's going on at ASML? Is it a warning sign for chips and, by extension, for the U.S. stock market?

Or is it just an ASML thing? Is the company losing what is a monopoly hold at the moment on high-end lithography? Spoiler alert, I don't think it's any of those, and I think it's a reasonable bet that ASML will bounce back. My colleague Andrew Barry at Barron's recently named it one of his top 10 stock picks for 2025. So let's get into it. I'll give you a little background, and then we'll come to my conversation with DDA. There might be some things that are repeated here, but that's okay because a lot of them are complicated. I'll warn you in advance, this conversation could become extreme, as in extreme ultraviolet lithography. You probably don't have to hold young kids' ears. All right, I want to start

with Roy G. Biv, who is the CEO of Jackson. Trick question. Those are just colors, right? I was panicking for a second. It's the anagram from grade school. The colors of the rainbow. Red, orange, yellow, green, blue, indigo, violet. I always thought it was unfair to purple. You got to take that up with probably Isaac Newton or maybe Grimace or someone like that. What I can tell you is that wavelengths for those colors are the ones that human eyes are tuned to see. And they range from 700 nanometers down to 380 nanometers. And I'll get to what a nanometer is in just a moment. If you go longer, let's say that you blow past red, you come to some wavelengths that you can't see but are nonetheless useful.

You come to infrared, which is used in remote controls. And if you go further than that, you come to broadcast signals because long waves can travel far. Old-fashioned television, that used wavelengths that some of them were about as long as meatball heroes. And those are increasingly switching over to being used for cell phones. And if you've got ones much longer than that, you've got waves the length of football fields that are used for AM radio. Okay, what if you go the other direction,

smaller, out past violet? You come to ultraviolet, which is what's used in tanning beds. And when you get below 10 nanometers, you come to x-rays, which you use to see broken bones. And this is the neighborhood where ASML plays. It has machines called extreme ultraviolet or EUV machines that can print resolutions on silicon of 13 nanometers.

And chip makers can use techniques to turn those into even smaller features. That's how they make their current nodes or generations of chips that are named for some of the sizes of these features. 7 nanometer, 5 nanometer, 3 nanometer. And I said I would tell you what a nanometer is. It is roughly how long a fingernail grows in a second. Oh my gosh. Right there, that's two nanometers. The time it took you to say that. A human hair is about 80,000 to 100,000 nanometers wide.

And this is what has been keeping alive Moore's law. Not really much of a law. It's more of a trend. Transistor density has historically tended to double every two years or so with a minimal increase in cost. And that has helped make computers ever more powerful. These are room-sized machines. DDA says they're the most complex machines that humanity has ever created. ASML has a near monopoly in this business, EUV lithography. Its technology took decades to develop. The first EUV test chips were made at SUNY Albany in 2008. The first commercial

product to use EUV-enabled chips, that was 11 years later, 2019. Samsung's family of Galaxy Note 10 smartphones. The complexity of these machines helps keep competition at bay. So do ASML's supply deals, including one with Germany's Carl Zeiss. They make the world's flattest mirrors. And a company called Trumpf. It's a Trump with an F on the end. They make the world's most powerful pulsed industrial lasers. China is barred from buying ASML's most sophisticated machines. It's nowhere close to being able to make its own, as DDA will explain. A couple of more details. ASML's current top machines are estimated to cost more than $220 million. It's working on the next generation. Those are called high numerical aperture or NAEUV,

and those can print eight nanometer resolutions. They can support nodes starting at two nanometers, and they could begin showing up in financial results soon. The next generation of machines is called high numerical aperture or NAEUV machines, and they can print eight nanometer resolutions. Those will be used to support nodes starting at two nanometers. If the cutoff between ultraviolet and x-rays is 10 nanometers, and if ASML's machines are going to drop from 13 nanometer to 8 nanometer resolutions, why don't we call them x-ray machines instead of EUV machines? I don't know. I think if you make them, you get to call them what you want. But the bigger point about that next generation of machines is that they could start showing up in financial results

soon, and of course they will be more expensive. Prices there are estimated at $380 million to $400 million. Jackson, have I made things better or worse with this background information? Are you more or less confused than you were a few minutes ago? I take it that ASML is making fancier, more expensive machines. is making fancier, more expensive machines. Machines that shine a light on silicon to create patterns that are used somewhere in the process of putting circuits on chips, making those circuits very close together. That's why the wavelengths have to be very small

so that the circuits can be very small. So the machines are getting fancier, better, more expensive. Why is the stock slumping? Ah, I'm going to let DDA tell us that and we'll circle back to those points at the end. So let's get to our conversation with DDA Simima. He's an analyst at B of A Securities who has studied ASML for decades and he was on the go, but he was good enough to give us a call from the UK. I don't know if there are any acronyms I should explain in advance. You might hear DRAM and HBM. DRAM is a type of memory. HBM stands for high bandwidth memory, the new stuff in other words. Let's start with who the users are for EUV lithography machines

and the chips they make versus which companies can get by with older technologies. A computer chip is made of transistors. And the more transistors you have per chip, the more computing power or memory density you have. And so to give you an example, the latest NVIDIA Blackwell chip for AI servers has got two times, I think, 96 billion transistors and two 800-square-millimeter chips or dyes.

So they are like monstrous, very big chips, extremely powerful processors and they require a SML machine. So who are their customers? Their primary customers are chip makers. So companies who have factories where they make chips. So either their own chips like Intel, like Samsung, like SK Hynix, or it can be also companies that provide foundry services or manufacturing services for third party. Best example, of course, being Taiwan Semiconductors. TSMC, they are the biggest customer of ASML, and they make chips for pretty much everyone to various degrees. Biggest customers are Apple, NVIDIA,

AMD, Broadcom, Qualcomm. The most advanced ASML machines, where they have a monopoly today, they are the only company that can provide these technologies called Extreme Ultraviolet. This technology is primarily used in smartphones, in laptops, in desktop, in servers, in autonomous driving, in gaming systems, right? And you use a bunch of chips, but the most powerful element in those systems is using EUV technology from ASML. And then in older technologies, older markets, like think about automotive or industrial applications,

you don't really need these technologies, these advanced geometries, so-called like two nanometers or three nanometers or five nanometers. You need things like 28 nanometer chips, 45 nanometer chips, even 90 or 130 nanometer chips, chips that were technologies that were available 15, 20 years ago. They are very mature technologies. And so they use older machines from ASML called DeepUV. So they are effectively enabling less fine, less advanced geometry, so bigger distance between transistors.

I asked Didier what has gone wrong for ASML in recent months? Why the cut to the revenue guidance? And the first thing he mentioned is China. China is only allowed to buy some machines from ASML, and it's worried about even tighter restrictions soon. So it has greatly increased its orders, and that's something that is now poised to go back to normal. Here's Didier. 2024, you know, China accounted for, you know, in the first half, it was like 50% of the SML revenues. If you look back, you know, prior to COVID and when this thing really took off, China was around, you know, low double digit or high single digit percentage of revenue for SML. China became like this, you know, massive, you know, demand driver.

So in 25, now as SML used to say, look, China is going to be flatty from here. Now they are saying, look, China is going to drop to about 20% of their revenue. So if you do the math, that's a mid-20% revenue decline in China. Okay, so China is one reason. There are two others. We've talked recently on this podcast about the struggles at Intel. One of the things that Intel has tried to do is greatly expand its manufacturing capacity to become more of a foundry, more of a third-party manufacturer for other chip designers. And it has not met with so much financial success in that pursuit.

So it's scaling back its plans and it's pushing out orders for chip making equipment. Samsung is doing the same. And there's a third factor that's really affecting both Intel and Samsung, and that is something of a chip slump. Now, I know that we hear often about raging demand for artificial intelligence chips and data centers and so forth. And I know that the U.S. stock market is riding high in large part because of those chip sales and sales of related software and devices.

But that's AI. Personal devices like phones, personal computers, cars, gaming machines. If you look at the Philadelphia Semiconductor Index, which has 30 different chip stocks in there, two-thirds of those stocks are underperforming the S&P 500 index over the past year. So things aren't great across all of the chip business, only some of the chip business. We'll comment a bit to the prospects for ASML turning things around. So that answers one of the questions we had talked about earlier, which is that the fact that ASML stock is trading lower and the company's cut revenue guidance, that doesn't mean that there's trouble ahead for NVIDIA or for the U.S. stock market.

AI is still doing quite well. It's the consumer devices that could use the lift. And we'll come soon to the prospects for ASML bouncing back. I asked DDA about the chances of ASML being replaced. Let's come to that right after this quick break. Welcome back. Jackson, we're talking about ASML. We're talking about its marvelous lithography machines, why the stock is down. Let's just play my conversation with DDA from there, and we'll hear about what keeps the company safe from competition. We'll hear about the keeps the company safe from competition.

We'll hear about the prospects for a recovery from ASML. We'll get some of the technical details on how these machines work, including the part about firing lasers at molten tint. I don't know everything about that process, which is to say anything.

Here's honestly what I was picturing when he was describing it to me. There is a ride at Disney World called, I think it's called Toy Story Midway Mania. And you sit on these spinning swively cars and you got a like kind of a gun thing, but you pull a string to fire it. I know the one. You keep going in front of these cartoony images and you got to shoot the cartoony images. On one screen, you're putting rings around aliens, and another, you're popping balloons with darts. Here's a fan video of the same ride at Disney in California, just to give you an idea.

Hey, partners! And what DDA talks about, 40,000 to 50,000 droplets per second, and you have to hit everyone twice with a laser and you can't miss any, or you might mess up your whole wafer. That's kind of what I think of. I wasn't that accurate, but I did get the high score on my family. I mean, my kids were probably five and nine at the time, but I am still the reigning champion of my family in that ride. Anyhow, that's the way I picture it. I might not have all of the technical details correct. And with that, let's hear DDA. Is there anything that's changing about the intensity of using this process in making chips? In other words, are manufacturers saying, wait a second,

these machines now cost $300 million and more. Can we find ways to work around them and do less of this in our manufacturing? Is that happening or no? I think that's really the biggest debate in the market right now. I think the market view is really that this is happening. You've got effectively two, let's say, alternative technologies which are not using lithographies to enhance the performance of those transistors. The one is called, it's a new transistor technology called Gatoround, so G-A-A. So the first Gatoround-capable chips

will come out of TSMC production line probably the later part of 2025 or earlier part of 2026. It's going to be used for AI and for probably laptops, PCs, maybe things like that. And long story short, you are not shrinking. So you move from three nanometers to two nanometers, and there is no shrink, even though it's a misnomer, right? It's called two nanometers, but there is no shrink. So there is no shrink, even though it's a misnomer, right? It's called two nanometers, but there is no shrink. So there is no incremental spend on lethal.

Instead, you are adding, let's say, transistor performance with a boost in other technologies like deposition of certain materials, etching of certain materials. The second one, it's a way to boost the performance of your chip by stacking chips on top of each other with memory and logic. So these two technologies, you can argue, are somewhat deflationary. But they're not replacements. They're not new things that do what EUV does. They're just ways to get more mileage out of existing machines, right? That's exactly right.

And I think the way I think about it is that anything that boosts the performance of semiconductors, anything that lowers the cost of semiconductors at the advanced nodes is good for ASML because it's going to encourage new customers, new people to move to those advanced machines. So to give you an example, one thing that is helpful for ASML right now is that obviously NVIDIA is building these massive chips and they consume a lot of wafers. The problem is that they are not at the bleeding edge. So Apple is the most advanced customer of TSMC and the most advanced in the world. They are 3 nanometer for the iPhone, and they will move to 2 nanometer in probably calendar year 26, Christmas.

NVIDIA, the majority of their products are built on 5 and 4 nanometers. That's so interesting. I think of NVIDIA as being out in front on everything, but you're saying they're not in chip design. So not in transistor design. So because NVIDIA, there are two reasons for it.

The first reason is that their chips are so big and they are so dense in terms of transistors that if they were to use, you know, two nanometers right off the bat

when they just come out of the production line of TSMC, the manufacturing yields would be very, very bad. So TSMC margins would be impacted, and so that would be unhelpful. So they need to get the process technology of TSMC very mature. So they typically debug that with Apple. The second reason is simply that NVIDIA,

you will never see Jensen talk about, you know, four nanometer technology, even though that's what he's using for Blackwell. He's talking about a complete system. And I always say, you know, NVIDIA, they don't sell you a chip, like Qualcomm will sell you a 5G chip. NVIDIA is selling you an iPhone. They sell you chips, networking, memory, the whole box, the operating system, the application layer. We've been talking mostly about processors, but this technology is used to create chips for memory too. And do I understand correctly that there's a larger growth opportunity right now in memory in the years ahead? How would you characterize that?

Exactly right. Things like AI servers are going to require a lot more DRAM chips, so type of memory chips, which are like buffer memory. And those AI servers are very, very greedy for a type of DRAM chips called high bandwidth memory chips. And obviously, as we move into more and more complex, large language models, we're going to require more and more memory to do the training of those algorithms. So today, those chips typically come from SK Hynix in Korea. They've got a very high, very dominant market share. And SK Hynix, if you want, have been the pioneer in using EUV technology from ASML for those 8VM chips.

So they had the lead versus Samsung and Micron in coming with HBM technology. And they've decided to double down if you want to accelerate their manufacturing process by using EUV from ASML. The reason why they've done so is because compared to normal DRAM chips that you find in a PC or in a smartphone,

HBM DRAM chips are at least twice as big. Because the chips are bigger, when you make them, manufacture them on a wafer, your yields, the manufacturing yields are going to be poor. So the number of good chips for wafer are going to be poor.

If you use EUV, the main value added of EUV is that the quality of the print is much better. Tell me about NAEUV. This is high numerical aperture EUV. That's the next thing. You talked earlier about deep ultraviolet and then extreme ultraviolet. And then I guess NA EUV is what enables the next leap forward. When does that become something that enters the marketplace and becomes profitable and makes money? something that enters the marketplace and becomes profitable and makes money? High NA, what the SMM machines do, they print lines and cuts like that in 2D at a very, very fine level with a resolution of only a few nanometers.

If you want to move to something like one nanometer, you're going to start to have to go through two times to draw a proper line or proper cut, which means you have to buy twice as many machines. So the benefit of high-end A is that with one single brush, which is much thinner, you can do that line and cut in a perfect way. So we think around the 1.4 nanometer node, which is probably 26, 27, most certainly at the one nanometer node, before the end of the decade, at least for sure, you'll have those high-end machines in production. Today, we know it's public. Intel have been the first recipient of those high-end machines.

But obviously, those machines are extremely expensive. We think they cost about, you know, 380, 400 million euros, something like that. For long-term investors who are looking at this stock, what keeps ASML safe? They have this enormous market share. What's special about them? I've read about their relationship with this German company that makes the mirrors, Zeiss. Is that part of it? What makes ASML so special? I think EUV machines and, by definition, high-end machines. And by the way, I think EUV machines and by definition high NA machines, and by the way, after that, you'll have hyper NA around 2035 or so, which are $800 million machines. But those machines are the most complex machines invented by men or by human beings, I should say.

So I think, you know, the moats that they have, the competitive advantage that they have is extraordinary. We can go through the whole history, but what makes ASML so special is, yes, they have this special relationship with Carl Zeiss. So they own 25% of that company. It's an exclusive relationship both ways. So those guys make the most, the flattest mirrors in the world. To give you an idea, it's so flat that it's flatter than a mirror that would be the size of Germany. And there would be no more like in the whole of Germany. So that's flatter than a mirror that would be the size of Germany. And there would be no mold in the whole of Germany.

So that's flatter than that. You said there would be no what in all of Germany? No mold. The second one is another German company called Trumpf, which makes their carbon dioxide lasers, which are used to explode the tin droplets in the vessel that create the UV light. You're firing a laser at molten tin droplets to create this light that you need. Is that right? Correct.

So 40,000 droplets per second. You can't miss a single one, right? In 24-7, 365 days a week. Yeah. It's getting so absurd, the complexity. I want to ask you lastly about the stock. But just before I do, what is the probability, just following up on that question, what is the probability that China right now is saying, we might lose our access to these machines

and we need these machines, and therefore we're going to cut out the rest of the world? We're going to make this a national imperative. We're going to make our own Chinese homegrown ASML, and we're pouring all of our smarts and resources into this. What is the probability that behind the scenes somewhere, they've been working on these same types of machines, and they're nearly there? Is there any chance of that, or is it out of the question? So first of all, ASML is forbid from sending EUV tools to China since 2020. So that's part of the Vassana agreement. But to your point, you know, it's all public knowledge. China wants to develop their own lithography 2020. So that's part of the Vassana agreement. But to your point, it's all public knowledge.

China wants to develop their own lithography systems. But there is one, let's say, known company in China called Shanghai Microelectronics, which is working on DUV tools. The CEO of ASML probably is the best source for that. He said that China is about 10 to 15 years behind on process technology. Now, I don't want to discount China because clearly it's a country that has got immense resources and very, very fine individuals and great universities, et cetera. But they will struggle to get those flat mirrors.

They will struggle to industrialize EUV. But I think the odds of them being very successful in EUV over the course of the next 10, 15, 20 years are fairly low. And just to finish off on that, if you look back in the early 2000s, there were four companies in lethal, ASML, SVG, which ASML bought, and Nikon and Canon. And the market share were roughly split between Nikon, Canon, and ASML. And today, ASML has got 92% market share were roughly split between Nikon, Canon, and ASML. And today, you know, ASML has got 92% market share. So they want to square purely organically through innovation.

When you put it all together, you have a positive recommendation on the stock. Describe for investors what the opportunity looks like. In other words, what do you think that the rest of Wall Street is getting wrong about this company? Where will it perform above expectations? And what are maybe the one or two biggest risks from here of it falling short of your expectations? Well, look, I think, you know, it's a bit of a tug of war between shorter term investors who are worried about potentially additional U.S. export control restrictions or Dutch export control restrictions on China. That would impact the stock in the shorter term. ASML, at the end of the day,

is dependent on consumers buying smartphones, buying PCs, buying gaming systems, buying cars. So the consumer demand has been very lackluster, especially in 2024. People would have expected that to pick up. So I think we would want to see smartphone, PC, gaming, et cetera, these demand to really improve. If you look at the longer term, if you look at that 2030,

which is only five years away, it's not that far away. ASML are telling you that they're going to do between $40 and $60 billion of revenue. They're telling you they're going to do 56% to 60% gross margins. So the earnings growth over the next five years is about 16%. So that's a very aggressive, very solid profit growth for that company over the course of the next five years. Thank you, DDA. Oscar, I hope that that gives you some useful information.

ASML stock does not sit at some rock bottom price to earnings ratio. I saw it recently at around 37 times the slashed earnings estimates for 2025. JP Morgan a while back said that the stock is trading at a 10-year low premium versus U.S. semiconductor equipment companies. So if it's not cheap, it's cheaper than usual. And as DDA points out, there could be handsome growth ahead. Things to keep an eye out for are the recovery in consumer devices and whether Intel and Samsung can lure some of those chip manufacturing customers from Taiwan Semi. That will be good for new machine owners.

Things to look forward to are memory using more EUV machines, just like processors have, and the shift to NA EUV, which could kick off a whole new spending cycle, and of course, continued strong spending in AI. Didier's price target for ASML recently implied about 26% upside for the American depository receipts over the year ahead. We'll see, and good luck if you're invested. Thank you, DDA and Oscar. I want to thank Roy G. Biv and Sheriff Woody and the gang at the Toy Story Ride. And thank you all for listening.

Jackson Cantrell is our producer. I'm going to do some research on that mirror the size of Germany with a mole on it. Right. I've read that. If their mirrors were that large, they wouldn't have an imperfection that's larger than a millimeter, or I guess the size of a mole on your skin. That's really impressive. I've got a much smaller mirror than that in my bathroom, and all I see is imperfections when I look into it. I see earlobe hair that didn't used to be there. You know what? We don't, we're out of time.

Just be thankful. It's not the size of Germany. If you have a question you'd like answered on the podcast, you can tape it on your phone. Just use the voice memo app and send it to jack.how. That's H O U G H at barons.com. See you next week.