Actively Speaking Podcast

Semiconductors: The Eyes, Ears and Brains of the Digital World

February 22, 2021 Epoch Investment Partners Episode 27
Actively Speaking Podcast
Semiconductors: The Eyes, Ears and Brains of the Digital World
Show Notes Transcript

Semiconductors are the eyes, ears and brains of digital world. They are being relied upon to solve greater and more complex problems all while being smaller, faster, less expensive and more energy efficient.  Listen in as Epoch Senior Research Analyst Matthew Chan discusses how the explosion of exciting breakthroughs in AI, autonomous driving, 5G, and cloud computing could likely drive double-digit growth in the industry. (February 2, 2021)

Speaker 1:

Hello,

Speaker 2:

And welcome to Actively Speaking. I'm your host, Steve b Weiberg. Join us each episode as we discuss current issues concerning capital markets and portfolio management from the perspective of an active manager.

Speaker 1:

Well, welcome back everyone to another episode of Actively Speaking, I guess this is actually our first episode of 2021, so I deleted Happy New Year to everybody. Today we're gonna be talking about semiconductors, which are, uh, certainly a timely topic, lot to talk about and to, to talk about it with us. We have Matthew Chan, who's an analyst at Epic. He's been with the firm for 10 years. So, uh, welcome, Matthew. Thanks

Speaker 3:

Very much, Steve. Thanks for having me.

Speaker 1:

I guess we'll start out, you know, I get, it's in my role here as host, I get to ask dumb questions, so I I I have to take full advantage of that. So, uh, I'm gonna start with a really dumb question. Why do we, why do semiconductors called semiconductors and, uh, you know, when do they get upgraded to be full conductors?

Speaker 3:

<laugh>? I'm glad you asked that, Steve. I'm, I'm pretty sure you're not the only one thinking that. Well, the term comes from the fact that, uh, by its very nature, semiconductors sometimes conduct, and sometimes they act as insulators, or in the industry we call'em dielectrics. Depending on whether you have voltage on or off, uh, it can sometimes conduct, uh, essentially sometimes they'll give you a one signal and sometimes they'll give you a sterile signal, depending on the electrical circumstances.

Speaker 1:

Got it. Okay. So, so walk us through the history, uh, briefly of, of semiconductors. When did they get invented and, uh, how has their use spread over time?

Speaker 3:

Absolutely. So semiconductors are essentially the brains and as well as the eyes and ears of machine world, they are essential in the increasingly automated world We live in. Chips that are familiar to many of us include, you know, the, in the intel inside X86 processors, you, you and I have on our computers. If you've got kids at home, you, you know about NVIDIA's, uh, graphics processors, GPUs, you know, on your phone is, is a qualcom baseband chip. And, you know, on your phone as well is, is a Sony's, uh, camera imaging sensor. So, so these are the semiconductors that are most known to us commonly, but semiconductors also include less cutting edge, less sexy, but equally essential chips, such as analog chips. These are chips that con convert real world signals into digital signals. They also include power management chips that manage voltage and battery consumption in, in any device, as well as the street components that manage and store electric current. So the very basic semiconductor chip is a transistor. Mm-hmm.<affirmative>. And that is a very, very building block of any semiconductor chip. Uh, with transistors, you can build basic logic chips. These are and chips or chips, man chips, more chips. And with these basic logic functions, you can basically build modules, modules to subtract, to add, and then multiply and so forth. And then it becomes, uh, and then when you multiply that you can do, do very complex functions. To your question, Steve, it was, uh, not that long ago, I mean, it was, transistor was invented 1941 in Bell Labs in Murray Hill, New Jersey, in fact. And the team was led by William Shockley, uh, 1941, who invented, uh, the, the first transistor. William Shockley went on to win the Nobel Prize 15 years later. But, uh, we've come a long way since then.

Speaker 1:

We certain, well, we certainly have. So, and talk about how, of course, the, I think one of the most famous quote laws or rules, uh, things that people know about semiconductors is how they've just gotten smaller and smaller, but also more and more powerful at the same time, uh, over time. So tell us about, you know, who, who was Gordon Moore and what is Moore's Law?

Speaker 3:

That's right. Uh, Gordon Moore, the famous founder of Intel, he laid down the gauntlet in 1965, and he said that number of transistors on a chip should double every two years. And pretty much the industry has managed to do that, you know, not, not for the sake of doing it for, but we've needed, uh, the chips to perform at greater speed. We've needed semiconductor to solve greater complexity, as well as lowering costs, as well as lowering the power envelope. Uh, these are all motivations for making semiconductor chips smaller and smaller so that we can have the more powerful processors that can consume less power and solve very difficult problems for us. And so, Moore's Law is famously known in industry, and it's led us to some very miraculous discoveries. And we're talking about five nanometers as being the cutting edge today. And for matter of perspective, one micro is a thousand nanometers. So we've come a very long way. And as an example of the power that we have in today's modern chips, if you open the latest iPhone, you'll find apple's a 14 chip, and there's 12 billion transistors on this tiny dye size of the snowflake.

Speaker 1:

That's amazing. So give us a, a sense of the, how big is this industry today?

Speaker 3:

It's an enormous industry. I'm glad you asked. Uh, we're talking about four 40 billion in sales for 2020 up about 7%, uh, year over year in 2020. And that's just semiconductor revenue sales. I think some of these in a more visceral way, we also need to keep in mind that we're talking about the billion units. So this is a very big industry. Uh, it means that, you know, while some semiconductor chips may sell for less than a penny, some semiconductor chips on the high end will sell for more than thousand dollars. So range is in terms of the, for these products.

Speaker 1:

Got it. So let's talk about this industry. It, I know as a, as somebody who's been involved in investing for many years, semiconductors used to have a, a reputation for extreme cyclicality. I mean, there's this well known index called the, the SOX index. It's the Philadelphia, uh, stock exchange, a semiconductor index, and mm-hmm.<affirmative>, if you go back, you know, 20 years ago, it used to be the case that when those stocks were booming, they would outperform the market by literally a hundred percentage points over a year, but then the next year they might underperform by by 50 percentage points. So there was definitely kind of a, there was a boomer bust image to, to this industry. Is that still the case? Has it changed? And if so, why?

Speaker 3:

It is in fact, uh, still a cyclical industry, but much less cyclical than it has been in the past. There's been a lot that has transpired in the last two decades. A lot of consolidation and a lot of exodus of participants who are simply not able to keep up. As a consequence, the industry has become very consolidated. It's consolidated into several superstars who really dominate their served and market and their served vertical. If, if I may, Steve, I'd like to offer you a few examples of this consolidation. Maybe we start, start first with advance manufacturing. Uh, these are the companies that are able to make the most advanced logic chips. We had about 30 players at the turn of century. Now you have only Taiwan Semiconductor and Samsung at the leading edge. This after Intel fell to keep up last year. In the years that I've covered semiconductors, I, uh, was on the side about 20 years ago, uh, covering the space. You know, I've seen IBM Global Foundries, c Charter Semiconductor, and a host of other companies just simply unable to keep up with leading edge manufacturing. It's a very difficult process to continuously innovate and get to the next process node and the issues that these companies face are laws of physics and also yield issues. And many simply have not been able to keep up. So if you look at advanced manufacturing only, TSM and Samsung today, both for vying for, uh, to produce three nanometers this year, if you look at the DRAM space is another very important vertical he ran, is dynamic random access, memory, otherwise known as working memory or buffer memory. Uh, we had dozens of players in the 1980s, the fields down to about six players by 2000. And then since the bankruptcy of in 12, you only have three players who dominate the market, and they, they're Samsung, Hynek, and Micron. And another, uh, another industry where we've had not had any new entrance. You know, one industry veteran, uh, equates it to new entrance coming into the field, try to catch a high speed train on foot, and it's evolving that fast. And, and you don't mind, Steve, I like to give you one more example. In the analog space, we've also seen massive consolidation. You know, for example, the industry, Vanguard is PI or Texas Instruments. They are the industry's most profitable player and also the largest player. Massive consolidation over the, the years. Uh, they bought a facility from chema, then they acquired National Semiconductor in 11, and it's the second largest player in the space. Analog devices has also been on acquisitions screen. They acquired linear technology in two 17, and they're in the process now acquiring Maximum Integrated. So with this consolidation, these companies have both, uh, become more profitable and more stable. Similarly, in the equipment space, we're seeing the same type of consolidation. Years ago, the photography company, ASML had two competitors, cannon, and, and, but today in advance, lithography, uh, otherwise known as E U V, extreme Ultraviolet lithography, ASSL stands alone. They're a monopoly after Cannon and Nikon bailed out, unable to keep up. And so this is a common theme in the industry. The consolidation has really yielded, uh, much less and much, much less volatility for the space.

Speaker 1:

Okay. And how about, I mean, there, I, I think there's also been, hasn't there also been a, um, you know, if, if you go back to that period that I was talking about 20 years ago when, when the SOX index would sort of swing wildly, you know, most of the demand for chips was coming from, you know, PCs, uh, as to when the PCs came along the mid eighties, and then spread throughout, you know, our, our economy. And there was a lot of demand for the first 10 or 15 years until pretty much everybody, whoever was gonna get one got one. And the demand would, it was when there would be a new generation of PCs would come out, like you mentioned, you know, I remember when the, you know, the 2 86 machines were outed then the 3 86 machines, then the 4 86 machines, which was a reference to the, the chip, the intel chip. But seems to me there's, there's, uh, we, we've moved way beyond just PCs now in terms of, uh, the end markets for chips.

Speaker 3:

That's right. It used to be that the Windows cycle and the PC cycle drove most of, most of demand semiconductor land. And that was, uh, basically dictate cadence, semi out the growth and, and fall, uh, today is very different. PC, believe it or not, is less than 30% of overall demand right now. And that includes servers. Communications is actually the largest end market today. It's about 35% of total semi demand. And we have contributions now, significant contributions from industrial and consumer and automotive each at 10% of the end market. Um, so as this end market becomes more diversified, we've also seen much less volatility in this space.

Speaker 1:

Yes. I, I think people are probably familiar, they've heard news recently about auto companies having to restrict production because they can't get enough chips. And, uh, you know, who, who would've ever thought that semiconductors were the, the, the bottleneck for, uh, auto production?

Speaker 3:

That's right.<laugh>. Yeah, it, you know, it, in fact, automotive is, is the highest growth market for semiconductors right now.

Speaker 1:

Well, let's talk about that a little bit. About, uh, differences in these end markets and, uh, terms of growth rates or are there different players for the, you know, specialize, that specialize in producing chips for these different end markets? Give us a feel for that.

Speaker 3:

Definitely. So I think there's two ways to market. One is by product category and another is by end market's start first product category. Uh, the biggest product category within the four 40 billion that I discussed earlier for semiconductor sales are memory and application specific logic. And within this is billion industry of revenue, man. A man is flash, otherwise known as storage memory. So within the theam space, you have three players that dominate space, Samsung, Hynek, and Micro in man. Uh, there's about six players. So it's, it's the aforementioned three players, plus you have Uhhi, uh, you have Western Digital, and now you have a Chinese player in the man space as well. Memory is very important. Cause as data continues to grow, we need more storage, but we also need more buffer or working memory, uh, memory that you can use to compute on this data. And so both Dran and man are very high growers. Uh, if you look at big growth, for example, Samsung offered, you know, its outlook for the industry, we should expect big growth of at least 15% for D and at least 30% for NA for the coming years. Another big area, uh, area of excitement is the application specific logic chips. Uh, this is also 118 billion industry, you've no doubt heard about the Avidia graphics chips. They're actually now using a data center as data center accelerators. In addition to being, you know, your video gaming cards, there's a lot of new innovations here. You've got companies like Amazon creating their own graviton chip. Um, they're designing inhouse and having psm, uh, and this is for their data center. Uh, you have Google with their<inaudible> processing unit, otherwise known as TPUs. Again, uh, TSM is, and these chips are Google's way of taking AI to the next level. In fact, Google is, is probably in leadership position in regards to AI because of its, uh, internally developed chips. Tesla is, is working very hard on the eight as chip. This will be manufactured at Samsung. And you know, you may have heard that Alibaba, the Chinese internet giant has their own in-house chip as well. Another, uh, example of this application specific chip, which is used for crypto mining. So these are all the contributing to the high growth in these, uh, application specific logic chip. Then I want to go to analog chip. These are very important chips that convert real world signals into the digital world, so, so that you can manipulate digital signals, but analog is, uh, space of significant market, uh, worth 60 billion in 2020. Uh, no doubt you know about the microprocessor space, and that's a fairly sizable market as well. 52 billion in size. And then there's a space known as discrete and passive semiconductor chips. These are very basic chips that manage voltage store electricity, and they are essential in any computing device or any electric circuit. And this is about an 80 billion industry. So that's from the product level. Steve, I, I, I would highlight memory again and application specific logic as the higher growers. Then if you look at it from the end market perspective, automotive and communications continue to see the highest growth followed by industrial consumer. And lastly, pc. And the reason for that is we are essentially x growth in the PC market. Uh, certainly, uh, COVID has provided a nice for all of us upgrading our equipment and buying laptops, but, uh, the PC market in, in the next couple years is, and has, has been this way, is X growth.

Speaker 1:

Uh, well, so, you know, apart from PCs as everything else you were saying, there's, there's clearly a lot of growth in, in this industry. So let's, let's talk about the investment case for this industry, because clearly there is a lot of growth overall. And, uh, if, you know, if you're looking for a growing cash flow stream, um, seems like this is an industry that, that has that, but the stocks have done well, many of them have done quite well over the last couple of years. So, uh, let's, let's talk about what the investment case looks like. What does, for example, how does the valuation, uh, of, of this industry compare both, you know, to the market, to its own history, and, you know, how do things like return on capital factor into this?

Speaker 3:

That's right. Performance has been strong, as you mentioned, the sox, which is the Philadelphia Semiconductor Index, uh, it's a cap index of 30 large semiconductor companies. These are global companies that are listed in the US directly or through s last, the last five years. Even though, uh, I would also add that, uh, uh, the, that has, and the volatility of that has definitely decreased in the last two decades. Uh, it has continued to outperform s but it's largely outperformed cause of stronger than s p free cashflow growth. So it's not, uh, multiple expansion today. If you look at the, IT trades at about 23 times forward PE and about 16 times forward e to ebitda. And this compares to the s p2 times PE and 14 times ebitda. So yes, little above s I would also add that historically one point, 0.2 times out of the s p, so not on line here. Um, but what is different now is, is your question earlier, is that you have this industry that is actually much less cyclical and much less volatile before, but it has an incredible returns profile owing to the cons, the massive consolidation we've seen, for example, the SOX average operating profit margin is 22%. This compares to, I think about 11% for the s and p, the return invested capital 14%, and that compares to about five, 6% for the s and p. So what you get in investing in the SOX is you get higher growth, much higher quality for small premium to market valuation. Uh, this is why we continue to be constructive on the space, uh, on the industry and in particular, global champions that dominate their market, uh, industry lingo. We call it sam, service Addressable Market. These are companies like Applied Materials, asml, Samsung<inaudible>.

Speaker 1:

Okay. Well that sounds, that sounds really interesting on the investment case, uh, for the industry. To finish up, let's talk about a new paper that you've worked on. It's, it's called, uh, Moore's Law and the Race for the Rest of the Chess Board. It's could well be up on our website by the time this podcast appears, but if not, it should be up shortly after that. We, we talked about Moore's Law earlier, of course, which was about the doubling of the number of transistors that could fit on a chip, uh, every 18 to 24 months. And the, the chess board, uh, reference is, uh, refers to an old story, undoubtedly apocryphal about the inventor of chess. And what happened when he presented this invention to the king of wherever he was living. And the king was so pleased, offered to give the inventor whatever he wanted, and the inventor said, oh, I don't want much. I, you know, I just want some food. So how about if you give me, put one grain of rice on the first square, two on the second square, then double it to four on the third square, eight on the next 16, and so forth, just keep doubling on each square. And that the king, you know, thought about that for just a second, said, oh, that sounds reasonable. I mean, I, you know, that probably doesn't amount to all that much, uh, grain. But, uh, you know, the, the truth is that when you, uh, keep doubling 64 times, you actually get to, on the last square, you need about five quintilian grains of rice. And of course, when you add them all up on all the squares, you're, you've got, uh, almost double that. So that, that's refers to the power of doubling. And so when we talk about Moore's Law and the doubling of transistors on a chip every 18 to 24 months, well that was as, as Matthew said, you know, that was about 1965 or so, that's, uh, you know, over 50 years ago now. So we've had quite a few doublings, and if you put it in the context of a chess board, we're sort of, you know, we've gotten to the halfway point on the chess board in terms of the number of times we've doubled the, those chips. And what it has led to is this, you know, true explosion in the capabilities. And we've seen all these things like, uh, AI and, uh, artificial intelligence have come along, that that truly would not be possible without all the, the doubling of, uh, semiconductors, transistors on a ship that we've seen over the years. So the, the paper I understand is, is about what's gonna happen on the back half of the chessboard as we keep doubling. And I guess my first question is, is that really possible? Are, is there, is there a physical limit? You talked about how these, you know, these things are getting down to like three nanometers and you know, I, I think I've read separately about, you know, we're talking about things that are like 10 atoms wide. Are we getting close to a physical limit where Moore's law will, will not really apply anymore?

Speaker 3:

That's right. There are fiscal limits, and the industry has been grappling with it. As we get to smaller and smaller geometries have more leakage issues that the industry has been trying to solve, otherwise known as quantum tunneling, what happens here is when you have this gate with phase three nanometers, it's very easy for electron to inadvertently cross the channel even when you don't wanna, and the industry's come up with very, some very creative ways to deal with that. Fin you may have heard, is a 3D structure that is essentially a 3D gate, and now Samsung is proposing something known as a gate all around structure also to deal with this issue. But eventually, you know, these laws will become, uh, greater and greater headwind. Uh, we have definitely visibility to the atomic layer deposition level, where at some point we're gonna be depositing Adams and Adam atomic width, that Adam could be the width of your gate. But it's hard to see beyond that at this point. The industry has been very creative in solving many of these physics issues. And, uh, if there's one thing you can, uh, you know, hang your hat on is it's human ingenuity. We've overcome a lot and I think, you know, there will be a new, a new breakthrough when that happens. But the paper you, you mentioned, Steve's a fun paper for, for me to work on, wanted to address, uh, where we are today. Certainly the, the many doublings you've mentioned has carried us a, a, a long way in terms of the, the power of the chips that we have in our hands. If, again, referencing your, your iPhone, uh, that chip is far more powerful than any computer that existed on earth, you know, 10 years ago. And that's in, in your pocket now. But as a society, we're trying to solve very difficult problems. We're trying to solve back to half of the chess port problems. And, you know, with the advances in technology, we're finally able to address some of those issues. Case in point is, if you think about what autonomous driving involves, uh, level three or level four autonomous driving essentially is, is when you know you have unaided by driver. In normal circumstances, the car is driving by itself. Uh, when you think about that, it, the complexity that's involved for the chips to make split second decisions, pedestrians from pothole and signage and other, as well as other intentions, you know, as the car navigates through a busy intersection. These are very difficult problems that we're grappling with today and cause of the power of, uh, Morris Law and the, we've had, we're actually at the point where we can actually solving some of these issues. As another example, there's been a lot of excitement about Bitcoin and crypto in general. Um, this chain type of, you know, monetary system, if youll, it does offer some very interesting, um, promises. But the M bitcoin coin network must crunch through 7.2 times 10 to the 22nd power hashes. This is type mathematical function that involves adding mathematical components. You have to through that to solve one puzzle in the chain. And that process, you're consuming 72 terawatts of power. So we can do it today, but as chicks get smaller and smaller, uh, the good news is that they'll consume less and less power and they'll be able to do it faster and faster. As another example of the half of the problems is, you know, all the things that as society we're trying to do as AI becomes more in of as well. And, you know, as an example of the work that Google has done in ai, you know, I would like to reference Alpha Zero. Uh, you may have heard of the chess engine from Google called Alpha Zero in training itself after nine hours of self-talk play where it played millions and millions of games, it was able to defeat the reigning chess program at the time, stock eight. Um, this happened in 16, uh, a year later, Google's alpha using Google's TPU was able to accomplish the same in the game of Go. And this game was for computer game team. Cause this game has 10 to the 50th possible moves. So, you know, I speak of chess and, and go and, but there is far more relevant applications that as a society we can tackle today. And I want to point to drug discovery. This is a very complicated space where, um, what you're trying to do is model every atom and the quantum effects of every atom. And when you can do that successfully, it'll really enhance discovery. We're all also dealing with the effects of global warming, and we need these more powerful computers to help us in terms of climate modeling. Uh, certainly, uh, I think, uh, it will also help us in the investment field leveraging the power of ai. So these are the challenges that we're society facing today becoming increasingly, exponentially difficult. And I, I hopeful that, uh, we as, uh, human ingenuity can keep up with this.

Speaker 1:

Wow, it's really, it is amazing. It's, uh, the things, even the things we can do today like a smartphone, uh, were would've seemed like science fiction to people 50 years ago, uh, when I was a child, uh, when the idea of a handheld device where you could speak to people and see them and look up any piece of information pretty much, uh, in the world, uh, it, it, it was science fiction then. And so it's sort of staggering to think of what's going to happen over the next 50 years as we continue to double the Moore's Law keeps going, and what we'll be able to do on the back half of the chessboard, the, the things that do in fact seem like science fiction to us today, that that could well be, uh, reality within the next few decades. We've been talking about, uh, semiconductors. My guest has been Matthew Chan, uh, Matthew, thanks for joining us. Thank you so much. And if you enjoyed this podcast, please uh, leave us a good review, uh, wherever you get this podcast from, and we'll talk to you again soon.

Speaker 2:

Remember to subscribe to actively speaking on Apple Podcast, Spotify, or Google Play. You can find all of our previous episodes and additional content on our website, www.eipny.com.

Speaker 4:

The information contained in this podcast is distributed for informational purposes only, and should not be considered investment advice or recommendation of any particular security strategy or investment. Product. Information contained herein has been obtained from sources believed to be reliable but not guaranteed. The information contained in this podcast is accurate as of the date submitted, but is subject to change any performance. Information referenced in this podcast represents past performance and is not indicative of future returns. Any projections, targets, or estimates in this podcast are forward-looking statements and are based on epic's research, analysis, and assumptions made by Epic. There can be no assurances that such projections, targets or estimates will occur, and the actual results may materially be different. Other events which were not taken into account in formulating such projections, targets or estimates may occur and may significantly affect the returns or performance of any accounts and or funds managed by Epic. To the extent this podcast contains information about specific companies or securities, including whether they are profitable or not, they are being provided as a means of illustrating our investment thesis. Past references to specific companies or securities are not a complete list of securities selected for clients and not all securities selected for clients in the past year were profitable.