In this episode of Markets in Focus

Recent supply chain issues have shed light on the importance of semiconductors in everyday life. Tariq Siddiqi, Senior Research Analyst at Eagle Asset Management joins Matt Orton, CFA, Director and Portfolio Specialist at Carillon Tower Advisers, to discuss how the chip shortage could impact the future of the semiconductor industry.

Subscribe where you listen to podcasts

Transcript

Matt Orton:
Whether we know it or not, semiconductors are the foundation of the advanced technologies that we all rely on, from cell phones, laptops, and washing machines to refrigerators, cars and airplanes and we probably didn't pay too much attention to how these chips are produced until the COVID pandemic where the global supply chain faced significant disruptions that are still impacting chip availability today. Global semiconductor companies have been among the best performing industries within equity markets over the past year, but there are some meaningful differences between the types of companies and their role in the supply chain that are critical for investors to understand. Today, I'm joined by Tariq Siddiqi, senior research analyst at Eagle Asset Management, for a deep dive into the semiconductor supply chain and a look at how they think about evaluating investment opportunities.

This is Markets in Focus from Carillon Tower Advisers. I'm your host, Matt Orton. Join me and my colleagues as we discuss the latest trends and developments driving the markets. Visit us at marketsinfocuspodcast.com for additional episodes and insights.

Tariq, thanks for being here today.

Tariq Siddiqi:
Great to be here.

Matt Orton:
Excellent, and so before we dive into some of the opportunities, why don't we start with a quick review of the basics? What companies are involved in the manufacturing of chips, meaning who designs them, where and how are they made, what companies are needed to make them and what the end users are? I know that's a lot to unpack, so maybe let's just start with the design and then we'll go from there.

Tariq Siddiqi:
So this is a big question, right? As you just alluded to, there's a lot to go through here. It can really be a college-level course, and books are written about just some of the sub parts of the system, but I'll try to give a quick 10,000 foot view of how we go from a design to raw wafer to the finished chip that goes into the various end products like the cell phones, laptops, cars, et cetera that you mentioned.

The process starts with basically a design. If you're a big semiconductor company, or now even the mega cap, internet hyperscale tech companies, they build their own chips now. They all started by using design software to actually lay out the chip, the blueprint of the chip. This software is called EDA, electronic design automation software, and there are only a few dominant players here.

These are incredibly complicated designs in very tight spaces and so it's a very complex process and requires a lot of money. Just as a quick side note, let's say you and I are building a new chip. For our leading edge chip at the latest note of five nanometers, we would need to come up with $500 million just to design and develop a chip. This is before an actual physical chip even gets produced, tested, or anything like that. And we'd be spending half of that in the software design. The other half would be spent on specialized hardware to design that shape, and of course, to run simulations and perhaps most importantly on very capable, smart R&D engineers. Once the design is set and simulated and the bugs and the kinks are worked out, we would either send it off to our own factory, known as a semi fab, that can produce this chip, or more likely nowadays to an outsourced factory called a semi foundry that manufacturers the actual chip. The big behemoth in semi foundry is in Taiwan, but there are a few other players in Korea, China, and the US.

Matt Orton:
So I just want to say it's pretty incredible that it can cost upwards of $500 million just to design and develop a chip. I think that's some really important perspective, and we always hear about the upfront expense for developing new drugs, but rarely for developing new chips.

Okay. So maybe you can walk us through the actual production of a chip now.

Tariq Siddiqi:
Yeah, absolutely. So all the chips that we're talking about are made on a silicon wafer, or a substrate. The raw silicon comes from processed sand, but it's safe to pick up our discussion here from the nice, large, round, perfectly clean, very thin dark wafers that I'm sure our listeners have seen photos of. Currently, these wafers are the size of 300 millimeters, or just under 12 inches. The two main producers are companies based in Japan. From here, the wafers are brought under lithography, a specialized late, deposition and etching steps happen. So the design, the blueprint that we made in that EDA software earlier, is effectively printed onto that raw wafer. This is called the front-end of the chip production process. So, we shine those special lights, create a cut on the wafer, deposit some material, etch and remove some other material, and again, shine more lights.

It's a complex process with hundreds of steps, but each time you're likely using a lithography machine from a single Dutch company. As a side note, this litho machine is super expensive, nearly $100 million for one tool. It is the most expensive piece of machinery, and effectively, this Dutch company is the only one that can provide this machine. There's a smaller secondary vendor out in Japan, but for all intents and purposes, you're going to this Dutch company if you want to be at the leading edge of stuff. And then you have the deposition of materials, etching out those materials, using machines, again from a handful of companies, based out of the United States and Japan. Via these processes, entire layers are being deposited, etched away and tiny, tiny 3D structure city lights, there's almost a city skyscraper that you're developing on top of the wafer substrate.

Now, all the while these processes are happening, you're using chemicals and gases that are super pure and only meant to be used in semiconductor processing. As you can imagine, even the smallest impurity will ruin the design that we just spent $500 million developing. So you have chemical companies and gas companies from around the world providing many different specialized materials and gases to keep the process going. Once the whole wafer has a design pattern on there and all that excess material has been removed, you cover all the design with more material to make sure it's all solid and you design the interconnects. Interconnects are sort of like if you think about little highways that connect different parts of the chip. Once that's done, you've got basically depending upon the size of the chip, maybe hundreds of chips ready to be tested on that wafer. Now, once you do the testing is done with specialized instruments, you go into cutting and packaging.

This is called the back-end of the process. In the back-end, the individual chips are cut from the silicon substrate, the wafer packaged into ways that can be installed into a PC or a smartphone, and this stuff happens mostly in Malaysia and Thailand by a handful of companies. Once again, the fully packaged and ready chips are tested, and after that, they go off to an assembler or somebody that actually builds the end product and then the end product is shipped to distributors and ultimately to you and I. This whole manufacturing, testing, deliver process can take anywhere from three to six months post-initial EDA software design phase.

So yeah, that's a lot. It's a complex process across various companies, materials, geographies. The key takeaway that we have is its complexity. Lot of really, really complex stuff happens here.

Matt Orton:
Yeah, that is definitely a lot to take in. Incredibly complex and it's also abundantly clear how many companies, all based around the world, are involved in the process. I'm going to come back to the complex global supply chain, but first, I want to build on the great background you provided, Tariq, and have you talk about the end markets. Where exactly are these chips being used that were just produced?

Tariq Siddiqi:
Yeah, absolutely. So the total semiconductor market is about $450 to $500 billion and the largest end market for semis is in PCs and servers, what is called the computing side of things. This is about $160 to $170 billion market, it's the largest market, and the server portion within that has been growing for quite some time as we talked about those hyperscalers and data center companies doing very well. This has been growing for some time and likely will continue to grow for some time. Now you've got after that mobile phone semis, or what generally is called communications, which will include things like the 5G phones, the base stations, anything to communicate with. This is about $140 to $150 billion market where semiconductor content is going up even if your phone units don't grow so much, the complexity continues to go up, right? So you have more chips being used in there.

After that, it drops down to the consumer market. This is about a $60 billion market for usage in various things within your home, game consoles, tablets, headphones, smartwatches, et cetera. Autos comes in next at about $40 to $50 billion and where they're likely to be a lot of growth as we see electrification of the automobile. And after that, it starts to get more hodgepodge like various industrial applications, medical, aerospace, defense, et cetera.

Matt Orton:
Incredibly helpful. And as I let the background sink in that you just provided us, I want to come back to my last comment where I mentioned all of the global players involved through the design and production of the chips. Given this value chain, it's evident how COVID could wreak havoc on production and contribute to the supply shortage that we're facing today. But there have also been some compounding issues and shifts in demand. So Tariq, maybe you can provide some color on supply and demand pre-COVID and then transition to what happened once the pandemic actually hit.

Tariq Siddiqi:
Yeah, so the pandemic was obviously something that nobody saw coming. Unlike a lot of other potential contingencies that all of us and the companies that we invest in plan for, this sort of came out of nowhere and it had a very strange impact on different parts of the world. Specifically for the tech hardware and the semiconductor space, it led to both a supply crunch and a demand shock. Most companies plan for supply issues. If you think about the Japanese earthquake a few years ago, Thailand floods, and even when they plan for demand issue, it's more on the downside like if you have a fire in a factory or there's sanctions by US government on Chinese affiliated companies, but most companies don't really plan for sudden 20%, 30%, 50% rise in demand. Most plan for slow and steady growth in semiconductor demand at one-and-a-half to two times GDP.

Well, with everyone all of a sudden needing a new PC to work from home, new iPads, consoles, et cetera, and the stress that put onto the internet connections and the associated data center spending, it was certainly once in a lifetime event that further got complicated by some of the China trade war issues that had been brewing for some time and some other natural disasters that only added fuel to the fire.

Now taking a step back, the issue is that the semiconductor industry is a maturing industry. It had been slowing from a hyper-growth in the 1990s to healthy growth in the 2000s to a much more stable growth now. As one would expect when an industry starts to mature, the capital spending starts to decline and the cash flow associated starts to rise, or at least it should because since the industry is maturing, it doesn't help to aggressively build new plants.

So, the industry has been in mode of consolidating, manufacturing, from 20, 25 main players 20 years ago down to about three or four major players now. This also made sense from the fact that as we discussed, the newest chips require the best equipment and that costs a ton of money. Just as a reference, the latest five nanometer node fab that we're talking about, a leading edge is a $12, $15, $17 billion fab. That's just one factory. So, it didn't make sense for many of the semi companies continue to spend on building their own operations when an outsourced foundry can do the same thing and then recoup their costs over several different customers.

Matt Orton:
Right. No, it makes sense when you put it in that perspective and especially given the scale of those numbers. I don't think a lot of us realize how big those numbers are, just like how much it costs to actually develop one single chip. So with this background, how long does it take before this gets fixed? When do things actually start to come back into balance?

Tariq Siddiqi:
Yeah. So in terms of a fix, this will take time. In some areas, there's some flexibility and some excess capacity. So with some new equipment and labor, companies have been able to meet that crunch. However, for new capacity and to truly address that demand, it will take time. The time it takes on breaking new ground on a shell of a fab, a shell of a factory, to actually putting equipment in there to making sure it all works properly. All those steps that I discussed earlier, they all have to work perfectly well. And then having actual impact on supply will take anywhere from 18 to 24 months, maybe even as much as 36 months and some very complex products. This is not an overnight situation here.

Matt Orton:
Yeah. So that's really, really good color and I do have one follow-up to that. So based on that color, it seems like the issues are widespread in impact chips broadly speaking. Why then do we only hear about certain shortages for autos or appliances? I tend not to hear too much about the impact of iPhones or smart TVs. Is that coming or have phone and computer makers better insulated themselves?

Tariq Siddiqi:
Yeah, that's a really good question. So, the semi industry generally has focused spending money on the latest and the greatest, which gives them the best growth or the best margins. A server chip that comes in for a data center is a $1000 part and that can get the gross margins in the 65%, 70%. And the server data center side, as we talked about, is one of the best growth area so people are focused there. A chip that goes into your car can be a $5 part, a $10 part, and makes about 30 to 40% gross margins, and autos in general have been a somewhat slow growth area. So the companies are focused on chips for automotive just haven't spent as much. Chips for automotive, or home appliances, generally tend to be a lot more lagging edge, older technology. They just don't need the whiz-bang computing that your iPhone 12 needs.

When we talk to companies that provide chips in this lagging edge area, when we discussed companies in 2017, 2018, 2019, they all said, well, we continue to gain more efficiencies by reducing inventory, running more and more on just-in-time type of model and running our fabs at the highest possible utilization rates to maximize the margins, and that actually made sense at the time. Then, the pandemic hit and the auto sales absolutely came to a halt in March, April, May of 2020, while the auto producers cut manufacturing and laid off people.

Now, the chip makers that sell to these automakers, well, they diverted. They're like, if you're not buying, I'm going to go sell to somebody who is still buying, the consumer electronics guys, the iPhone guys, et cetera. So when the snap back for auto sales that sort of came in second half of the year into this year, there was no inventory left for them. So what happens when you have super low inventories, high utilization rates, lack of investment for years, but then see a demand crunch as people rush out to buy cars and TVs, you have super shortages.

Matt Orton:
Yeah. That's very, very well put and really useful context. Also very easy to understand. So, I think it maybe makes sense to now look at it through an investment lens as that's what you do every single day. And there've definitely been some clear winners from the current semi shortage. Chip manufacturers have had a record year, semi cap equipment companies had been posting incredibly strong numbers as the environment seems supportive, given critical technological advances and increased complexity. Can this continue? Are valuations too stretched at this point? What are your thoughts?

Tariq Siddiqi:
Yeah, I think this is the multi-trillion dollar question that the investment community and the semi industry in general is grappling with, what do we do now? Let's talk about four major things that we do know. One is demand will slow, no doubt about it. As we lap the comparisons from a year ago and as the world slowly and unevenly, due to the various variants and so forth lockdowns, as they slowly open up, there'll absolutely be a demand slowdown. PC growth will slow down. iPads and new TVs that people bought, smartphone upgrades: All these end demand products will slow, we know that.

Number two, capex [capital expenditure] has been up nicely, about 30% or so expected for 2021. That has helped the capacity expansion, but it will take time for supply and demand to come in balance again. As we discussed early in some areas of semi-industrial, it will likely be a one to two quarter phenomenon, in some areas it will be a one to two year process.

Number three, healthy long-term demand drivers. 5G wireless technology, autonomous driving, Internet of Things, artificial intelligence, the explosion of data. All these drivers are going to push up the demand for semi chips. Just as a side note here, as bleak as it looks right now, I think one of the best demand growth drivers is in autos. As the auto industry goes from a traditional gas engine to an electric one, number of semi chips per car goes up significantly. Although this number bounces around based on the specs of the vehicle, a traditional car has about $300 to $400 worth of chips. An electric car with motors and battery will have something like three to five times as many chips. And once you start talking about fully autonomous ADAS driving with multiple cameras, sensors, LiDARs [light detection and ranging], and everything working well together, the chip content can be as much as 10 times. So I firmly believe that this is a very nice growth driver for the next decade or so for the semi industry.

Lastly, continuing complexity. The leading edge chips will only continue to get more complex. There's no doubt in my mind that the world will look for even faster chips, even more storage, and even faster internet connectivity downloads in gaming. These trends will only continue, and for all that, you need smaller and more complex chips with 3D designs and nanometer wavelength physics that are hard to solve. Related to that capital spending point earlier, much of the money being spent right now, it's not even helping capacity. It's just solving these thorny physics issues that will only get thornier as we go down.

So how do you invest through that? You move from a macro-themed investment style, basically saying, okay, semis benefit from work from home buckets, semis get hurt from a reopening of the economy, to a much more company-specific thought process. Which companies are going to benefit from new innovative products that are taking market share, which companies are going to benefit even if the capex cycle slows down, which companies are more exposed to the long-term demand drivers that we just talked about, which companies are increasingly becoming more important in the very strange physics that semiconductor industry has already been plowing through for years. And now it's only getting stranger as we reach three nanometers, two nanometers, one nanometer in quantum levels.

This is a very individual company-specific issues are going to be a lot more important rather than the diversified generalist semiconductor investment thought process. And, we have to keep looking for yellow flags, signs of overheating. Are the customers of these chip companies or the equipment companies over-ordering? Are they double ordering, thinking that by giving super high orders, they will be able to secure more immediate orders to meet their outsized demand for their own customers.

How the lead times are changing for when an order is given today to when it can actually be delivered is a very important metric to monitor and we have to be cognizant at some point, if capex doesn't slow down, there will be severe headache as a supply demand will likely go on the other side and overshoot to the wrong side. As we sit here in the middle of 2021, however, we have some time before that but we cannot be complacent. That balance between demand drivers and the yellow flags and the capex spending is what we have to understand and navigate through.

Matt Orton:
Yeah, I think that's some excellent perspective for all of our listeners to listen to as you look at such a large universe of companies in potential investment opportunities. Now, let's transition to thinking about domestic supply building and the kinds of things that we've heard the current administration talking about regarding US-based semiconductor manufacturing so we're less reliant on the Asian companies. Is this something that's within the realm of possibility? What are your thoughts on that?

Tariq Siddiqi:
Yeah, so that's a big issue, but that one we won't solve for quite some time. As we discussed, it takes a long time for a fab to actually get built and come online. There are a couple of things here. One, we absolutely should be building US-based manufacturing. Self-sufficiency is very important. Chips are central to so much of our economy and will only continue to get more relevant. You don't want to be in a situation where China threatens Taiwan or Korea and all of a sudden the major fabs in those two latter countries cannot provide important chips. So US companies, including defense companies. That's a big problem. So having some semi fabs in New York, Texas, or Arizona is very important and that's what the legislation that President Biden has put in with the CHIPS for America Act is a really strong, good step in the right direction.

At the same time, it's not as if America doesn't have many ways to counter the issues today. As we discussed above, quite a few very critical companies are based in the US and its allies, like that Dutch companies that we just talked about earlier. If China wants to be aggressive with Taiwan and try to cut off the supply, yes, it creates a new kind of headache, but then the entire Chinese semiconductor industry will have to halt because the US and European companies will stop providing those critical end materials with them.

So both sides know this balance. However, what I do worry about is that when we build things not in relation to actual supply and demand of the industry, it will lead to imbalances. All major manufacturers have announced multi-billion dollar major new fabs we built in the US either for the new foundry business or general capex and that will certainly lead to strong growth in the equipment companies in the near term. But when all those fabs do come online and ramp up, there will likely global excess supply of chips and that most certainly will lead to issues, and a major downcycle for those equipment companies and will create a lot of pain.

Anyways, we have time before that, but something certainly to keep in mind, the balance, self-sufficiency, and general supply demand dynamics. How we go about adding capacity without disrupting the supply demand while making sure the critical components cannot be denied to American companies and consumers is a really delicate balance.

Matt Orton:
Great, thanks Tariq. That is some very, very good perspective and I think it's important to look at the risks that you highlighted, both with respect to building out supply chain in the US and some of the risks that you have along with the rest of the semiconductor industry. So with that, that's really all the time that we have this afternoon but this has been an incredibly insightful discussion and hopefully our listeners felt the same way. Thank you very much for tuning in. Thank you Tariq for your time and to our listeners, until next time. Take care.

Thanks for listening to Markets in Focus from Carillon Tower Advisers. Please find additional episodes and market insights at marketsinfocuspodcast.com. You can also subscribe to our podcast on Apple Podcasts, Spotify, or your favorite podcast app. Until next time, I'm Matt Orton.