Surge in Semiconductors is the Wave of the Future
The semiconductor space is a constant battle against commoditization.
Editor's note: This article first appeared in the Q2 2021 issue of Morningstar magazine. Click here to subscribe.
During the pandemic, the semiconductor industry has secured its position as an integral part of daily life and industrial activity—and vital to U.S. economic interests. Surging demand has outpaced supply, and lawmakers are responding. At the end of 2020, Congress passed the CHIPS for America Act providing for investment in U.S. semiconductor manufacturing and design, and in February, President Joe Biden signed an executive order calling for a review of supply chains to increase production.
Technology strategist Abhinav Davuluri covers some of the biggest names in the industry for Morningstar Research Services. He says that consolidation and an expanding end market leave industry leaders well positioned for long-term growth. We spoke on Feb. 25, and the conversation reflects performance and valuations as of that date.
Laura Lallos: A year into the pandemic, how has the landscape for the semiconductor industry changed?
Abhinav Davuluri: A lot of the trends that we’ve seen impact the semiconductor space in recent years—artificial intelligence, 5G, cloud computing—have been accelerated during the pandemic because of everyone working and learning from home. These were things that were going to happen sooner or later. The shift to the cloud has been ongoing for the last couple years, as fewer businesses and enterprises control their own data centers and outsource that to major cloud vendors like Microsoft MSFT and Amazon.com (AMZN) and Google (GOOGL). But what may have taken five to 10 years was condensed into one year. As a result, many firms in the semiconductor space did extremely well.
Lallos: Now there’s a shortage of chips. What led to this supply/demand imbalance? How might government intervention affect the industry?
Davuluri: When everyone’s at home, not only do you see a massive increase in demand for laptops for education or work, but also for gaming and other consumer electronics, such as smart speakers or other smart devices. The overall ocean of demand has risen quite a bit. Then you have the supply-chain disruptions that occurred during the outset of the pandemic, stemming from initial lower demand from automakers, for example.
Automakers and parts suppliers for automakers were looking to keep a lean inventory level and expecting to be in this situation for quite some time. Now, they’re feeling the brunt of this shortage. They have very long lead times, and while their parts can be quite complex, they are not as expensive as, say, a chip for an iPhone. So, they are a lower priority at some of the manufacturing facilities for chips. Then automotive demand snapped back sooner than anticipated, as people wanted to drive as opposed to fly. When that happened, automakers were caught flat-footed and unable to get enough chip supply to meet that production demand. That’s why they’re in dire straits right now.
The automotive side is feeling the most heat, but the shortage is impacting every aspect of the semiconductor supply chain and every end market. The executive order sheds a light on the fact that the U.S. has outsourced a lot of this manufacturing. Much of it happens in Asia. There are still some domestic champions like Intel INTC and Micron MU, but for the most part, they make the chips that they sell, as opposed to making third-party customers’ chips. Most chip designers have outsourced the manufacturing to Taiwan Semiconductor Manufacturing (TSM) . All the Apple (AAPL) iPhone chips are made there. PC graphics chips for Nvidia (NVDA) and Advanced Micro Devices AMD are made there. Qualcomm QCOM smartphone chips are made there. The U.S. doesn’t really have a domestic foundry to make those chips.
The Biden executive order should ideally get the ball rolling for the U.S. to have more domestic manufacturing going forward, not only from a security standpoint, but also for diversification of the supply chain. There’s a kind of funny adage in the semiconductor space that building a fab—a manufacturing facility—is kind of like playing Russian roulette, except you don’t know if you died for four or five years. There are extremely long lead times to invest in these $20 billion facilities and get them running. The government will assess the current supply-chain situation and see where there are opportunities to either reduce reliance on foreign firms or build up the domestic fabs as necessary.
Lallos: How does China fit in here? Is it more of a threat or an opportunity?
Davuluri: The U.S. and the companies based here have made a practical decision to outsource. That wasn’t China’s doing. China’s taken advantage of our outsourcing and built lower-cost production facilities for manufacturing devices, like the iPhone, and, increasingly, built the chips themselves. But China’s still far behind the most advanced chip production players in Taiwan and South Korea, and even here in the U.S.
The U.S. is trying to limit China’s advances from a technological standpoint. The Trump administration was very adamant on trying to limit China’s edge in 5G, with the Huawei ban, and in artificial intelligence, via bans on U.S. chip companies selling some equipment and chips to Chinese firms. I think that for the U.S. to remain in the leadership position, it does need to invest in its own ecosystem. China is still going to have a big impact in the supply chain, but COVID has helped companies realize that they can’t have all their eggs in one basket.
But then, China is also a huge demand driver. For example, that’s where a lot of the iPhone growth has happened in the last couple years. China’s a very important end market. For the U.S., it’s a delicate balance of playing hard and tough with China while also making sure that U.S. companies can still sell into there.
Lallos: The industry comprises three different kinds of companies: designers, chipmakers, and the companies that provide equipment to the chipmakers. Is one of these areas better for building and maintaining moats?
Davuluri: The equipment side. There’s been consolidation in all of these particular verticals, but the equipment side has seen a ton of consolidation, and that speaks to how difficult it is to make the tools that actually make the chips. The five major equipment firms I cover account for about 70% of total equipment spending, and that number is likely to keep increasing.
Applied Materials (AMAT), Lam Research (LRCX), KLA (KLAC), ASML (ASML), and Tokyo Electron 8035:JP span the overall manufacturing process and sell basically all the tools to make those chips. They rely on existing relationships with advanced chipmakers. They work very closely with their customers to identify problems, and that creates a positive feedback loop where they can take what they learned from a previous generation and then implement it in new products. That makes it difficult for startups or smaller firms to break in. The equipment side of things is probably one of the most difficult for China to replicate, and it’s probably going to be at least five years, if not longer, before they’re able to be competitive at the leading edge. As for the chipmakers and the designers, there was a time when most companies did both: physically making the chips that they designed. But manufacturing chips has become expensive; you need to have scale and volume of chips to make it practical from an economic standpoint. The leader here is clearly TSMC, which has leapfrogged Intel in recent generations. They make almost every single smartphone chip in the world—about 70% to 80% of that market—especially at the leading edge. That is a lot of volume that has helped them fuel stronger demand, which allows them to keep investing, and that virtuous cycle has benefited them.
Intel is the U.S. leader, but they have had some manufacturing challenges in recent years; they have a wide moat but a negative moat trend. Samsung 005930:JP is in the middle. They are similar to TSMC, though not quite as leading-edge in terms of technology, but they also make a lot of chips for themselves, like Intel. Those are pretty much the only three firms that can do leading-edge manufacturing.
The chip designers have a lower barrier to entry and are the least moatworthy of the three. Companies like Apple and Tesla TSLA and even Microsoft and Amazon have gotten comfortable with hiring their own chip designers.
Lallos: Do you see Intel’s downward trend as inevitable?
Davuluri: What kept Intel ahead for so long was Moore’s law. Every two years, they’d increase density and increase performance, like clockwork. As things have gotten smaller and smaller, it’s becoming harder and harder to keep eking out those enhancements. Diminishing returns have made it more cost-prohibitive.
Intel had a dominant market share position in both PCs and data center chips, and in hindsight, they rested on their leadership edge. In recent years, AMD has been very competitive with their chip designs, and they’ve been getting them made at TSMC. Intel’s had some manufacturing delays, which has allowed the competition to catch up.
Meanwhile, the PC market has been in a slow decline over the past decade, 2020 notwithstanding. Last year was the first big growth year in about a decade because of the pandemic, but overall, the computer environment has shifted to smartphones and tablets. And those are two areas that Intel was unable to break into.
Intel did hire a new CEO, Pat Gelsinger, recently, and that has sparked some optimism in the market. Intel is contemplating a mix of more outsourcing and some internal manufacturing, and they may be able to find a better balance between those two. As we get into new markets such as artificial intelligence and 5G and such, perhaps Intel can grow beyond the PC and sell more chips that aren’t dedicated to that diminishing market.
There are a lot of headwinds, and they’ve been reflected in the stock price. With the new leadership and some of the new strategic initiatives that they’re taking, the next couple years will be indicative of whether or not they’re going to be successful.
Lallos: You’ve noted that the fabrication equipment industry is becoming less cyclical. What’s behind this shift?
Davuluri: Moore’s law is ending as we hit the limits of what we can physically shrink, so the cost of designing and making chips is increasing massively. TSMC, the biggest foundry, has increased their 2021 capex budget from about $17 billion or so last year to a range of $25 billion to $28 billion for 2021.
That is a direct benefit for the equipment firms. Investors looking to get exposure to trends like AI and 5G and cloud computing don’t necessarily have to worry who will be the winners and losers. These equipment firms are involved in making every single one of those chips. Picks and shovels for gold miners is a good analogy.
I think the industry still will be cyclical, at least from a quarter-to-quarter standpoint, but if we zoom out and think about a five- to 10-year timeline, the downcycles may be less severe and shorter in duration.
That’s a side effect not only of the consolidation on the equipment side that I mentioned but also among the customers themselves. There are fewer chipmakers, and they’re typically bigger because they’ve acquired some of their smaller competitors. And the chipmakers’ end markets are a lot more diversified. TSMC, for example, is not just making smartphone chips, they’re also making data center and automotive and PC chips. If there is one end market that’s ebbing a little bit, then they can transfer some of that capacity to make other chips in a market that’s growing.
The equipment side of things is dependent on capex, and that can be weaker during times of GDP contraction and such. But the leading-edge players like TSMC or Intel increasingly need to keep spending very high amounts of capex even in years where the economy is weaker. Apple still wants to put out a new iPhone regardless of the economic climate.
Lallos: Why has extreme ultraviolet lithography become a key technology?
Davuluri: As we hit this wall with Moore’s law, the industry is looking at other ways to increase performance and efficiency. Photolithography is one of the key steps in fabricating chips. The lithography tool shines a light through a photomask to imprint a pattern of the circuit design onto the wafer. The analogy I like to use is a paintbrush: More-advanced lithography tools are making the brush itself much finer so you can paint smaller and smaller feature sizes. EUV, or extreme ultraviolet photolithography, is the next advancement. The light with extreme ultraviolet is a much finer wavelength.
ASML is the dominant player here with 80%-plus market share. This is a difficult, complex technology. These EUV tools are on the order of $100 million to $150 million each, versus the prior generation that cost half that. Some of the major customers like Intel, TSMC, and Samsung coinvested in ASML’s R&D to help spur this technology on. After a lot of delays, it’s finally ready for prime time. TSMC and Samsung have begun using it for some of their processes. Intel should be using the technology for their 7-nanometer process that will come out likely in 2023.
ASML sits in a very comfortable position in the broader equipment space, and their stock definitely gets a pretty solid premium. Our concern from a valuation standpoint with ASML is what level of growth will their EUV tools actually have? Will it be as high as implied in the current stock price? This is a very expensive piece of equipment, and I think chipmakers will try to limit its use as much as practical, to remain pragmatic from an economic standpoint. There’s a longer-term existential risk that there will be a shift in the manufacturing spending from ASML to competitors. But at least in the next decade, their tools will be at the core of complexity and where some of the most value is added in the equipment space.
Lallos: How are the other equipment makers on your coverage list positioned going forward?
Davuluri: Of the five that I mentioned, KLA, Lam Research, ASML, and Applied Materials all have wide moats. I’d say that you really can’t go wrong with any of these, depending on valuation.
KLA specifically is one of our favorite names because what they sell is diagnostic and control tools. These are used to verify that a chip was processed correctly after it runs through, say, a lithography tool from ASML or an etch tool from Lam. As parts are getting smaller and defects potentially becoming harder to identify, KLA’s tools are critical to making sure that chips are functioning correctly or yields are high. They directly benefit from this rising tide of complexity.
Similar to how ASML dominates the lithography market, KLA dominates the process diagnostic and control market, with about 60% market share. They have industry-leading profit margins, really strong cash flow, and better operating margins relative to their peers as well. They are critical in the transition from technology to technology, while some of their peers are more dependent on capacity expansion, where customers are building out new fabs and adding a ton of equipment. KLA is kind of a pure-play investment on increasing complexity and more immune to short-term economic shocks. Let’s say there’s a sharp decrease in demand, and TSMC decides to cut capex by 15% to 20%. They’re likely to buy fewer tools for their actual manufacturing facilities, but they’ll still be working in the R&D setting and ramping up new technologies that will be coming out in future years. KLA is instrumental in helping them achieve that. That’s why we view them very favorably relative to the overall space.
Lam Research has shown some novel technologies in recent years in the areas that they focus on, and they’ve gained quite a bit of share. In the specific markets that Lam plays in—the etch and deposition markets—there is a lot of competition, with Applied Materials and Tokyo Electron competing in similar subsegments. But also, they’ve historically been more dominant on the memory side of things, with customers like Micron and Samsung. Meanwhile, they’ve been increasing their logic and foundry capabilities, and we expect them to gain more share there as well, with customers like TSMC and Intel. This is a shorter-term call on their competitiveness being a little bit better than some of their peers’—at the end of the day, four of these five firms are rated wide moat.
Lallos: You recently increased your fair value estimates for these major suppliers, as well as narrow-moat Tokyo Electron.
Davuluri: Yes. In addition to short-term demand being a lot stronger across the board, the 2021 expectations for wafer fab equipment were much higher than we’d anticipated.
Lallos: We’ve touched on some of the trends driving semiconductor demand. What are the overarching themes in this space?
Davuluri: The semiconductor space is a constant battle against commoditization. You want to keep innovating, keep pushing the envelope with performance or efficiency to generate a premium or a positive margin. If you stagnate, then competition catches up, and it becomes a very price-sensitive battle.
We’re moving from an environment where PCs and smartphones are the primary end market for chips to one where everything we use is. Everything is smart and intelligent and using some sort of AI. There’s computer vision, which allows a car to realize that the car in front of it has stopped suddenly so it needs to brake. Natural language processing enables our smart assistants to recognize what we’re saying and answer questions. Then there are more sophisticated AI applications like self-driving, or recommendation engines, or any sort of healthcare application. Pretty much every industry is going to be touched in some way, shape, or form by increased connectivity and increased intelligence.
There’s also all this data that’s been collected and is just waiting to be monetized. There are a lot of companies with data sitting in the cloud or in their data centers, but they don’t necessarily have a cost-effective way to monetize it. The AI trend is allowing companies to unlock the value of all the data that they hold.
The winners in the semiconductor space are going to be the ones that help those customers figure out solutions, whether in connectivity or 5G or Big Data and other AI applications. This is what is going to push the overall semiconductor and computer industry beyond personal devices like PCs and smartphones.
Lallos: What risks should investors be prepared for?
Davuluri: The COVID-related demand trends have been strong, and there’s always risk that as things start opening up, some areas may be negatively impacted, which could cause a short-term correction. Or there may be a glut of inventory after we’ve been dealing with all these shortages, and that may overcorrect. But if we look at a five- or 10-year horizon for the overall semiconductor space, I think all the trends that we discussed are likely to be up and to the right.
Laura Lallos is managing editor of Morningstar magazine.
Laura Lallos has a position in the following securities mentioned above: AAPL, AMZN, GOOGL. Find out about Morningstar’s editorial policies.