AI Tipping Point: A Deep Dive into Semiconductors
Which semiconductor companies will be long-term winners from the AI boom? What risks do supply chains and energy usage bring to the sector? This episode of the All Angles podcast features Genevieve Gilroy, sector lead for semiconductors and consumer staples at MFS. Genevieve describes the challenges and opportunities in the semiconductor space and explains why valuations matter when navigating the AI hype cycle.

Vish Hindocha: Hello, and welcome to another episode of the All Angles Podcast. Today we're delighted to welcome Genevieve Gilroy onto the podcast. She's the sector lead for consumer staples and also covers the semiconductor industry, which is where we'll spend most of our time today. In this conversation, Genevieve talks about how to stay calm during the hype cycle, how to think through risk reward trade-off, the types of questions that she asks management as she passes through a regime shift in the AI and semiconductor space, and she has a wonderful book recommendation for us towards the end. AI is a space that is front of mind for so many of our clients, asset allocators, portfolio managers and just about everyone in the world at the moment, so this is a really timely and relevant conversation and I hope you take something from it. Welcome, Genevieve.

Genevieve Gilroy: Thank you.

Vish Hindocha: A tradition on the show is just to learn a little bit more about you, so I'd love it if you could give us a bit of a potted history about yourself. How did you get here to do the things that you're doing now?

Genevieve Gilroy: Well, I actually have spent my entire career at MFS®. I started at MFS right out of Boston University in our Equity Research Associate Program. I worked on the consumer staples team, and it was really at MFS that I discovered a love for investing, with so many people that I worked with at the firm that were so inspirational in terms of finding how incredible this career opportunity could be. I was in that role for about three and a half years before moving into the analyst role, initially covering protein processors. So very value oriented, working a lot with our midcap team, our small-cap team, super cyclical businesses, and learned a ton about thinking through normalized earnings and a lot of different valuation approaches.

And then I eventually added on non-alcoholic beverages, so Coke and Pepsi, and again working much more with the core teams, the large-cap international teams. So really broadening the horizons. And then about three years after that, I joined the technology sector team and now cover all things semiconductor across large cap, small cap, across the software that is used to design semiconductors to the capital equipment that's used to manufacture the semiconductors and work really closely with our global team covering that space. And it's been of course a very exciting number of years for that space.

Vish Hindocha: That's amazing. So on the face of it, though, you've got consumer staples, beverages, non-beverages. You and I have discussed Coke and Pepsi, or plastics, and I learned a tremendous amount. And that feels like a whole podcast episode on itself, so we're plant a flag there to maybe invite you back again. And semiconductors, so they seem very different disciplines and yet you're talking about connected . . . I'm just curious, is there anything that you think distinguishes those spaces or anything consistent thread —philosophically — that you look at when you are looking at where to allocate capital, what wins, how you think about competitive positioning, any of those themes that translate across, or are they very, very different spaces and you have to think very differently and have a whole different set of mental models around them?

Genevieve Gilroy: A major difference as I think about consumer staple — and from a very high level as I think about investing behind semiconductors versus investing behind consumer staples — is just how we think about duration and confidence in duration. So as you think about consumer staples, that's a space that has been characterized over many, many years for advertising and very long periods of advertising leading to duration of the brands. And the value and the multiple that you put on that duration is very high. There's very high confidence and there's low cyclicality in those businesses and they've earned a high multiple for that reason. And so when I compare that to semiconductors, it's interesting because across technology — that's a space that's characterized by lack of duration. This is a space that's consistently at threat of disruption, and so the technology is constantly changing and also characterized by cyclicality. So unlike consumer staples, which are characterized by even in a year of a recession, we would generally expect maybe flattish revenue growth for consumer staples.

Across semiconductors, every three to four years, you could have booms and busts that are difficult to predict. And so that space, as we look across it, much more attractive growth long term, but also you're consistently dealing with the risk of — medium term — lack of confidence in fundamentals and much longer term lack of competence in fundamentals, just given the duration risk from the threat of technology disruption. So it's a really interesting contrast of groups to cover. I would say one area that has actually helped me, as I've thought about my semiconductor coverage, was really my first coverage that I ever had at MFS, which was covering the protein processors and the grain traders. And I worked a lot with our small-cap and midcap team to really think through how do I think about normalized earnings and how do I think about valuing cyclical businesses? And so, I think that's maybe an unexpected space where there's been some leveraged learnings in terms of how to think about valuation of groups that are not necessarily what you would intuitively think as similar.

Vish Hindocha: Yeah, that's great. That's so insightful. And maybe we'll come back to sort of other lessons learned through time in both spaces but, given you're talking about some of the characteristics of the cyclicality, the kind of constant disruption, I do want to sort of zoom in on the semiconductor space. So AI is so front of mind for so many people right now, with obviously the explosion after the large language models, ChatGPT in particular, and everyone's talking about AI. You can't avoid it. How is it playing out in your space? There's so many prevailing narratives around how people and investors in particular should think about this, whether there's going to be much more of a winner take all mindset or whether it's understanding who the picks and shovels are in the space and making sure you get exposure to different things. How are you thinking about it? How has the euphoria impacted your coverage in the way that you maybe think about risk and opportunity from here on out?

Genevieve Gilroy: Absolutely. And just thinking through how AI has levered through the entire tech space, I mean, of course it was completely unexpected when it came out and has driven an explosion in interest across cloud service providers, across enterprises, across consumers. And I think everyone is really figuring out the long-term implications from all of this. And so it's interesting as a global technology sector team, as we initially approached this explosion in interest, what we were really looking for is trying to dissect across different business models. Just exactly what you're speaking to, which is just how this would impact companies differently. And as we talk about software businesses, there were some questions around “is this disruptive to the software business model.” “Is this a positive for the software business model?” And there were a lot of questions, and I think there were questions across enterprises. “Just how is this going to flow through and impact their business models?”

And the one place that it was interesting where, as we looked at it, it was really only a positive with semiconductors. And so right now, of course, we're seeing that with NVIDIA, just the largest earnings inflection in terms of magnitude in the history of business, really. But there's many, many companies throughout my semiconductor space that are also benefiting from this explosion in training of large language models. So thinking through the networking chips that are going into these data centers, the memory that is going into these data centers and a whole host of other products that are being designed into the servers that are training these models. And also, even thinking through just all of this really is an extension of Moore's Law slowing. So as you think through why we're utilizing GPUs, it really is because they are an extension of bringing compute progress forward because Moore's law has slowed and is no longer giving us the performance impact that we've needed.

We're seeing compute fragment away from the Intel CPU into GPUs, into custom chips that many of these cloud service providers are talking about. We're looking through how those custom chips as a derivative of Moore's law slowing could be benefiting other companies. So thinking through this is a positive for the design software that is used to design semiconductors. If we have a greater amount of chips proliferating away from the CPU, then that's a positive for design software. And of course all of this is also a positive for the companies that manufacture semiconductors, so thinking through the foundries or even the capital equipment that is put into a foundry or put into a semiconductor manufacturing facility. And so, really a lot of the work has been as we compared semiconductors relative to certain other spaces, the confidence that this would be a long-term positive for these businesses is pretty high and looking through places where that opportunity may be long-term underappreciated.

Vish Hindocha: I want to get to what you think might be long-term underappreciated, but just I wonder if it's maybe a small step back. So you talk about this is a space that is used to disruption, it's kind of prone, privy to disruption, yet we've got sort of a regime shift happening within the space. Moore's law is slowing. You've seen some of these exponential hockey sticks around groups like NVIDIA or other spaces around it. And you mentioned the technology team. Is there any sort of change in the way your process . . . how do you think about that space in the middle of a regime change and inflection point? And again, we could talk about in other domains you can see that, maybe in health care or in the past in other sector teams as well. So I'm just kind of curious, just as a small step back, how you as a team might analyze risk and return differently, recognizing that the ground is shifting underneath your feet, right?

We're in the middle of this regime shift of how we might think about it. I want to get into what company management I talked to you about, but I'm sure they've all got a wonderful story about how they're playing it. And how do you decipher, as you're doing that due diligence, what is real and who's going to win in that space and what might need a bit more work in terms of competitive moat or a business plan? So any thoughts there in terms of what the team has been focused on in terms of risk and return or engagement or how you think about questioning corporate management around some of these themes?

Genevieve Gilroy: Yeah, absolutely. And I think you hit the nail on the head in terms of this space is moving at breakneck speeds. And if we were to plan out what is the ultimate market size for all of these chips that are supporting AI, if I look out five years and I think through what are all of the assumptions that go into what that would look like, it is incredible just thinking through from a unit's perspective, from a price perspective, from a market share perspective, from a gross margin perspective the range of outcomes across so many companies that I cover, particularly for NVIDIA, is wider than it ever has been in history. And so really what we're trying to do as a team is talk to as many companies as possible, talk to as many experts and talk to each other and really just think through range of outcomes.

How wide could that really be across all of these businesses? And respect within valuation that, just given the range of outcomes across every variable at this incredibly involving period of time is so wide. Respecting that within valuation and respecting the difference in ranging of outcomes across companies. And so thinking through, again, there are many businesses that are benefiting within semiconductors from AI, but one example is NVIDIA. But there's many, many questions around thinking through, right now all of this GPU spend going to NVIDIA has been driven by the training of large language models. Eventually, we're going to move to a place where the querying of the large language model is actually typing in your question to ChatGPT and having it send back an answer. All of that is what's going to drive semiconductor spend. And so we don't know at this stage just how large that will be relative to how many GPUs have been needed to train all these huge models.

And again, competition is entering, so AMD, Intel, even companies like Microsoft, OpenAI, are designing their own chips. So the confidence that we can have long term in market share, the range of outcomes is just very wide. And that has implications for prices and for gross margins. And so as I think about a company like NVIDIA, again, the range of outcomes is so wide relative to thinking through, for example, the opposite of that would be the software that is used to design semiconductors. So benefiting from companies like AWS, Microsoft, Meta designing their own chips and having that be a tailwind to long-term growth, but with much, much smaller range of outcomes. And so thinking through just different approaches to valuation for businesses where that range of outcomes is so high and really trying to stay anchored in that risk reward that we think through and comparing that across coverages across our global teams.

Vish Hindocha: That's really powerful to think about how you stay disciplined in this kind of moment of change is the most important thing, rather than abandoning all your models and just sort of going with the prevailing narrative. I guess maybe two variables that are moving around that also it can maybe exacerbate the range of outcomes, at least in the popular narrative. And Genevieve, tell me how you think about this, but one is sort of geopolitics. So lots of people are talking about deglobalization, about onshoring, about supply chain diversification, et cetera, et cetera. And about how we don't want fragility in the system by clustering either production manufacturer or shipping of semiconductors in one or two locations around the world.

Is that something that you'll see come through? You talked about how some of these companies are starting to insource their own chips and their own GPUs. Is that something that you're seeing, paying attention to? Again, is that people entering a new space where, presumably, harder for you to get some confidence around their competitive edge, but equally — hopefully — diversifying a potential risk over an extended time horizon that hopefully makes you feel better? So how do you square those things?

Genevieve Gilroy: I think risk around deglobalization has been one of the topics that we focused on the most as a global team over the last several years. And it is interesting across my space because I think semiconductors are the space where that has one of the biggest impacts across the market. And the reason for that is that it is among the most globalized industries to begin with as you think through every single step of the supply chain, starting with the semiconductor design companies, which are centralized in the United States. The capital equipment companies, which some are in the United States, some are in the Netherlands, some are in Japan. And you think through the companies manufacturing conductors, largely TSMC out of Taiwan, and all of the chemicals that are used in the semiconductor manufacturing process are from all around the world. And really, it's a space where so many small niches — because this space is so, so difficult — what really has been a more successful business model is a very focused R&D engine.

So focusing either on the design, focusing on the design of the manufacturing equipment, focusing on the process technology like TSMC does. These are businesses that are highly focused in whatever niche they play in the semiconductor industry and that's led to an incredibly diverse supply chain. And it's so interconnected with one another. It would be such a challenge for any one of these large businesses to be pulled out of the system because they are so interconnected. And so, really what the process has been like for us has been understanding and thinking through just the financial statements. Understanding if geopolitical tensions were to escalate or even remain status quo, just what does that mean from a unit space for certain companies, like the analog players like Texas Instruments and ADI?

These are companies who sell a considerable amount of their semiconductors to Chinese customers. So thinking through, is that a risk long term from a market share standpoint, from a unit share standpoint? If Chinese companies are increasingly designing in Chinese semiconductors over the American players and then thinking through manufacturing. So to what extent are companies manufacturing exclusively in one location or one geography? What does that look like from a cost perspective if they need to diversify? Does there exist enough capacity elsewhere in the world to appropriately diversify? And I think for some companies like the analog players versus leading edge semiconductors, it's very, very hard for a company like Apple to diversify away from a company like TSMC, given their chips are of an advanced technology that only TSMC can do.

Whereas some of the older chips can be more effectively diversified away. What are the capex implications of diversifying away from a concentrated manufacturing footprint? Are there capex implications for the company? Are there capex implications for the foundries? And would that be a net positive for the capital equipment companies that are selling to those foundries? And again, it all comes full circle from the capital equipment company is thinking through our Chinese capital equipment company is going to increasingly be taking share within China. So it's an incredibly intricate set of questions that we have that we're consistently talking to with every company. And I think it affects these businesses everywhere from a unit, a price, a margin perspective, a capex perspective and trying to compare across each of these businesses and really understand where we're exposed.

Vish Hindocha: That's such a powerful example when we think about on this show . . . sometimes we talk about stewardship and engagement and ask for examples. And what you've just described there, it sort of perfectly captures what many of our guests often talk about, which is the subtlety and the nuance. And to use your word, it's very intricate sometimes, a set of questions that you have to think about and appreciate that that company is responding in a particular type of context that you then have to take into account and view the bigger picture more end to end. Which is really, really interesting as you drew that sort of chain for us in our mind in terms of how things go from beginning to end of that. Another variable that sometimes I guess moves the needle and often gets discussed in the space is sort of energy or power.

Net zero commitments and climate change often . . . sort of front of mind for many investors and governments and regulators. Sometimes what gets brought up is the energy requirement for quantum computing, for crypto, for some of these other spaces where many of these semiconductors get used. Is that something that you think about in your analysis? Where does that feature in terms of materiality for you when you're looking at a semiconductor name, for example?

Genevieve Gilroy: Absolutely. I mean, it is really interesting because power is of the utmost importance to all of these companies. And thinking through all of the cloud service providers that are developing these large language models, all of these models are limited by power. They're power constrained, so power consumption to them is a sustainability focus, but it also is critical to their business model. And so, as we think through what has driven the transition to utilizing NVIDIA's GPUs over the traditional CPUs, part of that has really just been that power limited aspect of compute. Really what NVIDIA's GPUs are doing is accelerating one single workload, offloading traditional compute to take on that one single workload in a way that can — on a constant basis for constant costs — can use considerably less power than traditional compute would.

And so I think power is incredibly important to all these cloud service providers. And what we're seeing from semiconductor companies is an enormous focus on power, driving power down, how to manage around flowing as efficiently through the compute system. So it's something we focus on from a sustainability perspective, but also just from a business model perspective for these companies.

Vish Hindocha: That's really cool. I was just thinking about all of this and how much it's moving so fast. And maybe for our listeners, we should date stamp today's conversation as being January 24. And especially in the context of geopolitics, as you mentioned, Taiwan semiconductors or TSC, just acknowledging that we've just had the Taiwanese election that happened and the results. And to your point on even if the geopolitical tension remained at status quo as we are here today, there's still ramifications for businesses and for sentiment and for how companies are going to react to the current environment. So the last question as we think about this particular area of your coverage, Genevieve, and you talked about how you think long term what's going to be differentiated, you talked about range of outcomes, how you think about your engagement activity and how you integrate some of these issues.

Is there anything that you would say differentiates your view or your team's view, the team's view versus maybe some of the prevailing market wisdom or narratives? Anything that people should be thinking differently? One of the great investors that may be in this space, Peter Thiel, often ask the question what do you believe that no one else believes in this space? Is there anything that you and the team believe or think about that very few other people are paying attention to?

Genevieve Gilroy: I think we're likely paying attention to all of the same things, but really I think what differentiates the way that I've been thinking about it is focusing and being anchored in valuation and really staying away from the hype. I mean, semiconductors have always been a space that tends to trade peak multiple on peak earnings and through multiple on trough earnings. And this is a space that tends to get very excited and overexcited at the hype of the cycle. And not just the AI cycle, but any kind of traditional semiconductor inventory cycle and then tends to get very, very bearish at the bottom of the cycle. Valuations tend to move to trough levels. And so I think what we've been really trying to anchor to is really thinking through that risk reward, really anchoring in what we can know versus what we can't know in this ever evolving period of time. And again, just staying anchored in the risk reward, not getting wrapped up in the hype, but truly understanding and digging into each of these businesses and understanding what truly could be the potential opportunity versus risk across these businesses.

Vish Hindocha: Thank you for sharing that. Be interested on how you stay calm throughout the hype cycle, but I want to move to fun facts at the end, if you'll humor us. So we have a few questions for some of our guests just to get to know you a little bit better. Outside of MFS, when you're not thinking about the weird and wonderful world of semiconductors or protein businesses or consumer staples, what are you devoting your time to?

Genevieve Gilroy: I would say I love taking long walks with my dog. She's three years old and we love to just go through the woods, center ourselves in nature and really just find that peace and breathe that fresh air in the effort of continuously remaining calm and peaceful. So that's always been very helpful. I love to read. I spend a lot of time reading and, again, just in the effort of centering and learning. So that's probably what I spend the most of my time doing.

Vish Hindocha: Nice. On reading, what is the book, article, podcast, piece of literature that you've recommended the most?

Genevieve Gilroy: Yeah, it was probably a book that I read this year. It's called The Overstory by Richard Powers. And I know a few of my colleagues have read it as well and spoke very admirably about it and I would encourage anybody to pick it up because it really made me think about many of these questions surrounding climate change and our ecosystem in a very different way in terms of the responsibilities that we have to each other and also to all of the other organisms that really live on this planet. So it really stretched the way that I think about the issue of climate change and the issues that are facing all of the elements of natural capital that we have on the planet and really touched me.

Vish Hindocha: That's beautiful. That's something very close to my heart, so we'll add that definitely to the list. Genevieve, what's the kindest thing someone's done for you?

Genevieve Gilroy: Quite frankly, I'll think about this from a career perspective, but I think one of the things that sticks out to me in my career was, which is also my life as well, which is the mentorship that I think I've received from people here at MFS. And there are a few specific people that stick out, but so many people in terms of really creating that bug for me for investing in stocks. I started at MFS as someone who wasn't totally sure of what their place in this world would be and there were a number of people who really went so far out of their way to actually take that time to sit down and teach. And I think that's among the most kind thing that someone has done for me and has had long term implications. So yeah.

Vish Hindocha: That's great. Thank you for sharing that. Genevieve, thank you so much for spending the time with us today. Appreciate with the amount of disruption going on in your space, that taking the time out to talk about what you're doing I think means a lot to us. And that generosity of your time with our listeners is really important. But actually, you've shared a tremendous amount of insights in terms of how you think about it, the right questions to ask, how to stay calm through the hype cycle, both peak and trough, and to say disciplined on the process with the team. So thank you so much for being with us today.

Genevieve Gilroy: Thanks for having me.

The views expressed are those of the speaker and are subject to change at any time. These views are for informational purposes only and should not be relied upon as a recommendation to purchase any security or as an offer of securities or investment advice. No forecast can be guaranteed. Past performance is no guarantee of future results.