EDITED TRANSCRIPT
Lukasz de Pourbaix (LDP): AI is no doubt very topical; it's in the papers every day. We had a discussion around this theme in an event we recently hosted talking about navigating the AI boom. One of the questions that constantly comes up within the AI theme is the question: are we in a bubble? What are some examples of previous bubbles we've gone through?
Maroun Younes (MY): What characterises a bubble typically is where the narrative takes over, so it becomes less about the fundamentals and what businesses are doing, but more about the hope and the narrative of the future, driven largely by sentiment.
Part of it is fear of missing out. Part of it is driven by the narratives that these businesses will dominate the future, for example. Ultimately what you get is really prices detaching from reality or detaching from the fundamentals. It then becomes less about what you can justify given the current state of the businesses and more about what these businesses may or may not deliver in the future.
In terms of some of the more notable ones, the first one that comes to mind is the Tulip bubble in the Netherlands around 400 years ago. Some more notable bubbles in the last 100 years are driven by new technologies such as radio in the 1920s and the dot.com bubble in the noughties - a big one. There's been a few other regional bubbles, like in Japan in the late 1980s, impacting Japanese real estate and asset prices.
There are some smaller ones here and there. Crypto has gone through a few of them.
LDP: It sounds like there are some common themes in all of those. There is a book called ‘Irrational Exuberance’ by Robert Shiller where he talks about some of these thematics. The media talks about the relinquishing of risk and it's all upside. I’m interested in your thoughts around some of those characteristics, if you think about where we are today.
Some of the broader market structural things in the market we are observing are the magnitude and length of potential bubbles is elongating because of the market structure. Common things being referenced include the growth of ETF's, passive investment. This feedback loop, where the weight of money going into things along elongates the potential for bubbles to go for longer and be more powerful? Is that something you're observing as well more generally in terms of that dynamic?
MY: I haven't seen anything empirically to say that the points you raise definitely make sense. But I guess if you do go back in time and look some of the other bubbles historically, they have tended to go on for years - the railroad boom was close to a decade. So, it would be hard to say and prove empirically. But the bubbles are getting longer or shorter.
Historically, if you think about the conditions for the narrative to take over for people to get sucked right into the vortex of it, normally that's a process that takes quite a bit of time and I don't think today would be any different.
LDP: As I mentioned, there is some debate is you know as to whether AI is a bubble waiting to happen or is it a structural growth story that actually has fundamental legs? The reality sometimes is a bit of both, right? What would it look like if we suggest in the case of AI, the pro is yes, it is a bubble and the con is no, it's not a bubble?
MY: Firstly, I'd say they're not mutually exclusive. Most of the bubbles historically have been driven by something that's fundamentally changing the world. So again, railroads, it was changing the way passengers move around the continent, changing the way a freight moves. Radio fundamentally changed the way we communicate back in the 1920's. The internet fundamentally changed things 20-30 years ago. AI will change things in the future. I don't necessarily think the two have to be mutually exclusive in terms of where it could be a bubble.
Certainly, the amount of capex that has gone in a very short space of time, we rarely see this much capacity coming in a particular area. That would potentially lend itself to an argument around the bubble. Also today, the revenue and returns that we're generating from all of this spend is yet to deliver anything meaningful. So, if this were to persist then certainly this would also lend itself to being a bubble. It’s a bit early to tell because the returns still may come through over time.
The excitement, the hype, certainly the narrative is there. The duration in terms of how long out into the future investors in the market are looking. I think those softer factors certainly lend themselves to a pro bubble argument.
LDP: You take the view that maybe not a battle, but it is a structural story. What are some of the things in your mind that would be supportive of that case?
MY: The main players today, with the exception of OpenAI (the owner and maker of ChatGPT), more reasonable valuations they're currently trading at. Thinking about the dot.com bubble, it was characteristic for companies to be trading on 60-70 times price to earnings, sometimes not even making any earnings, just trading on a sales multiple. It's fundamentally very different today because what you're seeing is again, except for OpenAI, the majority of the big names in the AI race are businesses that have been around for decades, are established, very large, very cash generative, and vastly profitable.
This is very different to say the 1990s where any person could raise a few $100 million, have a company name with a .com on it and then list and off you go. The main players today, Amazon, Microsoft, Alphabet, Meta, Oracle - these are businesses that have been around for 15,20,30+ years’ worth of history, trillions of dollars and generating hundreds of billions of dollars in annual cash flow, and the majority of them are trading at 35 times or even less price to earnings, very different again to a Cisco or a Sun Microsystems in 1999 was trading predominantly on a sales revenue. Hardly any profits astronomical 70-80 times.
The exception is OpenAI. The company has half a trillion-dollar valuation, producing around 15-20 billion in annual revenues, more on an annualised run rate, so probably closer to 30+ billion annualised run rate. For 2025, 15 to 20 billion in revenues with half a trillion dollar valuation with capital spending commitments over the next five years.
In excess of a trillion dollars, that one stands out. But if you look at Amazon or Microsoft, they have reasonable valuations and they've got a legacy, immensely profitable business. Fundamentally, it’s very different to what we've seen.
LDP: It will be fascinating to see. I think OpenAI is looking to IPO at some stage. So, it will be really interesting to see how the market reacts.
Your day job is looking at the broad investment universe and working with our research team to identify quality names in a sector like this where it can be challenging. In the case of the AI theme, a lot of the question is around what is the use of case and how is that monetized over time? There are still some question marks over that.
What are the types of things you're looking for? What are the typical characteristics you like to see when you're looking at a company that's either directly or indirectly linked to this AI theme?
MY: Because we're in the small- and mid-cap segment, a lot of those names that I mentioned earlier, the leaders in the AI race, are beyond what we can look at. This means our focus is on indirect plays. The direct plays tend to be larger. It's more about scale of data and those sorts of things. What I'm looking at is effectively indirect businesses, the picks and shovels plays.
These businesses will play a role in a particular niche in enabling some of the larger players with what they do. There are a few different examples in the Fidelity Global Future Leaders Strategy today. One example is TechnologyOne, a leader in cooling, designing and installing HVAC system. Data centres generate a lot of heat so having reliable ventilation and cooling in an energy efficient way is quite important.
Another exposure is through utilities. Right now, one of the biggest bottlenecks for being able to build data centres is access to power, more than access to chips. So we're seeing names like Microsoft, who have a bunch of chips which they don't have power to, just sitting there idle. So that is a big constraint and it won't solve itself easily in the next year or two. If you think about what's required you need permits, planning, zoning - whether you want to build a nuclear power plant, whether you want to commission things or decommission things. All these things need refurbishments to be brought back up to modern standards and to make them a bit more efficient.
All these things will take time. Whether it's Greenfield or Brownfield energy expansion, these things are not going to solve themselves in the next year or two, it’s probably going to persist for a number of years.
For the picks and shovels, they're less exposed directly to the return of AI. They're more exposed to the hyperscalers continuing to race amongst themselves. Obviously, if the returns don't come through, those hyperscalers will eventually dial back their capex spend and that will impact the picks and shovels plays. For the time being, and for what we can see in the likes of Microsoft and Amazon future spending plans, these picks and shovels plays have 12,18,24 months of good visibility backlog into future work coming through the pipeline.
LDP: It's an interesting point because I think one of the interesting things a lot of investors are thinking about is their exposure to the AI theme in an index or in an active strategy in the large-cap part of the market. If we do see a pullback in that theme and generally that would certainly impact the broader markets.
The S&P 500 is a big component of that Index, 40%. The mid-cap and small-cap part of the market which you invest in, how would you envisage that if we were to hypothetically see a pullback and the fact that you aren't investing directly in those big names you are?
MY: It's a very interesting question because if you look historically, the majority of market corrections tend to be led by SMIDs on the way down. They tend to be more cyclically exposed. They tend to be in aggregate less resilient businesses and higher beta.
Normally when the general economy starts to roll over, they're one of the bellwethers in terms of rolling over first, and you see a flight to quality up the market cap spectrum. This is very unique because this boom has not been led by and, but also the boom is usually led by SMIDs as well. This boom is not led by SMIDs, it's actually led by the mega cap end of town and your question is specifically not around the general market recession, but an AI led correction. So that one, I am not sure if SMIDs would lead on the way down, it may be the mega-cap names if it’s purely around scepticism or pessimism around what AI means for investors. I think you would see something different, where it would be potentially led by the mega-cap end of town, then it would have contagion effects into the broader market because it’s 40% of the S&P 500. So, the market in aggregate is almost certainly going to fall. We’ve never seen such a large concentration exposed to a handful of names.
Then you get some contagion effects because you get negative wealth effects, risk appetite coming right ow, people becoming very risk averse, taking money out of equities and putting it into cash etc. That would have a broader impact on the rest of the market and SMIDs would be no different.
The best analogy I could look at is the dot.com bubble where you had the NASDAQ peak in around March or April of 2000. The S&P which has a lower tech rating back then and certainly a lot less than the NASDAQ, that didn’t peak until around September or October. For a few months there you had this this unusual thing where the NASDAQ was going down, the S&P 500 was still chugging along in the upwards. Trajectory before it finally rolled over, because then you had a contagion and you actually had a general recession coming in, but certainly the reduction in the S&P 500 back then was a lot less than what we saw in the Asda now. The S&P today is nowhere near as diversified as what it was back then, so it probably behaved more like the NASDAQ did.
I think the SMIDs category certainly you may see something similar and also if you think about valuations in this mid-cap universe today is broadly in line with this long term history, whereas valuations in the S&P 500, the NASDAQ, the mega cap end of town are in their top decile of historical data that we have. So, certainly there's more of a reason to be optimistic that SMIDs can hold up a little bit better, relatively speaking. They'll still fall in absolute terms, but certainly there's a lot of reason to believe that they won't fall by as much as what you may see in that top end of town.
LDP: We saw a little of that when we had that DeepSeek moment where obviously, a couple of question marks were raised about the business model of some of these large players and some of the mid cap part of the market actually held up better in that environment.
If you look ahead, where are you looking for opportunities in the AI space?
MY: We're always on the lookout, but I think just given where we are in that AI cycle and the potential for it to manifest a little bit more into a bubble. What I'm particularly looking for is businesses that have some exposure that have not yet been swept up in the mania. Because ideally what you want is a business where it's a division or a small part of it exposed to AI that could actually record growth. But it's not really caught on by the broad market because right now.
What we're seeing is lots of companies that are exposed to AI directly or indirectly, but the markets recognise that and they're trading at elevated valuation multiples, which means a downside risk should things spiral out of control is a lot more. So, it's really trying to turn over as many rocks as possible in this space. Try and find businesses that that the market is yet to catch on in a big way, that actually, these things are exposed to are positively exposed to AI. Where the valuation makes sense then then dig further into that. It's a very much a bottom-up, stock by stock, company by company type of process.
