This podcast is for professional investors only.

Erika van der Merwe (EM): AI is moving into the real economy, changing how we work, consume, and travel. It's also shaking up the investment world, generating impressive stock market gains for the perceived winners. Is the market now starting to move beyond its obsession with the usual suspects in the AI race? And perhaps more fundamentally, can investors realistically determine the fair value of AI-themed stocks while the technology is evolving rapidly and unpredictably? Well, I'm pleased to welcome Jack Neele and Daniel Ernst to hear how they're dealing with these questions. Jack is portfolio manager for Robeco Global Consumer Trends, and Daniel is portfolio manager for Robeco’s Digital Innovation strategy. Welcome, gentlemen. Love having you here.

Daniel Ernst (DE) and Jack Neele (JN): Thank you.

EM: Well, first of all, what should we know about how AI is changing our lives? Are we seeing evermore practical applications being launched every single week? So what's happening? And how do you expect these new developments to change our lives? How will our lives be different, let's say, three years from today?

DE: Can I make a prediction? Not in three years, but in ten years no-one will be talking about AI. It will be embedded in everything that we do, similar to software or the internet. We will use it every day, but we'll largely take it for granted. And that process still is very early. ChatGPT launched two years ago, and we're just now kind of seeing the applications. And some of that might be quite boring, like filling out your expense report, but it could also be life-changing. So, Demis Hassabis, who's the CEO of Google DeepMind, he won a Nobel Prize last year. He was speaking, at Davos, last month, and he announced that this year they will have a drug developed, created by AI in clinical trial.

EM: Wow.

DE: That’s relatively incredible. It's a relatively new technology and we're already having boring impacts, like filling out your expense report, but also potentially saving lives. So still early, very exciting, but I think still in ten years, it will just be embedded there.

EM: Exactly.

JN: You know, I think what's changed recently, is that the focus has shifted more to these AI agents. I look at them as basically hyper-intelligent personal assistants. And we've seen it in both, in the enterprise and on the consumer side. Obviously, CoPilot is relatively well-known for summarizing meetings and even writing some emails perhaps for you. But we've seen, for instance, in the travel space that, Airbnb has been developing its AI concierge, which cannot only book trips for you, but also recommend what restaurants to go to, what things to do while you're there, maybe even look for the apartment’s WiFi code, etc., which we normally always do the first time we enter an apartment. And so you definitely see a shift in the behavior. But I think from an investment perspective, what we are focusing on, you see a lot of companies obviously being very busy developing AI, and working with AI internally, and for the majority of them, that will lead to productivity gains, so on the cost side. But there's a smaller subset that will actually be able to monetize them, i.e. generate revenue, generate profit potential. And I think these will be the more interesting companies over the longer term to benefit from AI.

EM: So, Jack, you've made that point on how it's changing things, also from an investment perspective. Has it expanded the investment universe for you, both Jack and Daniel? So on the one hand, you've got the direct beneficiaries of AI-related advancements, whether it's the hardware, the software or the services, some of which you've mentioned. On the other hand, also the indirect benefit that it creates because businesses are leveraging this technology to be more efficient or innovative.

DE: Well, if you go back to what I said about it being embedded in everything, so kind of mirroring what Jack said, with that, every company will have it but will it really drive value in their stock prices? Probably not. And so in prior technology cycles, it has tended to be that a select number of sort of foundational providers of that technology capture most of the value. So if you go back to the mainframe era, it was just IBM. In the PC era, it was some combination of Microsoft and Intel, and they called that Wintel. In the mobile era that's been Qualcomm and Apple. In the internet era, that looks a lot like something like, Microsoft and Nvidia that provide that sort of foundational layer, regardless what applications you use or how you build it, those toolsets get used. And that tends to be how we approach a lot of things in growth and thematic investing. So, there was a study by Professor Bessembinder in the US that found that something like over a 40-year period, 2% or 3% of companies captured 95% of the value. So that's typically what we're looking for.

EM: Jack?

JN: I think what you're seeing in the stock Market definitely, is that you see more dispersion in the performance. I would say last year, everything that was related to AI, whether it was chipmaking, whether it was industrials working on the electricals or liquid cooling of data centers from nuclear power reactors in energy, everything related to AI was basically on fire and going up. So far this year, you see a little bit more dispersion. And so investors are questioning also a bit more whether all of these capital expenditures – obviously you've seen many companies announcing billions and billions of dollars in capital expenditures – There's more questions about, what about the returns and when are we going to see these returns? So beyond the fact that, I would say Big Tech and the hyperscalers will generally be attractive, because intelligence is now a function of computational power. And the more computational power you have, then you're in a relatively good position to start with. I think, Mark Zuckerberg mentioned on the most recent earnings call that Meta plans to have a billion people using its AI applications. Obviously, they have about 3 billion people in their family of apps, but about a billion people actually using it... So obviously, these companies tend to be relatively well-positioned given the computational power that they have.

EM: Well, let's pick up on the point that you made, Jack, on the CapEx budgets of these companies. And I think part of those market-related gains, the momentum that we saw in AI-themed stocks, was the fact that they had these massive budgets and these have knock-on and secondary effects within the ecosystem, within the industry. And as you said, to be a player in the future, they simply have had to invest heavily to to not fall out of the race. And presumably it's very difficult for the companies and for the analysts and investors to discern whether that's wise investment. Is it too much or is it in the wrong direction? Are we at peak CapEx, etc.? So how do you go about that, applying wisdom?

DE: Well, going into this most recent earnings season, there was some concern that CapEx budgets might get dialed back. But in fact, all the big US tech companies dramatically increased their budget for the coming year. So just within 2025, CapEx estimates went up USD 27 bln to USD 314 bln. And for 2026, USD 48 bln to USD 367 bln. So they all had a very consistent message, which was: the risk is greater to under-invest rather than to over-invest. And you always see these sort of periods of like, okay, maybe there's too much capacity, and then we dial it back. But whether that was in the PC era, whether that was in the internet era, in the cloud computing era, but ultimately all that capacity has gotten used and we keep looking for more.The faster the compute is, the less expensive it gets, the more we use it. It's Gavin's paradox. So I think we're still relatively early, so it's a bit hard to say when this will slow down. But right now the numbers keep moving higher.

EM: And, Jack, during this earnings season that Daniel just referred to, there was the announcement from DeepSeek, on the latest model. And this prompted fears that it would cause these tech players to cut back on their budgets because, look, it's actually far cheaper than we all thought.

JN: Well, I think maybe the opposite has been true. So there's a couple of effects I would say one, it's put Chinese tech companies back on the map, basically. I think the narrative used to be that, the US will win the AI race, or the US is dominant in the AI race, and now all of a sudden it seems to be a much more competitive field perhaps, than previously thought. So maybe the need to spend more is actually bigger now than it was pre-DeepSeek. And I think that's also what we've seen from the companies' announcements that they're actually increasing their spending and not, not dialing it back. I would say, and that's maybe one of the other consequences here, is that what we've seen is that because of the import restrictions and maybe because the Chinese companies, on average, have less access, I wouldn't say no access, because there always seems to be a bit of a workaround, to the latest technology of chips, that they've had to resort to other means of efficiency. And I think it's quite admirable what they've done. And you do see that it's led to, a lot of excitement, also in Chinese suppliers like second-tier companies with exposure to AI. So I would say it's definitely brought the Chinese tech ecosystem around AI back to life, or back in investors radars.

EM: Now, related to that is the whole geopolitical aspect of AI. We had President Trump announcing the Stargate program. You've seen other regions, Europe also not wanting to fall behind, France being quite adamant to be in front of the race in Europe, etc. Daniel, what are you seeing in that sense?

DE: Yeah. Well, to just, go back to the DeepSeek question, because it actually came out after the news that they had spent significantly more, billions more, on developing it than what was actually reported, because they had access to the high-end US chips. So there was that. But this concept of cost efficiencies, that is just de facto standard in technology development. So Google, for instance, announced within the last year, they reduced the cost per AI query by 90%. OpenAI had a similar figure, the most recent Nvidia chip is 75% more effective than the prior one. So this is just standard in the industry. But every time along the way we keep using more compute. So I think that it kind of obscured what was really going on. I think what Jack said about, like, oh well, there’s actually this world outside of the US because there's going to be this chasm between them because the geopolitical restrictions I don't think are going to change. And so there'll be a China AI market, just like there is a China internet market. And there'll be a US and a Western market. So I think those are going to happen. And then after that, there's just what you actually do with the technology. And that's what's more exciting to me now.

EM: So will it be a kind of an AI race?

DE: Well, yeah, I mean, that's as we mentioned, the CapEx figures among just the US companies. But, there was a headline today that Korea is working on a USD 35 billion [program] – Korea’s just a relatively small country. Saudi’s building one, UAE is building one, so you're kind of seeing these sovereign models come around. The AI arms race is alive and well, and it's creating tremendous new opportunities. If you look back to the early days of the space race, everything that we do in technology kind of came out of that. So it's kind of exciting what will get built in the future.

JN: Well, I think that it's really important to these big tech companies. I mean, the CEO of Google said in a recent interview “I'd rather go bankrupt than lose the AI race.” So he definitely sees it as a race, maybe among American companies, but now we've added the Chinese as well. But they see it as an existential threat. And I think a lot of companies probably also look back at the internet era, where a lot of companies were either late adopters or maybe missing out entirely. And so they've seen the importance of these technology shifts, and they don't want to miss out. And hence they might overspend, in the short to medium term. But it's more important securing a long-term place at the table than overspending a bit now.

EM: So this is a very complex environment. So much is at play for companies, for sectors, for countries, for regions. If you are an investor, you’re having to determine whether the stock is overpriced or not, overvalued or not. So how difficult is it for you to determine the fair value of a stock, given that it's still so early on in this race?

DE: Well, your last point, I think, is most, important because it is too early. But, I think of fair value as a static term. If you have a relatively slow moving industry or company or business and you say: this is the fair value of it today But when you have a dynamic business, growing fast over multiple growth factors, the word fair value sounds like a present term rather than a future term. And so we think of value as being earned rather than assigned, and these companies are growing earnings. But just to give you some perspective, at the 1972, at the peak of the market, the Nifty 50 traded around 45 times earnings, at the peak of the internet bubble the Nasdaq 100 traded at 75 times. Leading companies were trading at 100-150 times. Today, the Magnificent Seven traded about 30 times. So more expensive than the market, but hardly in that bubble category that we've seen before. And on that internet comparison, Bespoke Research had a note out last week, with some graphs pointing to when Netscape launched in 1994, it was the first internet browser. Two years later, the Nasdaq was up 83%. Two years after ChatGPT came out, the Nasdaq is up 82%, and the lines look scary similar. Five years after Netscape: 450%. So we're two years in.

EM: There you go.

JN: And when famously Druckenmiller said, when I see a bubble I rush in, right, when I see a bubble forming.

DE: And the big difference between now and the internet era, which I began my career in, was those companies didn't have earnings, and their customers were primarily, levered telcos and ISPs that also didn't have money. So when the market turned, there was no debt funding to buy then. Today, the internet is being powered by companies exuding cash. And, they're large, profitable companies and they're investing to continue that trend. So there's some parallels and but there's differences.

JN: I think a couple of metrics that we think are important are... One is the durability of growth. Obviously, now a lot of companies are benefiting from the AI build-out or infrastructure build-out. But the question is whether that's just a temporary increase in growth that they benefit from the build-out for a few years, and then it sort of sort of normalizes, or whether it's a structural increase in growth. And there I think also the moat or competitive advantage comes in. You see that especially during these periods of change, that the companies that do have a moat or do have an advantage over their peers, they tend to massively benefit during periods of change. So I think those are a couple of important things we look at in companies. And from a portfolio management perspective, it's important to look at, what's sometimes called Hofstadter's law. And that says basically that stock prices can go up more than you expect, even when taking into account Hofstadter's law. So that sort of means, even though you know valuation can sometimes become irrational, it can still go on longer than you expect. And that obviously holds both for the upside and the downside, that normally follows after. But I must agree with Dan that especially among the bigger tech companies, obviously, these are, maybe a little bit more expensive than the market, but not super expensive. And they're still very dominant in each of their respective niches. So, I’m not really concerned about those.

DE: And the other thing people have been concerned with, okay – so where is the return on these AI investments? And so again, relatively early, but to throw some numbers at it, in 2024 cloud computing – so this is the big services that Amazon, Google and Microsoft primarily provide – expanded at a 25% annual rate. That was up from 20% in the prior year. So, revenue growth accelerated I think it's something like USD 225 bln in revenues. So that's not small numbers. Microsoft, which is not a small company, just their AI revenues grew, at an annualized rate of 175% year over year. ServiceNow, which has one of the agents in the market today, sort of not only to tell you how to fix an IT problem, but actually to kind of do it for you, their AI-assist product grew 150% quarter on quarter. So we're seeing some actual real dollars being generated from these investments. It's not just money going in, there's money coming out.

EM: Yes. And Jack, also linking back to a point that you made, now talking about the moats around the companies, like how protected they are. So I think part of the work you must be doing is knowing who will be the disruptors and who will be disrupted in all of this, and assume the ability to have the CapEx spending early on and to be consistent in that is a key factor.

JN: I would say it's hard to see who is being disrupted now as we're still relatively early, but I would say, from a portfolio management perspective, it's important to be flexible because it's even more difficult today to separate between signal and noise, and there's a lot of noise obviously surrounding AI, which is being amplified by social media as well. So that makes it perhaps a little bit more difficult. So being flexible is important, not to have too many fixed ideas, I think, Mark Twain famously said that, “it's not, what you know for sure that will get you into trouble. But it's what you know for sure that just isn't true.” And so if you have a high conviction in a certain holding that turns out to be not true or that turns out to be a loser that makes it, obviously quite difficult. So being flexible and remaining flexible to see who's being disrupted or who's at risk of being disrupted, I think is really important.

DE: And in prior similar tech cycles, there did seem like an obvious disrupted group. So in e-commerce, right, the bricks and mortar stores were very slow to adopt the internet. And so it was kind of... it seemed clear that these internet companies were taking share and the bricks and mortars were losing. In the software industry, the legacy on premise software companies were slow to adopt the cloud and the cloud-native companies outgrew. But here both internet companies, computing companies, software companies are all embedding AI in their core offerings. So we don't actually see the disruptive side. So it's a good thing we're long-only and not long-short. Because, historically, the way I've played these trends is to be long the disruptor and short the disrupted. But at the moment there's companies that I think are overhyped, but not necessarily being disrupted.

EM: I see your point. Now moving on to regulation. This is another hotly debated subject, also some geopolitical aspects to it. So first of all, what's at stake if AI developers don't have sufficient guardrails?

DE: Yeah. It's a great question. There was a survey that Stanford puts out every year, it comes out around June, I think, and this was the No. 1 thing, I think like 75% of AI experts said that AI-safety was being, under-focused on. And there have been some movements, to have the government get involved. So there was an executive order, from the prior US administration around AI safety that's been undone in the Trump administration. And what they're adopting is something more like what we had in the internet era, which is a light-touch approach. This is a new technology, we want to make sure that the best players on the field, don't have their legs tied together. So let them go and see what they do. A little different approach in Europe where there's more of a focus on security. And in China, which the last five years, the administration has been, I think somewhat, over-controlling of their top tech companies. XI Jinping just met with them, and made this announcement around something similar to what I think, Trump did, saying, “okay, we need these companies, to be our best chance at advancing this important technology. So have at it.”

JN: I think it's also maybe a cultural element, that the US is mostly focusing on the AI opportunities. And, so Trump has announced this AI and crypto czar as part of the administration, obviously Elon Musk as an advisor. So from a technology understanding, these are very much top of the bill, I would say, whereas in Europe, as Dan said, the focus has been much more around AI-safety. So you see, sometimes you see these memes going around on social media where the US is focusing on the leading edge of technology, developing these, high, leading-edge, technological applications. And Europe is focusing on non-removable bottle caps and paper straws. And it may be an exaggeration from the truth, but it's not completely false either, right, that there's a lot of focus on safety and regulation, whereas we sort of hamper, the technological development of companies in Europe a little bit.

DE: Yeah. That being said, there was a conference in Paris, this month or last month, on AI and I think the tone coming out of it was potentially that the EU lightens up on this stance. You have a private company, Mistral, which came out of Google, a couple ex-Google guys, that have one of the leading AI models in Europe. Obviously, in the UK DeepMind, which is part of Google now, was founded. So I think we're sort of seeing some recognition, that we need to, as Europeans too, let this thing roll a bit before, tightening it in.

EM: Now, bringing it all together, what would you say is the best approach to gaining exposure to the AI investment opportunities? So as we said, these are early days, but these very strong gains, let's say over 2024 that are still continuing might allow the average investor into thinking that, you know, this can continue. Although, Daniel, some of the stats that you gave us could confirm that it could indeed continue. So having said all of that, what would be the best approach to gain exposure,
without giving investment advice?

DE: Well, as we've written in several of our papers, we tend to focus on looking for those long-term compounders that have some edge, that are proving it or investing in themselves. They're showing profits today, and they're growing. And those foundational providers will capture the majority of the value over the long haul. So I think it's not different than any other technology or any other industry. You want to look at what the company's investment plans are and what products they have in the market, how they're competing and what profits are they generating and how they’re approaching the market. So I think it's not that different, even though it is fast-moving.

EM: So the tools aren’t changing, your methodologies and parameters are the same. It's all the same assessment of valuations and future prospects. Correct?

JN: Yes, I would say so and also think twice before investing in the second tier of companies just because they're cheaper, right? Because you have seen that the market leaders tend to extend their lead, especially during these periods. And so whether it's in digital advertising or in cloud computing or in chipmaking, you've seen the familiar leaders. Besides that, we are also looking at what's happening in ride-sharing, autonomous vehicles. We talked about the travel AI concierges, etc... So these are also areas that we're looking at to see whether these companies can play an important role, in the future development of these industries.

EM: Jack and Daniel, thank you. It's been a delight speaking to you. Great insights.

DE/JN: Thank you.

EM: And to listeners, thanks for being here. We publish a new episode every month covering a range of investment-related topics. This podcast and Robeco's markets podcast ‘In tune with the markets’, are both available on all major podcast platforms and on the Robeco website. If you subscribe, you'll receive a notification as soon as the new episode is published. In the meantime, please rate the show and share the show link with a friend. Until next time.