408. Is AI in a bubble?

24 September 2026 (pre-recorded 16 September 2026)

Please be aware that the accuracy of artificial intelligence-generated transcripts, such as those utilised in our interviews, may fluctuate based on factors like audio quality, subject matter complexity, and individual speaker enunciation. Consequently, these transcripts are unlikely to achieve 100% accuracy. However, it is important to note that, at FundCalibre, we do not consider the correction of automatically-generated captions to be an effective or proportionate use of resources.

Given the inherent limitations of machine-generated transcription, we strongly advise against relying solely on this transcript when consuming our content. Instead, we encourage you to use the transcript in conjunction with the accompanying interview to ensure a more comprehensive and accurate understanding of the topic.

Please remember we’ve been discussing individual companies to bring investing to life for you. It’s not a recommendation to buy or sell. The fund may or may not still hold these companies at your time of listening.

[INTRODUCTION]

Staci West (SW): Welcome back to the Investing on the go podcast brought to you by FundCalibre. Is AI in a bubble? And if it is, where should investors look next? We explore where we are in the AI investment cycle, why the most obvious opportunities may be changing.

James Yardley (JY): I’m James Yardley, and today I’m joined by Mikhail Zverev and Graeme Bencke, fund managers for the Amati Global Innovation fund. Thank you very much for joining us today, guys.

Mikhail Zverev (MZ): Hi, James. Thanks for having us, as they say on the American TV shows.

[INTERVIEW]

JY: Very good. And what a great time to be joined by two managers of an innovation fund, with AI obviously being so pertinent and so talked about in the press. And I think, obviously, the big question for investors at the moment, with a lot of people asking themselves, is this a repeat of sort of what happened with the tech bubble in 2000, or is this kind of a complete change of the world, a step up in growth rates, a new age we’re going into?

So perhaps you could give us your views on that. Where are we in this whole AI growth cycle? Are we still very early stage, or are we getting to the later innings now?

Graeme Bencke (GB): It does definitely seem like we’re in a bit of a bubble. There are, you know, certain behaviours that are definitely reminiscent of that sort of typical bubble stage of the investment cycle. And Mikhail and I have lived through a couple of these before.

If you look at AI CapEx as a percentage of GDP, it is pretty comparable to kind of the dot-com period or, you know, the telecom bubble at the time. And the trends we’re seeing in circular financing and the kind of high retail participation across the market in AI-related stocks, all these things kind of lead you to think that we’re getting towards that point of bubble territory.

There’s a lot of AI plays in everyone’s portfolios. From our perspective, we consider ourselves to be kind of pragmatic optimists, but we do feel that it’s time to have a bit of extra care in this space.

So we focus on companies that are likely to see continued growth, even if the spending by the Frontier Labs does begin to slow from here, or AI capability doesn’t improve. So we’re not betting on things continuing to get better. We think of these as being somewhat asymmetric bets.

So these are areas like memory. I mean, memory will continue to be in high demand even if the capabilities don’t improve, just because there’s going to be more and more use of AI across, you know, data networks and companies just using AI more and more in their day-to-day business. So memory will grow regardless, as well as areas like, you know, we just will need more power, so more electrification, improved grid infrastructure, that kind of thing is happening even if, again, AI doesn’t really improve. And these are probably decade-long trends.

So, you know, we’re trying to focus on those things that will continue to do well regardless.

MZ: Yeah, the one way to think about what we think is happening, exactly what Graeme is describing, is to look at what we own. You know, the evolution of our holdings within the broader AI theme is the best indication of where we think we are in the cycle. And, you know, if we look back at how our AI exposure evolved, we owned NVIDIA early in the fund’s journey. The fund’s just turned four years this summer, but, you know, four years ago we did own NVIDIA. That became a very well-known, widely held idea.

We owned SK Hynix, which is kind of the original memory, HBM AI memory pioneer. We sold that. We still have Samsung as preferred memory exposure. We owned photonics early in the fund’s journey and kind of for the last several years, but again, we’re selling that down. We sold Momentum, which has been a very long-held position in our portfolio for about four years, actually. We sold it relatively recently.

So, you know, if you look at overall AI exposure, it is getting more nuanced and it’s getting smaller. So, you know, as far as the investment cycle goes, it’s extremely well discovered and we have to exercise that care.

JY: So do you think this kind of data centre build-out CapEx play, you think that’s largely played out now? We can’t continue to see these growth rates continuing forever and ever with the hyperscalers, and you’ve started to move the portfolio to the next stage. Is that fair to say?

MZ: You could say that. I mean, the CapEx boom continues, but the second derivative is lower. You know, I mean, it can’t grow as much as it had been growing just because there’s not enough cash flow left. You know, you have to borrow. You know, there’s a lot of borrowing going on. Some of that is off-balance-sheet borrowing.

So the pace of growth that’s already, you know, a significant proportion of US GDP growth has to moderate slightly. Hence our need for, you know, where we still have exposure to the AI CapEx, which we think will continue to grow just at a slower pace, we want it to be asymmetric, as Graeme has described it. And we are a lot more interested in the lateral beneficiaries of AI or kind of other growth opportunities, you know, within the AI buckets.

Obviously, there’s a lot more to the world than just AI, but within the kind of AI sphere, we look for lateral beneficiaries and kind of less obvious implications of what’s going on.

JY: Could you maybe go into that in a little bit more detail then? So where are those opportunities which you’re finding now, and how do you see this kind of expanding going forward? I mean, is robotics an area you’re starting to look at, and physical AI and all of that?

GB: Yeah. I mean, a lot of those things. We did recently talk about, actually, physical AI. Physical AI is something that we think we’re going to see a lot more of in the future.

And, you know, maybe people may not be super familiar with it, but it’s defined as kind of artificial intelligence that sort of connects the software models with physical hardware. So this ability to allow hardware to sense and think and act in the real world. And it really brings AI into the factory or onto the road or into the home, you know, performing real physical actions.

And again, we’re pragmatic optimists here, so we’re not sure about, we’re not relying on, you know, these kind of very significant forecasts from people like Jensen Wang, $40 trillion addressable market, or Elon Musk’s vision of humanoid robots in every household.

You know, we talk to a lot of companies at trade shows and we go to a lot of technology shows to sort of see how these things are actually, what’s happening in the field and what companies are really doing. And in fact, a lot of companies that are using robots don’t think the humanoid form factor is a particularly useful structure at the moment, but maybe in the future.

But, you know, on the more practical level, we’re already seeing a lot of cobots and robots and automated mobile vehicles in manufacturing and logistics, and that’s already pretty ubiquitous. But that will give a significant growth opportunity over the next decade as they just become more and more present. So it’s not that we’re kind of betting on a future that isn’t already here. These are things that are happening today that are still relatively underexposed in areas like warehouses. I think still only about 25% of warehouses globally have any form of automation in them. So there’s a lot still to be done.

You know, areas like digital twins. So these are a software version of a physical infrastructure that allows a company to do things like predictive maintenance or, you know, improve processes across whatever that industrial system is. There’s a lot yet to be done there.

JY: And looking at your portfolio, I mean, you seem to have companies right down the market-cap spectrum. So is that where you’re finding the best opportunities, perhaps in some of the supply chain names, perhaps, which will be going into all of these products rather than betting on one particular winner for humanoids or one particular frontier model or something like that?

Is that your preferred way of playing it? Is that the reason? Or are you just finding faster growth in those kind of smaller mid-cap areas?

MZ: We’re not a small-cap fund, to be sure. We are an all-cap fund, so we go wherever innovation takes us.

The way we think about this is, when we find an area of innovation frontier, and let’s pick physical AI as an example, it’s kind of a collection of several things in there. You know, we see this as a sort of real, very interesting technological change which touches very large markets. It is happening, industry is adopting it.

So we want to investigate, you know, who are the players in that space? What is the cluster of companies that stands to benefit from that?

And we have this framework of pioneers, enablers and adopters. You know, the leading-edge inventors that come up with the kind of frontier technology that makes it happen, the picks-and-shovel suppliers, the enablers, and the lateral beneficiaries, the adopters.

And they can be, you know, anything. We owned NVIDIA in our time, we owned SK Hynix, we still own Samsung, but we also own companies which had, you know, kind of mid-to-high single-digit hundreds of millions of market capitalisation. So right down the spectrum, wherever we see the best business. And we’re looking for real businesses, profitable, capital generative, but also kind of the best exposure at the best price.

And, you know, to pick physical AI as an example, we did indeed find a mid-cap company, a company called Belden, which we think is a pretty undiscovered but potentially crucial fabric that kind of underpins physical AI. So the way we think about it is, just like AI in data centres needed networks to make it work and, you know, people investing in networking equipment players like Arista [Networks], which we have done, or Lumentum, or connectors like Amphenol, has been a very lucrative way to benefit from the data-centre AI growth.

We think the same playbook will be repeated in physical AI, except the industrial networks are very complex beasts. They are, you know, pretty exotic specialist protocols. They’re not your usual kind of office Ethernet. They have completely different tolerances in terms of ability to cope with harsh environments, latency, error correction. It’s just, you know, those are very mission-critical networks.

And so we see, you know, as the factories and physical infrastructure more broadly get wired up – and when I say wired up, I mean both wired networks and wireless networks – Wi-Fi 7, the next iteration of Wi-Fi, is now industrial grade, so we’ll see lots more things being connected using that. You need a trusted industrial-grade networking supplier, and Belden is that. And we think that’s a somewhat undiscovered enabler of physical AI in the industrial space.

GB: Yeah. There are actually a number of different areas that kind of expand from that.

If you think about physical AI, it really means that the machines develop more and more sensors, so eyes and ears and things like that. So we’re very interested in that area.

So we own companies like machine vision companies like Cognex and Basler, because all of these machines need to be able to see. And increasingly, they have multiple cameras on them.

Or things like Allegro. Allegro is a US company that does robotic motion sensors, for example. So as robots expand, whatever form factor they take, they’ll have movable joints, and those things need motion sensors. So you get these kind of specialist companies that are real beneficiaries of this, what we see, you know, as a decade or multi-decade rollout of this kind of physical AI across the broader economy.

JY: And where are you in the software debate? Obviously, a lot of the software stocks have been hit very badly over the last couple of years or so. They’ve done a bit better recently, perhaps. Are you just avoiding them, or are there select opportunities there?

GB: Well, no, we’re definitely not avoiding them. And we think, in many cases, the market’s taken too negative a view towards software, particularly certain elements of software.

We focus on companies that are quite specialist and very embedded in their industries. So these are typically software companies that are seen as being systems of record.

So if you think about areas like architecture, design, engineering software, these companies did suffer when the big software sell-off, the SaaS apocalypse as it was called, happened earlier in the year.

But we think these companies have very strong positions in their industries. It’s much more than kind of the fairly shallow software areas you think about in things like accounting or Salesforce automation. These are more what we call skin-on-data type businesses, where the software is just making a relatively simple process more accessible.

When it comes to these engineering, architecture and design-type software companies, there are decades of knowledge that’s embedded in these companies, in the software. It’s very difficult to replicate, you know, sources of data that other companies just don’t have, as well as this type of software being what’s called determinative.

You don’t want a generative model that’s effectively making up its answers to be designing the structure of a building. You want to know the real capabilities of the materials that are being used in that building. And so that protects these businesses, you know, from some of the areas of AI disintermediation we’re seeing in other parts of software.

MZ: History never repeats, but it rhymes. And we, you know, in software, we had a similar – not similar, but, you know, let’s call it rhyming – paradigm shift a few years ago when, you know, software moved from on-premise, you know, something that runs on your workstation in the office, so to speak, to the cloud.

And the commercial model moved from licence plus maintenance, to kind of largely a one-off sale, to SaaS, to kind of subscription per user, per unit of consumption. And, you know, those are very dramatic technological paradigm changes at the time. And some companies survived and some didn’t. And I think the companies that didn’t survive and survived were kind of aligned in the same deep-domain expertise versus shallow framework that we’re describing here.

So, for example, in Salesforce, you know, once upon a time Siebel Systems [acquired by Oracle] was the go-to company. It’s now Salesforce.com, which completely eviscerated Siebel Systems. In human capital management, PeopleSoft [acquired by Oracle], in a similar way, gave way to Workday.

Interestingly, our companies – Autodesk for architecture and construction, PTC for industrial design, and Bentley Systems for kind of infrastructure construction, so rail lines and bridges – not only retained their leadership in their domains during the SaaS and cloud transition, but actually strengthened their leadership. They became better, stronger businesses.

Now, nothing’s ever certain, but our view is that they have a pretty good chance of thriving in the AI world, whereas other companies might find a struggle to adapt.

JY: And moving away from AI, you mentioned there are other areas, plenty of other areas in the world where there’s innovation. So what are some of those spaces? Healthcare, perhaps, or other things as well?

MZ: Yeah, I mean, we had a general view, which we’ve shared with investors for a couple of years now, that there’s more to innovation than AI. And this is how we run our portfolio.

We benefited from AI. It contributed very strongly to the fund’s performance, but it’s never been more than kind of 15–20% of our exposure, all things considered, because there’s just so much more interesting stuff going on in the world which, you know, arguably the market is paying less attention to.

Healthcare is an interesting one. I’ll mention a couple of things in healthcare. Some of them are AI-adjacent because, you know, one of the things that AI is doing is accelerating the cadence of, let’s call it, drug discovery. You know, the initial glimpse of the opportunity – this molecule might do that good thing in a human body – which once relied purely on human ingenuity can now be accelerated or simplified using AI. And so the number of plausible candidates for new medicines is exploding. We’re hearing that from the industry.

But if you imagine the framework for drug discovery, yes, you come up with an idea first, then you test it maybe in a digital model to make sure you haven’t missed some obvious pitfalls. But then eventually you have to test it in what’s called a wet lab. You need to put it in a Petri dish, see if it actually does something to the cell that you think it does. And so we see this kind of wet lab in the loop, in vitro validation, as a really interesting stage. And this avalanche of new ideas is coming towards that part of the chain of drug discovery.

Both the capacity of wet-lab drug discovery research, the number of tools that do this job, but also the automation that enables this much larger number of candidates to be explored efficiently, will increase. So we have quite a substantial position in life sciences tools generally, in anticipation of that growth. But we also have a Swiss company in our portfolio that is the global leader in laboratory automation, effectively robotics for labs.

So that’s, you know, you told me to look outside of the AI space and I’m just giving you an AI-related idea, but, you know, we see other things in healthcare and beyond healthcare. Graeme touched on a couple of those as well.

GB: Yeah, I mean, there are lots of interesting areas that the fund is exposed to and that we’re very excited about. Some quite specific areas. One that sort of ties into healthcare is radiopharmaceuticals, which is moving from being a sort of diagnostic technology to a therapy. And there aren’t many drugs yet on the market that are sort of serving this particular niche, but there’s a lot coming. And we think there are big changes afoot there that the market is really not on top of.

Likewise, defence spending. I mean, obviously, we know defence spending is a big deal and there’s a big tailwind there, but there are certain niches within defence spending, like space, like drone and anti-drone, and cyber, that we’re really doubled down on.

And particularly if you look in Europe, because in Europe you get sort of a triple benefit of more spending, more focused spending on those three areas I mentioned, but also a localisation of defence spending away from the kind of historic big spenders, the US primes, towards European homegrown companies.

So if you can find those niche businesses that are operating in those specialist areas, we think there are some really interesting opportunities there. We own a company called Theon, for example, which is a leader in night-vision technology and is expanding into other areas of kind of this augmented vision that we think is really interesting in Europe.

JY: Well, thank you very much, Graeme and Mikhail. That’s been really interesting. Certainly no shortage of innovation in the world today, it seems, and I think your fund, it seems to be well positioned there. So thank you very much for joining us today.

GB: Thank you very much.

MZ: Thanks for having us.

SW: The WS Amati Global Innovation fund is a high-conviction, unconstrained global equity strategy that invests in companies benefiting from impactful, real-world innovation. The managers define innovation broadly, investing not only in the companies creating new technologies, but also those enabling their development and those adopting them to drive value. To learn more about the WS Amati Global Innovation please visit fundcalibre.com