Transcript: Investment Masterclass — The case for investing in AI

This is an audio transcript of the Money Clinic podcast episode: ‘Investment Masterclass — The case for investing in AI’

Claer Barrett
The personal computer, the internet, the smartphone, and now AI. These tech innovations have transformed the way we live our lives, not to mention the shape of world stock markets. And the pace of change is accelerating. So much so, artificial intelligence is said to have reached a tipping point. It’s fast becoming a feature of everyday work and home life. And my guest on today’s investment masterclass thinks this is just the beginning.

Ben Rogoff
Some of the smartest people in the room, as they say, are acting in ways that are slightly odd. Then my gut feel here, is that the very smartest people understand that the rate of change that we’re seeing, the model of a predictable rate of change of AI, is going to get us to some human equivalent general intelligence, as people call it, much faster than the market thinks.

[MUSIC PLAYING]

Claer Barrett
Welcome to Money Clinic, the weekly podcast from the Financial Times about personal finance and investing. I’m Claer Barrett, the FT’s consumer editor.

[MUSIC PLAYING]

My guest today describes himself as an AI maximalist. As the lead manager of the £3.5bn Polar Capital Technology Trust, he bet big on AI — artificial intelligence being the dominant investment theme of the post-Covid era. Well, the gamble paid off. And with recent results from AI chipmaker Nvidia powering a global stock market rally, I think it’s fair to say that many more investors now share in his view. Ben Rogoff, welcome to Money Clinic.

Ben Rogoff
Hi. Thanks for having me.

Claer Barrett
Well, we’re delighted to have you in the studio because with more and more of us using generative AI software like ChatGPT in our everyday lives, I mean, this is the tipping point, really, that people have been talking about. And the big hope is that it will boost productivity, right? Both at home and in the workplace, we’ll be able to do more things with less effort. So as an experiment, Ben, I thought I would ask ChatGPT, what should I ask Ben Rogoff from Polar Capital in an interview?

Ben Rogoff
Awesome.

Claer Barrett
Were you expecting this?

Ben Rogoff
Not at all. What did it suggest?

Claer Barrett
OK. Well, the good news is, I think I’ll still be in a job as a journalist for a while yet, as the questions were pretty anodyne. But the speed at which all of the information just spilled out on to the screen. That’s the really, really frightening thing. But anyway, robot Claer asks: can you provide an overview of Polar Capital’s investment philosophy and approach? How does Polar Capital differentiate itself from other investment firms? Not bad, AI.

Ben Rogoff
OK. Not shabby. I’ve been a partner of 20 years. I work with my partner Nick Evans for nearly as much time, and we’re growth investors because we believe that in the end, stocks tend to perform in line with their progress in revenues, cash flows and earnings. We run three long-only funds, investing purely really in tech. 

Claer Barrett
Long-only means you don’t sell. You’ve got the courage of your convictions.

Ben Rogoff
That’s right. Thank you. We don’t run hedge funds within our . . . in our unit. We’re looking to make investments in companies that over time, the share prices will appreciate and hopefully outperform benchmarks that we’re trying to exceed. And we take a diversified approach because in the end, everything about the world tells you that domain expertise gives you an edge. It doesn’t mean you’re going to be right about each and every investment that you make. And so, very liquid portfolios. That means that we invest in companies where we can change our minds. We can sell stocks quite easily if we decide that the risk/reward has changed on a stock.

So PCT has around 100 stocks in it. All of them, we hope, are exposed to the right themes within technology. Technology itself has just continued to grow share of GDP and capital spending. It’s been a wonderful place to have been invested. Obviously, past performance is no guide to future performance, but they say you only have to make a few good decisions in life. And one of mine certainly was opting to become a technology specialist 25 years ago or thereabouts.

Claer Barrett
Well, indeed. And you might say another one is investing in Nvidia. Now, I joked on the radio a couple of weeks ago that many people might think it’s a face cream, but actually it’s one of the biggest tech companies you’ve never heard of. For the benefit of listeners who haven’t, just tell us about Nvidia, which I believe is now the second-biggest holding in the trust.

Ben Rogoff
So the company is a chipmaker, and the chips that they make are graphic processing units, which back in the day, if you are really into gaming, you might buy your desktop PC with an Intel CPU — the central processing unit, the brain if you like, of the PC — and then alongside it, you might have had a special chip that was just there for the graphics.

Claer Barrett
A graphics card.

Ben Rogoff
Right. And that card really, that market has been dominated really since the late 1990s by Nvidia and also AMD, which is another holding about which maybe we’ll come to shortly. What’s happened over the last few years is that if you like the loci of compute has moved because the workloads, the unit of compute that we’re interested in today are these AI workloads. And the fact that, you know, you referenced how fast ChatGPT was able to give you some questions, anodyne or otherwise. The speed, the latency, that trade-off between quality and speed is best dealt with a graphics processing unit, not a CPU. And again, but that would be getting into the weeds, it’s really to do with this idea of serial versus parallel compute. And what GPUs are really good at, I mean really good at, and the reason that they are used to render images in a game where images are moving incredibly fast, is this idea of taking a workload or question a task and splitting it into many, many sort of smaller elements, solving for those at the same time . . . 

Claer Barrett
Right.

Ben Rogoff
. . . parallelism, and then giving you the answer. Whereas the old-fashioned approach to compute has been very much in a serial . . . 

Claer Barrett
Linear.

Ben Rogoff
Three times four is 12, and then I’ll do the next instruction set. So that’s why Nvidia finds itself absolutely at the core of the AI story.

Claer Barrett
OK. Now you say that 80 per cent of your portfolio can be defined as an AI beneficiary or enabler. So obviously we can get that with Nvidia. But what about the others? Unpack that term to me.

Ben Rogoff
I’ll try. We break the 80, let’s say, 80-85 per cent of the portfolio that we believe are exposed to AI trend themes into really two big buckets. The first big bucket are the enablers, and for us, that’s really the chipmakers. So companies like Nvidia, but also many others that we can talk about in a minute. And then the cloud companies — the likes of Amazon, Microsoft, Google — we believe, still like to be the conduit for most AI workloads. So we’re still very positive about those companies. And so they would be enabling that. They’re the infrastructure plays, if you like, on this new opportunity. And then the second bucket, and in there today we would include companies like application software companies, but also cyber security companies that are using or benefiting from either a greater threat posed by the use of AI, the bad guys using AI, but also a greater attack surface. For example, the questions that you ask ChatGPT to generate. What if someone had been able to get in there and come up with really nasty questions, or something slightly more malicious than that? So this idea of enablers and beneficiaries together, roughly, yes, 80-85 per cent of the portfolio.

[MUSIC PLAYING]

Claer Barrett
As an investor, you’ve long believed that AI is the future, but how is it going to change our world?

Ben Rogoff
We believe very strongly, and again only time will tell if we’re right, that AI represents the next general purpose technology. Like electricity, like steel. We really do believe that. That’s why we launched the fund, we did, six years ago. We just didn’t know when the tipping point was going to occur. And like you say, now we’ve kind of got these elements in place. Let’s not forget there are 6bn smartphones in the world. And as a conduit for change, that’s unprecedented. And so if you look through history at how long it took for steam or electricity and for the first wave of, you know, IT — ICT as they called it back in the day — it took half as long for those technologies to diffuse. And so if you follow it through, you go from sort of 80, 40, 20. It should take 10 years for AI to become widespread. I absolutely believe that the diffusion rate for AI is going to be vastly quicker than that.

Claer Barrett
Let’s take a step back here for a moment. I mean, you’ve got kids. I’ve got stepkids, grandkids now. The worry for me about AI, but also the wonder, is how is it going to change their world in the future.

Ben Rogoff
Profoundly, I think, is the simple answer, but in ways that are unknown. I was likening this to — a kind of a silly aside — but I collect all computers and a whole bunch of stuff that drive my wife mad. Of those things, the one collectable item from the 90s that gets the most playtime is my pinball machine. I have a 1993 Twilight Zone pinball machine. It’s awesome.

Claer Barrett
Oh, wow. 

Ben Rogoff
And we still play on it. And why is that? Because pinball machines have got an inherent chaos about them. When the ball hits the flippers, it’s mathematically impossible to know where the ball will end up after 12 sort of bumps on the bumpers. And that creates this kind of chaotic thing that draws you back to it. Now, of course, I’m making the point, this could have been applied to markets, but it also is this sort of what happens at this point in technology evolution, which is we have this new GPT. There are things that we can know about that. We’re going to become more productive for sure. The first wave of AI is going to make humans much more productive, in the same way that the first wave of agricultural mechanisation made humans much more productive. There will always be occupations that get caught up in these GPTs. But as a whole, I’m an optimist about what AI will mean for humans for now. Then we’ll go through a more disruptive phase. The disruptive phase, we don’t know what form it will take, but ultimately, as AI models tend to human-level capabilities, there will be more disruption. The leap of faith — the bit that, you know, why do I play a pinball machine if in the end it’s a chaotic system — the leap of faith is that we know that humans are incredibly adaptable, and then this is the bit where it gets really exciting, is that I believe that AI will probably unlock idea generation in a way that we haven’t seen for a very long time, and that’s, I think the kernel of what will come next will come from that unlocking of ideas.

[MUSIC PLAYING]

Claer Barrett
Mmm. Well, Ben, I wanted to ask you, as a historian, about hype cycles. Because this is the other big risk here. A hype cycle is a phenomenon where the market overestimates the short-term potential of something like this new technology, and underestimates perhaps its longer-term potential. Now, you started your investment career as the dotcom bubble burst. There has to be a danger, doesn’t there, that AI could be another?

Ben Rogoff
There are lots of banana skins and pitfalls about investing in technologies that potentially change the world. And again, without wanting to give investment advice, that’s probably why one does it in a diversified portfolio, because it does allow you to make mistakes and be able to sort of take broad side and still come through. So could it be a bubble? It could. I mean, there are a few things that make me feel that it isn’t.

Claer Barrett
OK. Tell us, Ben. 

Ben Rogoff
Well, I think the starting point is that this is a profoundly different way of computing. Maybe look at the various different modes of transport and say, well, they’re all transportation modes, but they’re fundamentally different. But if you believe, like we do, that we’re moving into this new kind of parallel form of computing or accelerated computing, then, yes, you know, you’re looking at something that looks similar to what’s just past but is architecturally different. It would be like looking at a Toyota Prius and a Formula One car and concluding that they were the same. But of course they are, but they’re also fundamentally different. And if we’re moving to a world where AI will infuse everything, which it must, because in the end, this computer is able to do stuff that humans cannot. It may not be as accurate yet. There may be lots of things that are not yet solved in AI. But like you said at the beginning of the chat, there are some incredible positives already coming out of an industry that really ChatGPT was only out in November 2022, and we are already at a point where Google’s latest model is able to hit human performance across 52 different subjects. Not tasks, but subjects.

So we know that this is going to infuse everywhere. We don’t know what quite that form will take, but we can say that we’re at the beginning of something. The point I’m trying to make is that this . . . we’re really in the foothills of building out a new infrastructure, which, by the way, has an incredible reach in that 50 per cent or thereabouts of the world’s work is knowledge work. And that’s the ultimate scope for AI. It’s a $44tn market opportunity at face value. And again, please don’t . . . I know I’m sounding, I’m like I’m prophesising here, but I’m trying to put it into context of we’ve really only had three or four quarters of the growth that Nvidia has been able to enjoy, and that feels very alien and premature to call that a bubble.

[MUSIC PLAYING]

Claer Barrett
One of the other big things that could slam the brakes on AI, of course, is regulation. Now, that’s a big unknown. You’ve spoken about how sovereign states are worried about the impacts on language. I mean, lots of us are worried about the impacts on employment, job losses, but also how AI could turbocharge fraud. That’s something that as a journalist I often write about. Now as an investor in AI, how do you feel about all of that?

Ben Rogoff
I think that regulation is an interesting one. I think it’s a sort of double-edged sword. It has this clearly a negative, you know, in the form of things like export controls. Is Nvidia allowed to ship the latest chips to China? It’s not anymore. And in this last quarter, China fell to a sort of single-digit share of revenues because the US has said it can no longer export the very fastest, latest generation of chips. So regulation certainly has a risk there. There’s risk in terms of having to protect people from, like you say, fraud, from deepfakes. Regulation also as a tailwind. That’s an interesting one, because if you look at the experience of aviation, early in aviation, aviation was fraught with danger. The accident rate was much higher than it is thankfully today. And actually, regulation makes flights safer.

Claer Barrett
Yeah, safer than driving a car.

Ben Rogoff
Exactly that. Sam Altman at OpenAI is explicitly calling for more regulation, because regulation will help people embrace AI in a way that has indemnified them, or at least better indemnify them against some of the risks. I think there are others, you know, the monetisation timeline. What if the productivity gains that we think you’re getting from co-pilots don’t necessarily persist? Other risks might include model performance and how that this rate of improvement that I referenced, what if it slows? That to me is the single biggest risk to the AI story, that if we get to a point where models are 97 per cent accurate and can’t get to 98 or 99, at which point some of those larger market opportunities take longer to unlock.

It feels a little bit like, again, I wasn’t . . . I mean, I was young at the time, but the nuclear arms race was something that happened. And then after the frenzied period of building more weapons than anybody ever needed, the largest superpowers are able to sit down and actually come up with a framework for this. And I think the frameworks will come later. I think right now we are at the beginning of an arms race because the capability of AI to transform very large parts of economies across the developed world particularly, means that the governments and policymakers will talk of probably a tougher story than they’ll actually deliver, because nobody will want to fall behind.

[MUSIC PLAYING]

Claer Barrett
So Ben, let’s come back to some more hardcore investment questions about your portfolio and your strategy. Innovation — we’re not short of it in the AI world, but is it coming from those large-cap companies or from start-up companies like Mistral?

Ben Rogoff
Interesting question. So we focus exclusively on listed companies. And so we obviously keep an eye and do work on private companies to make sure that we’re not blindsided, but also because we need to really be up with cadence of innovation. And of course, some of that’s happening in the public gaze. And lots of it’s happening in the private market. So we know that the privates have got some . . . there’s some super, super exciting companies coming through. We look for a better IPO market and some of the sort of maturation of those businesses because, sure, they can be very exciting, but it probably is still relatively early. Given where OpenAI is, as a leader with 2bn of revenues. More broadly, so far at least, the last, well, almost forever, it’s demonstrated that the post-internet world has lent itself to large caps.

So my default here is that AI is really going to benefit disproportionately large companies and certainly unlisted markets, for much the same reason that the internet has been a benefit to those very large companies. Scale, data advantage. You know, the cost of delivering AI workloads is not immaterial. This is expensive stuff. You know, an AI server is 25 times the cost of a general-purpose server. You know, again, Formula One to Toyota Prius. And that’s why Nvidia’s data centre revenues have exploded. Because they’re not just selling AI servers, they’re selling servers that are 25 times more expensive than the ones they weren’t selling in the general compute world. So I think that large cap should do very well. But I think for listed companies that are small, that don’t have large proprietary data sets, this is going to be a challenging time, actually. In much the same way that when the telegraph, back in the 19th century, slightly obsessed by leaning on history to help . . . 

Claer Barrett
You are a historian, we’ll forgive you.

Ben Rogoff
Amateurish. Two. One. So I wouldn’t necessarily describe me as very good on either. But again, I very much believe as an investor today that much of what we experienced has been experienced in slightly different form by previous generations. We all think what we’re going through is unique, but of course it isn’t. The telegraph was a fascinating development that, you know, the precursor for the internet, really. And one of the things that it did — it did lots of stuff — but one of the things it did was really hurt regional markets because information flowed in a way that it never flowed before. And I feel that has the same sort of potential ramifications for AI on smaller markets, on smaller companies that have been operating within niches that can now be unlocked and identified by large companies that can bear the cost of AI in a way that smaller ones are going to struggle with.

[MUSIC PLAYING]

Claer Barrett
Well, interesting we’ve been talking about large companies, those large caps which feature in your portfolio. I mean, there is an argument here: they’re so large and so dominant, they already account for a frightening percentage of global index funds as it is. Now, you’re an active manager. So your proposition to an investor is: you pay us a higher fee, we’ll pick the stuff that will do very well. I mean, there is an argument that because these funds are so big, the index investor who pays much smaller fees is going to have access. But there’s also another question that springs to mind, which is: could a robot take over your job as a fund manager in the future? Tell us your thoughts.

Ben Rogoff
(Chuckles) Oh, dear. Probably. Yes. Well, let’s do the easier one first. So in terms of our proposition, I think it’s very fair. I think we’ve been operating in a backdrop where active management as a percentage of funds under management has been shrinking. Candidly, if this was to continue, if only a handful of stocks are going to rule the returns in the world going forwards, life will be tricky for active managers. I think that when you look at concentration, generally it’s something that we don’t like, obviously, because it introduces a new risk to portfolios. If you look at concentration in the market, I think we’re back to sort of almost like late 20s, 1920s-type concentration, which isn’t necessarily awful news, but it tells you how far we’ve come in terms of this trade. So at some point you will need someone to help you out of these stocks in the same way that IBM was a very investable technology company for a very, very long time. Warren Buffett owned a good chunk of that business, and we were anxious about IBM from, you know, cloud, really what the cloud might do for that business. And we slowly unloaded our position and then we exited. And that will be the template for how active managers like ourselves that have been involved in these very large companies will ultimately, that’s the template, will deploy. So at some point Google for example, may not be on the right side of the AI trade. It’s possible that smartphones could be at risk from a new form of compute.

Claer Barrett
And that would threaten Apple.

Ben Rogoff
That would threaten Apple. So Apple is now, you know, a meaningful . . . absolute position for us, but one of our largest underweights. Google is let’s say 5 per cent of our portfolio. But again a meaningful underweight against the index. And so you start to flip from being the biggest companies, natural monopolies, basically nonfungible assets. If you want to be involved in smartphones, you’d probably be in Apple because Apple makes 85 per cent of industry profits. And if you’re excited by advertising, you’re probably going to be in Google because it has 90 per cent market share. But at some point, history informs us that incumbency becomes a challenge because you have old markets and big profit pools to protect. And AI, as a GPT, as a general-purpose technology, will challenge those profit pools. It must. So it’s been tricky for active managers, ourselves included. But at some point the market must broaden. And we know that AI is going to create disruption. And disruption creates opportunity. It also destroys value. And understanding the timeframe for those dynamics, I think, is going to be critical in the coming years.

[MUSIC PLAYING]

Claer Barrett
Mmm. While we’re here, could you briefly just tell us about the trust’s performance and how you currently are performing against those benchmarks?

Ben Rogoff
I can. So we’re really pleased with the absolute returns we’ve delivered. Relative to our benchmark, which is the Dow Jones world technology index, most of our peers have struggled to keep up with that index, because the index is dominated by those very large companies. So we’re delighted to have had as much of those names as we’ve had. We’re in the top quarter or the top half of competitive funds across all timeframes. The longer you go back, the better we look. And that sounds like an excuse, but more the value of compounding. If you’re always in the top third or top half of the peer group over short timeframes, you end up at the very top over longer timeframes.

Claer Barrett
Mmm. Now we’re sitting here chatting in the FT studio in London. Does it pain you, Ben, that UK-listed companies account for less than 1 per cent of your portfolio right now?

Ben Rogoff
Pains me in a sense that it would be great if we had a larger listed universe of UK technology companies for me to invest in. Candidly, it was an observation I made many years ago that we just didn’t have that range of companies. And I think there’s a very dynamic technology community in the UK. I know that Jeremy Hunt and others have talked about wanting to get Silicon Roundabout to this world and trying to turn it into a technology hub. You know, I think there’s tremendous talent in the country. I think that we have very little to play with as investors in the listed market.

And I think that informs us, ultimately, we also struggle a little bit with liquidity, not because of our size, but because the underlying UK equity market is pretty illiquid relative to other markets in the world. So we obviously are very heavily pointed towards the US, sort of somewhere between 70 and 80 per cent of the portfolio at any given Sunday. But more recently, actually we’ve been investing in Japan and in Taiwan, notwithstanding, you know, risks of investing in other markets. But this idea of a new architecture for AI, every single element of the Formula One car is different to the Toyota Prius. You know, they’ve got tyres and wheels and they’ve got all the nuts and bolts, but every single bit of the car is different. And those bits and bobs for an AI server are all high-performance bits and bobs, and they’re made in Taiwan and they’re made in Japan. There are dedicated companies in the investment market that can build entire portfolios around the UK opportunity set. It’s just that we see bigger and better opportunities elsewhere.

Claer Barrett
So, Ben, before I let you go today, anything else you want to share with our investor listeners? 

Ben Rogoff
Well, just maybe one last thing. So we’re super excited, as you can tell, about the AI opportunity. We think we’re relatively early. And of course there’ll be cross-currents along the way. But nonetheless, we’re super excited about the AI opportunity. I think I wanted to share, I suppose, is this idea that we’re not alone and that actually some of the smartest people in the room, as they say, are acting in ways that are slightly odd. And I think that’s interesting.

Claer Barrett
Tell me more.

Ben Rogoff
Well, say, for example, Geoffrey Hinton — who had been at Google forever, he’s referred to as the “godfather of AI” — leaves Google, Alphabet, in order to, quote, be able to talk about the dangers of AI without any negative connotations for Google. That’s unusual. We had in December last year, the CEO of OpenAI, Sam Altman, losing his job and then being rehired in the same week. That’s very unusual. I’ve done this for 20 years, I don’t recall that ever happening before.

Claer Barrett
These were dramatic moments in the FT newsroom, also. 

Ben Rogoff
Well, there you have it. And then Sam Altman in the public domain talking about the idea of raising $7tn to build AI chips. Again $7tn is twice the size of UK GDP, I think seven times the size of the entire semiconductor market today. So again: odd. At face value: odd. So what’s going on? We’ve also seen Mark Zuckerberg talk about building systems for general intelligence. So the point really I’m trying to make here is that the focus today is on co-pilots — software that make us more productive. But I think when you look at the way that some of the smartest people are acting, they’re acting strangely. And if I was around at the time of the gold rush and I had found a seam of gold, I wouldn’t be telling the world I’d be buying picks and shovels as fast as I could before you realised what I had found. And my gut feel here is that these type of data points that I just referenced are unusual, and they to me suggest that the very smartest people understand that the rate of change that we’re seeing, the multilevel, predictable rate of change of AI, is going to get us to some human-equivalent general intelligence, as people call it, much faster than the market thinks. And I think that the actions and unusual actions of these brilliant people suggest that the pace of change is faster than most are modelling.

Claer Barrett
Well, Ben Rogoff, it’s been a great pleasure having you speak to us in the studio today. He knows if we conduct this interview in a couple of years’ time, maybe I’ll be replaced by a robot. But then so could you. (Chuckles)

Ben Rogoff
Indeed.

[MUSIC PLAYING]

Claer Barrett
That’s it for Money Clinic this week with me, Claer Barrett, and we hope you found this episode useful. If you did, spread the word and leave us a review. We’re always looking to chat with people about their money issues for the show, so if you’re interested in being part of a future episode and are looking for some expert money advice, then email us: money@ft.com. You can also follow me on Instagram. I’m @ClaerB.

Money Clinic was produced by Tamara Kormornick and Persis Love. Sound design was by Breen Turner and our editor is Manuela Saragosa. You heard original tunes this week by Metaphor Music. Cheryl Brumley is the FT’s global head of audio.

And finally, our usual disclaimer. The Money Clinic podcast is a general discussion around financial topics and does not constitute an investment recommendation or individual financial advice. For that, you’ll need to find an independent financial adviser. That’s all the small print for now. See you back here next week. Goodbye.