Polar Capital Technology: best and worst AI stocks

Polar Capital's Ben Rogoff sits down with ii's Sam Benstead to discuss how his £4.6 billion technology fund invests.

The discussion includes recent performance, the impact of artificial intelligence (AI) on the economy, developments in AI technology, and what changes Ben has made to his portfolio.

Rogoff also reveals who he believes will be the winning and losing companies due to this new technology breakthrough.

• Our Services: SIPP Account | Stocks & Shares ISA | See all Investment Accounts

Sam Benstead, fixed income lead, interactive investor: Hello and welcome to the latest Insider Interview. Our guest today is Ben Rogoff, who is the manager of Polar Capital Technology Ord (LSE:PCT) Trust. Ben, it's great to have you in the studio.

Ben Rogoff, manager of Polar Capital Technology Trust: Thanks for having me.

Sam Benstead: We last had you in the studio about two years ago. So, how has the portfolio performed since then?

Ben Rogoff: Well, it's always nice to be able to report on a very strong period of performance and it has been. The technology market's been strong and the trust has been stronger still.

So, over those two years, I think trust net asset value [NAV] is up around 78% and our index is more like 66%, but they are remarkable returns, really, over any given two-year time frame, reflecting a strong backdrop for technology, but specifically for AI.

Sam Benstead: What have we seen over these two years in terms of the technology space? What are the key changes?

Ben Rogoff: It's a good question, and what I probably should have said earlier for the returns part of the question is that we're really delighted with the returns we've delivered, not just because we've got the AI theme correct within the portfolios, but it's been a really difficult market for many growth investors.

It's been a very narrow market. Obviously in our world, the Mag-7 has dominated, and I think small-caps have meaningfully underperformed, [and are] only up 20% over that time frame in the US. So, it's not been an easy market.

To your point about what's gone right and what's changed, the answer is that AI has really continued to make very significant progress and it's done that on a number of different vectors. The most important one is that models have got better.

If you go back to November 2022, so outside the two-year window, but really what started this was obviously ChatGPT at that point. You had this incredible quarter from NVIDIA Corp (NASDAQ:NVDA) in May 2023 where the company just blew away in numbers, but really the guidance they gave was record guidance. No one had ever seen an upside to price like that, and that really marked a different market, and that's mostly captured in those two-year things.

So, what happened? Yes, the models got better. ChatGPT became GPT-4. We've had competition from lots of other companies at the frontier of AI. We have seen capex numbers, the ultimate barometer, really, on how AI is doing and how people feel, these hyperscalers, how do they feel about the progress that AI is likely to make beyond what we see today?

The answer is that it's in the spending numbers. This year the industry is supposed to spend more than $370 billion (£272 billion) on AI infrastructure. Those numbers are wildly different to where they would have stood two years ago. So, we've got AI progress, more users, up to 800 million weekly users at OpenAI.

All the vectors that would chart the progress of AI have been steeply up to the right, which has really created a backdrop that has been fabulous for AI-related assets, but pretty tricky if you're not on the right side of that trade.

• AI demand generates ‘phenomenal’ quarter for tech firm
• Don’t chase AI – back the companies using it best

Sam Benstead: And how has your portfolio been moved over the last couple of years? Well, what you've been buying and selling?

Ben Rogoff: When we first saw ChatGPT, we were pretty excited by it. There was quite a lot of scepticism in the market, and not just in the market, but in the wider world, where in the end [people were saying that] AI just makes stuff up, it's probabilistic, and some people described it as a Victorian parlour trick, but we felt very differently about it.

Just for background, we launched our own dedicated AI fund nearly eight years ago, it was called the ‘Automation and AI’ fund when we first launched it [now called Polar Capital PLC-Artificial Intelligence], so I suppose you could argue that we've had a head start.

We've been looking at AI for a very long time. We were super-excited about the Transformer models that really are the underpinnings of ChatGPT, but Gemini and all the leading models that you see today are built on Transformers. We felt that the trajectory of AI was going to be profoundly changed. We did some small stuff between the launch of ChatGPT and that Nvidia quarter that I referenced, but after that we pivoted very hard.

I've been running the trust for a long time now. We've built it over time as being a very diversified six to eight core themes and trying to deliver more growth in the index, but actually that changed. After that Nvidia quarter, we feel that AI is a next general purpose technology. We've felt that now for quite a while. The market's coming around to our view. General purpose technology has changed the world.

We may talk about that more as we go on, but how can you have a portfolio that doesn't want to capture moments like this? These are highly unusual moments of discontinuous technology change and so we have meaningfully shifted the portfolio, really doing lots of work on the new AI infrastructure stack.

One of the things I should probably mention is that we've just come off the back of a wonderful period for cloud. Cloud was about the commoditisation, the homogenisation of compute. No one cares what the server is, the PC that you use, because in the end, it's all about the cloud delivery. That's not the world that we're in today, where hyperscalers and others are building performance compute that require the fastest chips, the densest clusters, the lowest latency.

We're building a very different compute infrastructure stack than almost anyone has seen in the market today and so we've pivoted very hard towards that. We can talk about some individual names if you're interested.

On the other side of the equation, we've been meaningfully reducing exposure to the old tech. For us, old tech really looks like software, which have been the most incredible conduits for super-normal profits for a very extended period of time. Most people, myself included, have been brought up on the primacy of software businesses.

Those businesses feel at risk, and again we can talk about that if you're interested. We've been reducing software. We've been reducing anything really that doesn't have AI exposure in order to give us a really good chance of capturing the upside of this new cycle.

Sam Benstead: Let's go into companies then. So, where are you putting money to work in these AI winners that you've mentioned, and which companies are you taking capital out of, as you don't think they can survive in this new world?

Ben Rogoff: The portfolio is being relatively actively managed. We're moving the portfolio almost every day because the pace of change within AI is really very frenetic.

The core of the AI story really has been about Nvidia, and Nvidia producing graphics processing units (GPUs) that can be used to drive this new form of compute. Nvidia is trying to release chips on an annual cadence, and around that, companies are trying to configure compute clusters around those new chips.

On top of that, you've got hyperscalers that are the Googles and Meta Platforms Inc Class A (NASDAQ:META) and others that are trying to configure their own form of AI cluster.

The point we're trying to capture is that this is very fluid. Companies that you think are involved in the latest iteration or the frontier of AI may not be in the next one, and that requires a lot of work, candidly.

We're a team of 11. We do 1,000 company meetings a year. We're working harder than we've ever worked because we feel that the opportunity set is as good as it's ever looked.

So, what are we doing in the portfolio to capture the AI story? It really does depend on where we think we are. For example, right now we've still got a very large position in Nvidia, but we augment that with companies in the chip world, in the memory world, semiconductor equipment makers, but also networking, which is an area that most people haven't looked at for a very long time.

If you recall, networking companies were very much the way to capture the dot-com period because we were carrying internet traffic that didn't exist before. That became a very difficult thing to do, obviously, post-dot-com bust.

But, today, the architecture of AI requires these incredibly dense clusters, compute clusters, and that requires incredibly fast networking, both within the data centre and outside it when you connect data centres in order to train AI models, particularly, but also to carry some of this incredible traffic that we're seeing already in AI in inference volume. So, networking is very much back in vogue.

We have also, and I think this is quite different to both the index and to our peer group, embraced companies that are in the power and cooling world.

Power is a critical bottleneck today, and if you extrapolate out, the needs for power in an AI-centric world are going to be very difficult to meet. I suppose that power consumption has been growing at half a per cent per annum, and there are estimates out there that say we might need to be 10 times that in an AI-centric world.

So, we've been moving the portfolio to try and capture where these bottlenecks and super-normal profit windows exist.

On the other side of the equation, we're beginning to, I suppose you could argue, disinvest from some of the Mag 7 names. Not all of them, there are some incredible assets there and they're a remarkable bunch of companies. But Apple Inc (NASDAQ:AAPL), [and Alphabet Inc Class A (NASDAQ:GOOGL) or] Google, feel less good conduits to us for AI, I should say.

We've also meaningfully reduced our software exposure, as I mentioned earlier.

• Still time to buy into these two tech winners
• Sign up to our free newsletter for investment ideas, latest news and award-winning analysis

Sam Benstead: Nvidia is your largest position at about 11% of the portfolio, I believe, and it's gone up 300% over the past two years. Why do you still like it after such phenomenal growth?

Ben Rogoff: It's definitely not outrageously expensive and I think my whole career, really, has been dogged. It's probably the wrong term, but for most of my career the US tech sector hasn't looked cheap and I think there are a lot of people who wish they'd been involved in what has been one of the best parts of the market for an extended period of time because optically some of the best technology companies on earth [do] not look cheap.

Google [Alphabet] came public on 35 times earnings and just how good has that company been as a conduit for a return?

So, Nvidia is not too expensive to own. It is the company at the epicentre of AI. It is the GPU supplier, but increasingly the connectivity supplier, and [it is] trying to create, basically, AI in a box, and defining the industry. Again, we have to have exposure to companies that are defining, three years in, a new industry.

It is profoundly different industry to the cloud industry, or the cloud compute world that we're leaving. So, we still have it in the portfolio, we don't see it as expensive, but you might argue that at this scale it's very difficult for the company to keep beating estimates in a very large way.

I think that's a fair pushback and it's a very large part of our benchmark. It's around 12% of the portfolio now, which is towards the higher end of where I want any individual position to get to really, just from a risk perspective, as investment is full of unknown unknowns. You can't know where there might be a challenge to that business.

But fundamentally, we've continued to really like the business. Going back to that idea of it being too large to keep upside surprising, well, people were saying that two years ago.

Having said that, if you think about it as the kind of core, the largest individual name that we own in AI, we actually have more exposure to AI outside Nvidia than we do in that individual name.

Sam Benstead: On this AI theme, where are the hidden winners and losers that other investors might not be paying attention to?

Ben Rogoff: Obviously, I'm going to be careful about how I answer this because the team's doing an incredible amount of work. Some of the names that are hidden, we're happy to share, and there are others that we'd [rather not], I suppose. Colonel Sanders or Coca-Cola don't want to share all of the recipe with you or your viewers today, and hopefully everyone understands why.

But, as it relates to the infrastructure side of this story, where most of our exposure is today, we have exposure to companies that are probably well known but perhaps less well-held as AI conduits. Obviously there's Nvidia, but Nvidia is competing with Broadcom Inc (NASDAQ:AVGO). It's a fairly well-known mega-cap building application-specific integrated circuits (ASICs). So, custom chips designed to do some of what GPUs do for the very largest AI companies.

Advanced Micro Devices Inc (NASDAQ:AMD) is another name that we've owned for a very long time. We like its positioning as a sort of number two/three alternative. You know, the Googles and Metas are going to want to have second sources to Nvidia, obviously.

So there's compute, there's memory, you need high bandwidth memory to build these new clusters, so we have exposure to some memory stories, SK Hynix, Micron Technology Inc (NASDAQ:MU).

Let's talk about networking. So, there's companies like Credo Technology Group Holding Ltd (NASDAQ:CRDO) or Stereolabs that are involved in helping those hyperscalers create these more custom configurations. Those names are probably less well known to some of your viewers.

We also like companies that are involved in providing the optical componentry that helps you to link data centres because power is scarce and if I need to access a certain amount of power to train an AI model, I may have to link smaller data centres together and that will help companies like Corning Inc (NYSE:GLW), the fibre optics business. Also companies like Ciena Corp (NYSE:CIEN) that have been optical networking for the last 25 years.

So, there's a flavour at least of some of the names that are perhaps less well-trod in the AI world.

• The tech funds that have actually beaten the Nasdaq
• Is your global fund really a play on the US?

Sam Benstead: And just how important could AI be for the future of the world? What could it look like in 10, 20, 30 years time? And what opportunities are there?

Ben Rogoff: It's huge. It's a general purpose technology. Steel, electricity, the internet, they change everything. I mean, that's not true. Of course, that is sort of hyperbole and nothing changes everything. But we believe, and we've believed for quite a while, that AI is likely to prove the next GPT.

If you think about what steel did, it starts off just as a better way to make rails for railways. But very quickly, you start building buildings differently. Cities are created and tin cans, razor blades and lots of other things. There are cars and home appliances made possible by cheap steel.

So, the world has profoundly changed by GPTs. Specific to AI, we're already seeing some very significant changes. We're seeing the disruption start to show up in stock markets.

You can see it with Meta talking about being able to create advertising campaigns for their customers, and look at the damage that's done to ad agencies.

Look at some of the weakness in information services companies as AI capability is reaching a level where the proprietary data sets that exist may not be quite as valuable as people thought they were because now these models are able to interpolate and start to potentially substitute for these proprietary data sets. So, we are beginning to see it. Software is also falling foul.

But back to the question, the disruption is going to be very, very significant. I wrote a piece years ago forPCTN report around containers and container ships. It was about the container box that transformed global trade. If you were a dock worker in Liverpool or San Francisco, the container ship was a disaster. You lost your job and the port became obsolete. The people of Oakland were perhaps less unhappy because they picked up the traffic. But in the end, global trade explodes.

AI is going to create winners and losers for sure. There are going to be people on the wrong side of it. There's going be a lot of disruption. What we would describe as white-collar work is going to be meaningfully disrupted. But I tend to, as I have done consistently in my career, think glass half-full. I'm an optimist.

I think that in the end AI is a phenomenal democratiser. We're all going to have super-intelligence in our pockets. Tesla CEO Elon Musk is talking about 2030 for super-intelligence. If that's true, even if it isn't, we now have access to information sets that were reserved for kings in history. I'm a huge believer that it's going to democratise.

One last thing, if I may. People tend to always focus on the negatives because the negatives sell newspapers or whatever the digital equivalent is today, and because change is scary. But, think back to, say, the 19th century - I've written about this one as well in previous annual reports - [and] agricultural mechanisation that really transformed farms.

There was a thing called a McCormick Reaper. It gave way to the phrase, you reap what you sow. Harvest time requires getting as many people as you can together, because if you don't get the crop picked, it rots, it wastes. So, McCormick's Reaper was a small thing that you added to a horse and it transformed the harvest. [It] would have been bound for people, farmhands, and it would have great for farmers. But, more importantly, 50 years later, it had essentially eradicated peacetime famine.

So, that's what gets me excited. People are going to focus on the negatives of AI, the disruption, the things that are going to changeour daily lives. But in the end, I think it's going to unlock massive positive externalities.

Sam Benstead: Ben, thank you very much for coming into the studio.

Ben Rogoff: Thank you.

Sam Benstead: And that's all we've got time for today. You can watch more Insider Interviews on our YouTube channel, where you can like, comment, and subscribe. See you next time.