Polar Capital Technology Trust plc (PCT) Earnings Call Transcript & Summary

Okay. Good afternoon, everyone, and welcome to Polo Capital Technology Trust. Just I'm Catherine Cripps, and I'm the Chair and I've been the Chair since 2022 for a year previous to that. Before we get started, just a couple of housekeeping points. In the unlikely event of an emergency, please follow the fire sign is at the back. I don't think you behind us and take -- follow the instructions that the Herbert Smith people will give you as you leave the room. May I ask also that you switch off your mobile phone or at least put them on silent so that we don't disturceedings. Thank you. So thank you to you all for making the effort to attend what I'm sure will be an interesting and informative session. We've also got a Zoom video link enabled. So there are some people joining us virtually. They can see and hear us, but we can't see them. They may well pass some questions through to my colleague, Tracey, over in the corner there later on. We will start with our managers, Ben and Ali giving us a presentation, which I know is the real reason here. They will sorry, Kelly, our very efficient co. just remind me, I need to declare that it's now after 2:30, and we are and we can declare the meeting open. Sorry about that. I was so keen to get on the moving. So there will be some formal business coming after Ben and Ali do their presentation. And also, they will leave some time for questions from you in the audience to ask. So please do part in questions. We try as a Board question them quite a lot. So -- but other questions always appreciated. So well, before they start the presentation, probably best just to explain who we are at the table here. These are my fellow directors. So Kelly, I've already referred to here is not direct, but she's Jane is the Chair of the Audit Committee and oversees the production of the annual report. We've got Tim and he is until the end of the day, and we've got Stephen next up, obviously, Ben at the end there. And then Charlie and Adiba to my left here. So we are also here all afternoon and afterwards for questions. And you will probably introduce some of your team, but they are not on the board, but they are in the front row here. So they will be present afterwards when we have to discuss any of the nuances of the investment pro.

And any difficult questions?

And any difficult questions reflected. So without much ado, I'll hand over to Ben and Ali to lead us off. I think actually, Ali is going to lead first.

Thank you very much. And thank you, everyone, for coming this afternoon. It is absolute pleasure to see you all here as always. We will try and take you through a few thoughts, a very little bit about performance last year and what drove that. But most of our time will be spent on the future, which is what we're all excited about and what we're investing in. So it was a remarkable year, I think it was the word we used for PCT in FY '26. It's fair to say that results were somewhat flattered by a depressed starting point. If you remember, April 2025 was liberation days, the market sold off pretty heavily, and we benefited from some of that. But still, this is a fantastic year for absolute relative returns. I think probably the best year in the Trust's history and certainly on a relative basis. And we'll go into some of the reasons why that is. Again, just to remind everyone that our performance is no guarantee of future returns, but we're obviously delighted with having stand out last year. And for us, it was -- came at the end of some years of strong performance versus our peers, but maybe some tougher ones versus the benchmark as it was dominated by a small number of very large positions. We feel like we got all that back and more last year and to a lesser extent, the financial year before. Since then to bring us a little bit more up to date. We've given a little bit of that back in relative terms, but actually had a better September. This is only through the end of August. And much of that was a function of a difficult difficult July when we saw a real drawdown in tech momentum, some stocks that were up a lot in the first half of the year -- or the first 6 months of the year, then pulled back really quite sharply. You've seen interest rates moving higher, the macro environment has become a little bit more challenging. But the trust has continued to move higher. We just put that in some context over prior periods. We're extremely proud of the performance in the first decile for all of the longer periods in question, and then there's some of the discrete annual performance there as well. So this is almost ancient history now, but what happened last year. It was a good year for Global Equities. We've got the third year of strong gains, and that was driven by all the things you'll be familiar with strong global growth, central bank rate cuts and ultimately, AI CapEx, and that was really where our portfolio was most focused. Tax stocks led again. I should point out that this charge is only to the end of June. And since then, small caps over large cap has actually reversed a lot of that -- part of that sell-off momentum obviously took down a number of small cap positions as well. But we've only got 1 slide on what happened last year because we're spending most of our time thinking about the teacher, but it was really a market all about AI. And if we've sort of start from the bottom layer, if you like, AI model progress, which we see as the animating force of the AI trade of the market even at the moment. AI models continue to get better. And they got better, we think, at an accelerating rate. More models came up more frequently, they were more capable, the jumps are getting bigger. The amount of capital and compute needed to train and to serve these models also grew exponentially. With that, brought very, very fast AI adoption much faster than people were expecting. I think some degree including us. And we saw an extraordinary growth in the revenue of some of the AI labs, which I'll come on to a little bit later in the presentation. This was reflected in higher AI CapEx expectations, in a really, really material increase in what we already some very big numbers as companies invested more in training new AI models and then in the infrastructure to serve them at imprint time. Our performance, if I sum it up, was strong because of our AI maximalist positioning. We positioned the portfolio, and we moved it around a lot, but the common theme was we were trying to benefit from upside surprises to AI progress, upside surprises to AI adoption and upside surprises to CapEx. And all of those things happened in spades in the last fiscal year. We were materially underweight software, which has been on the sort of the loser from this trade or the perceived loser from this trade, and we haven't rebuilt that software better in a meaningful way. We've added back some cybersecurity and some infrastructure software exposure, but we are still relatively bearish on application software. And we benefited from being underweight the MAG 7. Now that's not just a homogenous group, but we have been beating a retreat from that group of companies because as they are having to make these huge capital investments they are becoming -- more 2-way debates are emerging on those companies. There is a very strong bull case on each of them from AI, but there is also a much more material bear case than there was in a pre-AI world. So our base position has been to own less of those types of businesses. This is the team. Some of you are here today. The rest are traveling banknote places around the world at the moment. We continue to invest in the team. You may remember we talked to Pat joining us a couple of years ago to invest in our AI function. I hope there'll be some questions about that, because we really think we've made some material progress in our own use of AI in our investment process over the last 12 months, some of which has reflected the improvement of models, some of which also reflected Paddy's very hard work of driving AI through our team as well. You know who we are very well. But in case you don't, I think the -- 1 of the key differentiators in recent years has been the adoption of an AI lens. And this is a metaphor we've used to try and explain to our clients and our shareholders. The first thing we think about when we are looking at any company that may enter the portfolio or that is already in the portfolio is where is that company positioned in an AI first world. Because if you think AI is the next electricity or the next deal you wouldn't want anything in your portfolio that was on the wrong side of that. That is our mindset. That's how we're approaching this sector. We think we've got a unique or a differentiated approach to technology disruption, a different framework of thought -- and we think that our history of actually selling things is just as important as buying. Everyone likes to talk about the stock port that have gone up, actually, a lot of dealing with technological change is moving away from things that are falling on to the wrong side of technology as we would see it. This is the long-term performance to move past because I want to talk about the story now. When there is so much happening on the news flow is coming thick and fast with new models and exciting new developments. It can be quite hard to work out where we are in this story. So what we thought we might do is to try and give you some context and set out a story in sort of 3 phases. And the first phase, which really came to an end towards the end of last year, we would see as the Phase I, the proving out of a new technology. So if we went back 3 years, would have been talking about ChatGPT as the iPhone moment for AI. At the moment when we realized that something had changed. So there was a new form factor on top of the technological progress that have been on the need and that this was likely to open up new markets. What we didn't have at that point were business models, revenue, a sense of what this would look like as it matured. And we were focused on things like academic paper citations, very, very early developments and capabilities. but there wasn't actually much to look at in terms of business at this point. A year later and the debate has moved on somewhat. -- model progress has accelerated, we think, in 2024, although that wasn't necessarily obvious at the time. The arrival of reasoning models that could put together different things and come to conclusions that could work in a chain of thought was a pretty key development in the space. We spend a lot of our time focusing on the new compute stack that this AI revolution would require. And we liken that we've differentiated that from cloud computing, which we saw as the sort of Toyota Prius model of the world, which is kind of good enough at most things, not amazing at anything, utility like compute moving to a McLaren F1 version of compute accelerated, very performance, very expensive and a different thing, even though they both have 4 wheels and a steering wheel. That's what we thought the focus was on there. A year ago, we had our first major challenge to the story in the form of Deep Seek and the Deep Seek moment, which we and other investors have to very quickly take a view on whether this was the end of the trade because there was a model from China that was able to do everything cheaper about a faster, and that was going to destroy the CapEx trade about whether this was really competitive with the models from OpenAI and Anthopic and Google or whether it is complementary. And I think the evidence time has been very firmly that it is complementary for now, but that wasn't immediately obvious at the time. We started to see new form factors of research products. We started to see very large infrastructure commitments. I mean, $500 billion was a lot of money then. And it's now -- the industry will be investing $800 billion just this year and well north of 1 trillion next year. So things have moved very, very quickly. Our focus remains on model improvement on this idea of Jevons paradox that as you make the use of something more efficient, you use more of it, not less. I think that's been a critical concept to get your head around, and it's not totally intuitive. But for us, the real story of the past 3 years of this Phase 1 has been accelerating AI model progress and there have been regular pushbacks to whether there is a wall in model progress, whether we've run out of data to train on, whether there's enough power or compute, and we can certainly get into those debates if you're interested in Q&A. But for now, the chart on the bottom right has been what we think is sort of called the end of Phase I. And this is an index measure of model hallucinations. And if you remember the early models and if you ask them something, there was a certain portion of the time that they would make up an answer or it would hallucinate something because these are probabilistic models. A human score on this benchmark is about 5 to 10 and you can see that as generation of models have progressed, we've gone from being much less good than humans to better. And that is the moment we are now arrived at, which has brought, we think, the advent of agenetic AI.

Thank you very much. I just have to adjust for about a foot-high difference between us afternoon all. Thank you so much for being here. And yes. So thank you very much, Ali, to -- so yes, it brings us to Phase 2 of the story. And you'll know from annual report to go, that the front cover of our deck was much more polarblue with Neon -- and it was really about the idea that we were excited about agent. The story of 2026 is a genetic AI and the arrival of it and what it means and how it's heralding a completely different phase for the AI story. In fact, probably the best way to think about it is the idea of nonhuman scaling. In the first world, and in perfect AI, our interface with AI is via a chatbot in the same way that we would have interfaced with computing by a Google search bar agenda is about applying AI, a human on 1 end, many machines on the other end. And very much the best way to imagine that world is to kind of harp back to agrarian or an industrial revolution. Where the relationship between labor and capital is profoundly altered. So what we think comes next. Well, as Ali said, the model performance is already proven. There's lots of debate still being had by people that don't like where we're at in the world. But the reality is models are now well beyond human capability, at least an individual human tasks, much like the car was in 1900 and what we would expect to happen now is that gradual to suddenly think that is just so hard for humans to grasp, but here is the same street, years apart, no cars, no horses. Now again, this likely option curve is going to be very, very fast and furious, but it's worth remembering that in this period before -- well before the Wall Street crash sort of ended that excitement around the automotive industry, I think something like 80% of the car manufacturers of America were combust. So we have a lot to do here. Even though we expect there to be very rapid AI adoption from this point onwards. That's not to say that there won't be nontrivial challenges along the way. And of course, I would say this, but we believe it very strongly that, that provides us active investment managers with a particular purpose now, not just to bring you into the trade, hopefully, if you're not there or to remind you of our excitement around the AI investment opportunity, but also to try and avoid the automakers that you've never heard of. And try and find the BC and Ford and General Motors. We also have an awful lot of -- or the industry in the world has a lot to do -- a lot of work to do in terms of making roads safe agreeing the rules of the road where even from a very prosaic basis, we're going to need a lot of petrol stations. We're going to need to train a lot of people to change tires. So there are lots of things that still need to happen, but this vision of the world to play out, but what we're trying to demonstrate is the AI model of today has reached a level of capability where this is possible. And we've shown this chart before, there are very few times in the world in human history where this actually happens. The technology industry is infamous overpromising, underdelivering and then delivering you much more when you stop looking or when you stop caring. And I think, therefore, it's right for investors to default to bubble. -- names like this are very rare. And so I'm going to not necessarily believe it. There's an awful lot of investment narrative around the world that just doesn't believe it. So we're going to take you back to a period where everything changed in a very short space of time and how critical cheap steel proved in the transformation of how humans live and work, formation of cities and then what follows after that. And we also include a slide because it's very conveniently maps to the revenues of Antropic, which is one of the frontier labs really driving the AI story with their Calude models, which I'm sure many of you have used. And I mean, this is highly unusual moment. I mean it was fairly unusual when the company had 20-fold its ARR and then 30-fold. Now we're at 65-fold and likely higher. So I think it is worth just considering this for a moment that this is highly unusual. No company has ever reached this point has never scaled this fast. And again, I'll go back and take -- I suppose, some credit the team where we talked about this. There's a lot of skepticism about how fast this could happen. And we made the point many years ago that this would happen much faster than the Internet because we have billions of smartphones because the Internet delivery mechanism, we had cloud computing, we had storage. We had chip, the GPUs that power the model training have been around for 25 years. So a lot of the building blocks we needed. Anyway, that's all past and what have you, but this is a very unusual moment -- but why has it happened now? Well, Ali said, the models have gotten now human plus. So that opens obviously a door to beginning to substitute human work. But if model attention or if the capability of models only extend to a very short amount of human equivalent work, there's not much money in that. So if you went back to the days of check to put, again, this chart, I'm sorry, we should have made a bit big for you, but you can see it on the Internet, you could give it on the website. And this chart on the left is the idea that over time, the amount of time that a model can -- the amount of human equivalent time that a model could be trusted with that work and get it right, 50% of the time has increased exponentially. And so if you went back to the launch of ChatGPT, the amount of work, the time of human work would be -- would have been about 30 seconds worth of equivalent work before the model started making mistakes and so on and became very unuseful. Today, we're into the 10, 12, 18, no one really knows exactly how many hours of human equivalent work that models can be trusted to do -- and of course, what that's done is open the door on complex work, and that's where the money is. And that's why corporates are adopting this. This is what we believe is equivalent to that Skyscraper at a moment where we sold the industry and cheap steel and elevated sulfur vertical height -- the duration of tasks that AI models can achieve is equivalent to conquering cognitive debt. And with that, again, opens up a completely different conversation. And now we're talking about, again, how far and deep the AI competitor and that's why companies are adopting here. It wasn't very long ago that we were being asked regularly about well, there's an MIT study out there that says only 5% of iPilot work. That's because the models weren't great and because they have to be slotted into a workflow that has existed for a very long time. Those problems are nontrivial. They are going to take time to resolve, but the model itself has made phenomenal progress, and there is a productivity gain to be had today. And that's why Antropics revenues have gone to where they have. We haven't necessarily sold yet for the consumer case, but we're solving for productivity. Consumers don't really care about productivity, so untold, but companies do. And that's again why we're seeing downloads of MTP. And again, Patty's done a great job for the team, but there is some leasing win from model improvement he's filing. That means hopeful I haven't abate. So better model, duration, price adoption and disruption. We talked a lot about this. I still think despite the fact that there have been some high-profile passage to the AI trade already, but the market continues to massively underestimate the scope and magnitude of disruption that's coming from there. And again, I wasn't alive at the time we read in history books, but I don't think they will tell you very much about how relaxed people have been at previous moments like this. where matches a human need to be able to contain a risk, but that we can somehow compartmentalize where it is and where it isn't that will bleed into. But of course, it's very difficult to do that when models are improving at the rate that they're improving. And as they improve and as the cost of token continues to collapse, it becomes incredibly difficult to sort of do a kinome and say that's enough, that's as far as it goes. In terminal values that are at risk. And I will come to that in a minute. But the heart of the disruption, at least for the digital world that we care about most is the code base upon which all digital businesses have been built. It was once scarce and is now abundant. And if you thought about combined harvester and what that meant for the value of land or for the value of incremental farmland and what that might have meant for the labor that work the land, that's the right way to think about this. Something that was scarce and human has become abundant and machine. And it makes sure I'm comfortable reading -- there is a good -- there is lots of good news here, particularly if you're looking to lean into agriculture experience where the cost of weak collapses would have been deeply worrying for people working in the land. I'm sure at that point, humans within 50 years of the mechanization of agricole 20 years longer than they did previously. That's phenomenon. That's what we're excited about. Of course, we're investors, but happy to not be excited about the idea of a technology that could profoundly change the relationship with, I suppose, humans and human and longevity and the quality of life. So going back to the Core Point, it used to be human. It was expensive across $7 to $15 a line, it's essentially for. And so the installed base of code that your business has built on has now become much less valuable in fact and how companies deal with that. We're also going to see experimentation and competition in ways that most incumbent companies haven't had to deal with for a very long period of time. We've just changed one of the critical inputs, and that's why we're going to end up with much more disruption. You see this DIY idea why buy a piece of software, I could just grab an image, shut it into cloud, ask laro basically become a software program, spin up a reader of our image in real time. Again, if you're not doing it already, and I'm not giving you health advice or investment device and what have you. But just my mom is now using the stuff that she gets to a doctor by blood is it might be. And Claude is building basically a history of how medical needs. It's amazing. And so many of the things that would worry here are no longer worrying because she's informed in a way that she couldn't have been the field. It's very much like the democratization process that we went through with the Internet, but just a lot more important. This isn't about democratizing sources of information or silos of information. It's about synthesis. Ultimately, large language models can read and make sense of huge amounts of data in ways that humans cannot -- and again, it's just really we're scratching the surface for dragging an image and reading an MRI scan. That's really just scratching the surface. We're going to deubstruct the digital world. This is a view we've had for a while. It's helped inform our underweight position in software. It's allowed us to exit what bets we have in information services companies in the trust, but also in our AI fund that had a larger exposure to those companies because they have proprietary data sets. But what we've seen is that WattamAI and this abundance of code is enabling is really the unbundling stripping back of these businesses to what their core propositions once were. And in the end, it's because the user of that unit of insight, let's call it that, is no longer a human. And so if I'm a paralegal and I'm reading -- and I'm swamped with information that I need to be alive to aware of then someone telling me that information in a package that allows me with a user interface that I recognize that I find what I need is an absolutely critical component of my job. But if it's a machine that's incremental, doing the incremental analysis on something, that machine does it require the same front end. It doesn't require that all of the countries that are human required, it just gets to the heart of the problem. And we're seeing that in software where there's more competition, where unbundling really. Let's leave it there information services, there's definitely value and proprietary data, but not everyone's data sets or proprietary pieces of that data set are proprietary and lot bundle it together, fabulous for a human, but a machine doesn't need all that. It just goes straight to whatever requires. Digital platforms that bring buyer and seller in a world where there's an unlimited number of hotels to choose from as a customer, Wonderful. Go to a single website, give me 7 choices, and I'll pick one of those hotel rooms. But in this model, the hotel room is booked by a machine, by an AI agent. And that AI agent doesn't need you to reduce the amount of choice in tea. It doesn't have attention capacity. It doesn't value time in the way that humans do and so again, finally, content creation, bringing together buyers and sellers of media. That's great, but we've now found AI models capable of producing human like in time, of course, better than human content. That's why we've opt this pattern that we've been talking about for a while where models get better and the scope of disruption continues to extend. Again, I'm not clever enough to know when this finishes. Almost certainly, the market will cease to care when AI model progress slows, but that's we're seeing AI models increase at an accelerating pace, why those conversations being about potentially slowing down development at the frontier. So all of that comes together is revenues. This is a sector -- this was a story that lacked revenues as a proof point, and now it is not. And now the debate that we have around CapEx and how big could it be and where can we go, are at least grounded by a very, very large revenue base. $220 billion, I think, is the number, but I mean 5 minutes ago, that was $175 billion. And with each model and with each capability that we unlock, again, highly, highly unusual. And both of the labs now are said to be around GBP 70 billion. So we're talk about GBP 140 billion in 2 companies in terms of revenues -- and nothing -- there's no precedent for this. This is a press. How big is the market led here, we try to extensive -- all I will say is that if you think about it in terms of the software market we can get to, obviously, many times larger, but that's really just a visible market opportunity. That's not what we're playing for. That's not why people are putting in trillions of dollars of capacity. It's the other step. If we think about how much money we could size in the market by time saved, we're $1 trillion today, probably more today. If you think about the size of what we spend on tumor wages, $40-odd trillion, we think about the global OpEx at $60 trillion. So again, there is a reason why the very smartest people on there and I'm not suggesting that includes us. But the Amazons, the Googles and everyone that matters is leaning into this trade because the market opportunities are very, very large because we've got a technology that is now able to not just augment, but ultimately substitute for labor. And that's why we're in this race. And again, the rate has been characterized in lots of negative ways, actually, for the last 3.5 years. It's irrational. It doesn't make any sense. It's I don't know, beauty project, whatever it might be. We were alive and alert to the AI storage we launched an AI fund a long time. So we're always excited about it. But when we made the big portfolio here, it was with the very smartest people at our back. They were doing this. We lent in to what Sundar and others were saying at the time. This is a very critical moment. The Frontier labs are winning. There's a very good story out there and don't get me wrong, it's a reasonable risk around commoditized models as an alternative in a race, this was the gold major little distance running for the older people in the room, and there are always other runners. The London Marathon has thousands of people telling us, but only a handful can win -- but you need to have all of the other people there for it to be a race and for that to be a price. And the price be we've discussed it, the rate at this point has lots of runners and rides. But really, all of the revenues are being captured by a handful of companies. And again, it's a conversation that we should probably continue Q&A, but we remain very much aligned to the idea that this isn't going to commoditize anytime soon. In the end, steel became a commodity input. I'd like to say this one, but Mr. Bessemer did Rich, there are going to be wonderful opportunities for the frontier labs and others, and we want to be part of that we're-- we don't want to do on the other side of the trade, which I think most investors are on hoping that what we've seen so far turns into something more democratic, more evenly distributed for not show. Certainly, that hasn't been the case today. And that's -- and then this raised the price, not everybody in the race has all that they need. Some people have the model. Some people have the silicon, some people have the infrastructure, some people have a cat. That's -- but nobody can afford to lose. And in a world like that, you end up with these interdependencies that if you're a negative person, you would argue was circular financing or it was a deck of cards. That may still be the case, like infrastructure cycles don't end well in the end and we know from history. But that doesn't mean that the interdependence is we it may do, but it really, in our view, reflects the fact that nobody can afford to lose. And even the very smartest people in the industry don't know how this is going to play out. Very recently, NVIDIA or hubbing face, which is about a purest play on open-source models that they could have owned. And at the same time, I've talked about making an investment in our topic, which is at the very other end of the spectrum, the ultimately closed model, the frontier close model. If -- I mean, gentleman's an awful lot. But if he doesn't know this is going to play out, that tells you that no 1 does. And that's why we have businesses investing in each other because in the end, nobody can afford to be 0 in the most important rate of their lives. So our CapEx keeps moving higher. It's why we create bottlenecks that will resolve and the new bottlenecks have created. You put a new road in and all of a sudden, that traffic finds the bottleneck somewhere else. And while there's a demand supply imbalance, which is very much where we are today, we expect more investment bottlenecks and therefore, more opportunities for this team that does 1,000 company meetings a year, that's obsessed about this stuff to exploit those bottlenecks as they arise. -- risks, lots of them, of course, they change. Right now, we're talking about certain risks that didn't exist macro risk, politics for sure. For us, the single biggest risk is the model stop progressing our path that creates this unknown destination. Once the destination is known, this will seem to be as interesting story, great opportunities. I think profit pools are being thrown shaken around in ways that are probably very uncomfortable for people. Again, don't have the forum to talk about that, but look for more conversation around that in upcoming reports. But businesses that rely on Anasha businesses that interface with humans. These are businesses at risk. In a while where the incremental customer is a machine and where an incremental customer that's a machine doesn't care about time, doesn't think about inertia in the same way concentrate would be alien to a machine. Quickly on the board base. Normally, what talk about the pool case, all of the narrative on the negative. But it's just worth reminding ourselves, by Ali said, early automations of technologies are always lightweight imperfect versions of themselves. You can't understand the value of the iPhone to Apple if there's no app store. You can't ask where we are with the AI story, if you allow yourself to be basically locked into the current narrative step out, things that the AI is good at, things that AI is bad up, but it becomes a moot when AI is better than humans everything, even if some of the things it's better are less good than others. Absolutely comparative advantage. And again, why we say bullish? Because in the end, it's the invisible market. You can't know in 1860 that steel is going to make Gillette possible, and you certainly couldn't know about the refrigerator or whatever it might be. But all of those invisible markets, we think going to open up an abundance of markets with abundant cognition. The right way to think about AI isn't a substitute for human. That's how it may begin. But really, it's the ability to scale cognition in a way that was impossible in a human world. We want to solve a problem, need to hire some more people. We've applied discrete human cognition to problems, and we look at us as a species achieve phenomenal things. What happens in a world where you can apply cheap abundant cognition to problems that existed, but couldn't be tackled economically or problems you didn't even know you have. If you think about the moon landing, how did that happen? -- thousands, tens of thousands of humans specialized coordinated by NASA. If you look at what we've seen at the frontier with math problems being solved that extend for 90 years. It's a coordination layer from a leading front lab it's able to harness to corral 10,000 agents that solve problems that humans could have maybe solved but haven't got round to or problems that the tumors didn't know they have. And again, how can you not be bullish about a world of abundant cognition and being able to orchestrate it. Does it come with risk? Of course, it does, but let's not lose sight of the bull case. Expect to have to continue this all aware from the very beginning, this team has had to deal with inbound. It seems to be happening on a faster and ever more kind of frightening rate of things that we should be aware of and therefore, not be in the trade. And so far, it's paid to stay in the trade to not be shaken out of this incredible investment story, obviously, do the work, make adjustments in the portfolio where you think that there are adjusted to be made. But so far, we've been climbing all of war -- we are not applying peak valuations to peak earnings. I think many of our stocks have seen their PEs half in the current year because this is a very significant rule of worry out there, but do guard yourself against periods -- we just went through one. It wasn't a great. I think I have the best summer sell, and that's fair enough because new cycles, volatility is endemic to them, and there will be setbacks in this story. But we remain very much AI maxims this period of the 949 backdrop brilinta period for tech results, but lots of 7%, 15% pullback is how you should at least how we think about the likely outcome from here? PAUSE Very quickly, I haven't -- I know I've run out of time. I just want to quickly share the portfolio and then open up to Q&A. So as you would expect, this is a portfolio that is very AI heavy. And that's why we're AI maximists -- but there's the shape of the portfolio. In terms of the market cap and geographic exposure, it's not that dissimilar to how it's been in the past. Like Ali said, we are underweight the MAG 7 names, but our positions around those names can change quite from a -- our turnover is elevated for that and for other reasons, reflecting the opportunity set, but also a much more dangerous backdrop if we fall on to the wrong side of the AI story. And here are some of our biggest overweight on the left, in the middle or underweight the bowls of 0s. And like I said, the underweight mega caps is probably the standout point on the right. And I think with that last point, valuations. So we still remain -- and again, I think this is really important to make. We were once a 6 to 8 core theme investment vehicle to hopefully be a capture for investors for tech. We can do that. We can find other things to do if we're not doing AI. It's just that we think that the AI story is, as you can imagine in your presentation in all the years before, we think this is the most important thing that's happened for a very long time. So the good news is we continue to be careful about how we invest. We still have 95 stocks. We still diversified, like I say, geography and market [ cap ] we also continue to build a portfolio here that has more growth for a modest premium relative to the benchmark, which is what we've been doing for 20 years. It's just that our concentration is around the AI story. And I think Ali sort of alluded to earlier, when we have a good year now, we should have a really good year to hope. And when we have periods of buffeting you should expect it to be a bit. We're traveling pretty fast. And I think this is true, but the faster you travel, the more comfort that you should expect. And so the alternative to that is to us to diversify your portfolio, which we don't want to do. So expect us to overrun at forums like this to overcommunicate, but I think that's the best we can do is to try to take you with us on this journey and to guide you a little more about the volatility that comes with a little bit more concentration, but hopefully, that's offset by are excitement, but also the path that we think AI is likely to play out. And with that, let me take you back to the appendix. And China over to Q&A?

Yes. So if people want to ask questions, please just raise your hand, and 1 of my colleagues will come around with a microphone. And these from your questions to Ben, Ali, the team or any of the Board as well.

2 years ago, more or as the spend is ramping up. I think sitting around that that your type are -- probably no, you are -- that's set and the players would be able to play gold to rely on a par much through -- absolutely no as will happen to close How do you think we are? through Istria Skype in New York in Carape. There's going to be a time where it will add good earning in touch soon. But when it does and earlier reback down from the supply chain expect you got plans conception what might happen

Here the thousands of heating, some mutation music, while I am on the beach. No, I mean -- have we got plans Yes. I mean, we do. I think that we're very -- so the question is a great question, like and again, I'm not -- this seller I'm being rude about a great question, but -- it's the same great question that was true last year, and it will probably be the same great question we hope for next year and beyond. But we've been very honest about the fact that history has been unkind to infrastructure build finish building your canal and then someone comes along with a railroad or something. So we tend to get to a point that greed becomes much more traveler than fear. We extrapolate, we build, we take a peak point and then we build off that and it all ends very horribly. I don't think that this should necessarily be any different in the end. I mean there are a few things along the way that give us comfort that this isn't a dot com repeat performance like we've been very vocal about that for years. First one is that Well, these are some of the very smartest people on there. Now we don't know their motivations. There are businesses that they have to protect, Gogo has a business to protect as well as an opportunity to attack. So there are always -- you have to be slightly alive to the fact that your interests are probably aligned with the people that you're investing with, but you can't assume too much. We can take some comfort from the fact that I mean right now, the paybacks to the investment are almost in. I mean we're talking about companies joking about the fact that what was a debate a year or goes ago was, I think Michael Burrey was saying 3.5 to 5 years, people are filling the accounting treatment of the CapEx -- and now we have people standing in Forum saying, I'm getting such a great return on my CapEx, I could expense it. So again, I'm not -- that's spot pricing, and that's not real, real, but still, we've come -- again, the idea that these useful assets will fall away -- and then there's the financing question of who's paying for it, but I think Ali and the team have been very good on this, which is other industries have gone through periods where you used to buy things for cash. And then you move into Willsher you finance them why why am I paying for carbon I can ease it well, am I paying for a plane when I can lease it. Well, I know there's a residual. So I'm not really answering your question. I think the highest point to all of this is that right now, the very clever people are still investing and looking for that incremental gigawatt of compute which I think is really the most important thing that we can observe. But you do have to believe -- I don't think -- again, I'm not sound like Tadao here, but I do think you do have to believe that there isn't necessarily leafing required that other people have made where we are going to unlock invisible market opportunities. And when we do, it will become very obvious why people had invested in it will be too late. One of the things that worries us about Apple we have Apple, we've owned it for 25 years. Teply understand why it is that they've taken a different path to all of the other MAX7 stocks. But if something does land, like Meta's new model. If we are now -- we have agents that we're going to organize our lives and finally work up through our to-do list and all that good stuff. And how will Apple build capacity overnight in order to compete with that? And we could look back at what market good bag and 20 gigawatts of compute is being absolutely at a phenomenal decision that was. So I think the net of this is, is that you can be alive to the risks and still be excited about the immediate or medium-term backdrop. And if we see returns to AI progress beginning to slip or the narrative changing or AI regulation becoming actually restricted rather than something that we think is probably more just a bit of indemnification, please, then we'll adjust our portfolio. The most important point I should have said about 8 minutes ago, is the book is very liquid. We've put a huge premium on being liquid. And I know that means sometimes there are parts of the world that we can't access in the private world we have high-profile companies and we would love to be holders of some of those assets, certain times in the life cycle, but we also love the liquidity. We also love being able to change our minds. And when we came out of the pandemic, we had a completely different portfolio to the one we have today. So very build up on AI, the whole team, I think is fair to say is built up on AI. But if that changes, we'll make -- we want this to be an evergreen investment for people. And if it isn't AI, it will be something else. That we deserved by the length of the answer, please do ask question. far more distinct members of the team that I want to bring up After all that.

You said that they're buying shares in each other. This leads to into a anomolistic charge mechanism. I have all of those, but its own charging has been taken to color. The problem is, this is the new digital items that real Japan is embedded in so national health be big service companies. And the problem is now running. I mean alternative has been out a tie, some are embedded in posacompanies are here in a broad -- going back, the thing you didn't mention about the people in that 100 was the task back to the broker up these companies because they were using the system as large companies only 1 another. They get together the and fix the prices and the government just angles on the end of the line.

Really good questions. I mean maybe Alan, you want to take some of that and come up. But I'll just say -- maybe you don't I would say that the most important race that's being run right now is between the U.S. and China. And I think it provides us with a lot of cover as investors and have views, we all have views that are beyond the remit of our jobs because we have people as well as investors. But as it relates to investment we would be anxious if this was perceived as the Manhattan project and couldn't be therefore done in private hands. We're more worried about that. Like when you listen to letters being written that are sort of saying, we might need some oversight. And again, there's an awful lot of skeptics that's just nothing more than sort of Finnie regulatory capture in Blaha, which would be probably more akin to that old world of pulling up the ladder on competition. We don't see it that way. That doesn't mean we're right. Our view on that is that models are improving at a rate that there's a risk that their capability runs away from human ability to marshal it. and the rate with China means that we have to keep running it, but I need to be indemnified. I don't want to be -- I don't make responsible for shipping a model that took that back -- so I need to know that someone -- I'm okay, Donald, it. I don't know if it counts as a sort of legal cover to be called a trader if you don't, but it probably would help in a core now that he said anyone that doesn't do this as a trader to America. So I think, look, in the end, the MAX 7 were left alone largely won't here. The regulatory backdrop as -- regulatory risk was 1 of the big pushbacks to MAG and MAG 7 has dominated technology returns for a very long time. And so I'm not saying I'm wrong, but I also think that they can be true -- that can be true, but it can remain subordinate to a bigger story. And as long as the race is between the U.S. and China, I'm relatively sanguine about regulatory oversight, but let's be honest, if the hugging base thing is a precursor of something much greater, the regulatory backdrop could change overnight. We're very much alive to that. But again, that's why they're in the risk parts of our statements. And I still think that there are risks worth taking for the upside opportunity. Sorry, I was slightly less of both, but not good. Question?

Can I see any quantum computing stocks in the portfolio? Can you explain your in on there? It'd be fair to our quantum truck Secondly, is poor allowed to invest in unlisted stock because I know Scottish Mara's got to have a lot of supply and tropic and other things see them.

Scottish, that's a banking company in Yes, we know -- let me ask Nick to come up and ask the content question. But the answer is, I think we can invest in private companies and a very long time ago, we had a couple of tiny holdings in funds that invested in venture capital. We haven't undermined tenure we haven't invested in the week. Like I said before, we go back to the idea of liquidity preference. There's definitely an opportunity set that we're not going after when we don't do private investments, but we also don't give up anything in liquidity. And I think -- sorry to say it, I'm not referencing Scottish Mortgage, obviously because -- we've got some phenomenal wins in their private book. But I think there is tendency isn't there to overindex your wins. No one talks about that. When you bought a few bottles of wine for seller or something, and 1 of them turns out to be absolute humdinger win the ground National in 1986, you tend to still talk about that win, but you don't talk about the other 30 years of rating where you came in last. And I think that there's an awful lot of that private books that people forget conveniently. I think if you're going to do it with an expert do it with someone that does it all the time. It's their bread and butter to do it in BC, do it somewhere else. Don't do it private. And Believe done a really good job of carving out a niche doing that. That's not what we do. And with that, Nick, answer the Quantum question, which really was what you're asking.

So we keep a close eye on Quantum. We have done for a number of years. It remains a very, very interesting way alternative muting it still remains also very mature. There have been huge gains in the quantum space, but we are not yet in a environment where quantum computing is close to solving a problem in a way that tail computing, including accelerated computing and AI computing can't do. And actually, what we're seeing is a coalescence of the industries, not a bifurcation. So a couple of years ago, Quantum was its own stand-alone industry and companies were chasing quantum as its own alternative computing method. Now what we're seeing actually is that the industry is looking to merge with fatal computing. And in all likelihood, as developments continue, it will be just another part of the computing stack that is on offer. So there are 2 ways that content computing is being attacked. The first is through a small number of stand-alone listed companies and a very large number of private companies. These are all largely pre-revenue companies. And that's not the type of investing that we do. There are many, many players that is not clear who is going to emerge as a winner or even who might be a likely winner given how quickly, but also how early-stage quantum computing generally is. At the same time, you have the hyperscalers who are all -- they all have efforts in the space. Some of the artisan partnerships. Many are running their own quantum labs themselves. And they are starting to protect research. They are potentially leading the industry in terms of the pace of development that they are reaching relative to the stand-alone. And for them, this will be just another tool. -- that they can load into their cloud offerings. And you will have -- you will deliver a problem to your cloud computing workload or your stack. That will then be divvied up. A certain part of it will be solved by CPUs. A part of it will be solved by AI workloads and accelerated computing. And then there will be a small sleeve of your problem that might be best served by quantum computing. And we think that actually, that's probably the most likely way the quantum computing will be delivered. So we keep a close eye on the supply chain, the runners and the riders, the stand-alone companies. None of them really are accessible investments for us, given that they are all free revenue and starting to do and of the moves that for us traditionally are red flags, raising debt at unfavorable rates doing M&A. So we keep a very close eye on it, not least because our chose or quantum expert. -- but we're watching from a far for now. And I think the hyperscalers are perhaps moving to the forefront of the industry at this point in time. For online you choose again, yes,

I'll stick to one for this session.

There's tea and cake, so you can ask as many as you like. Ali says he's here all night.