Investor Webinar: Your Questions Answered | March 2026

Well, I think that you have to take a step back and say why have software stocks had a difficult time. And what we've seen in the past three months — so really pretty recently — is a huge step forward in the capabilities of AI systems in writing software. And that is really exciting because software is the underpinning of the modern economy. For me, it has been the biggest constraint for companies, you know, getting access to software developers, to the talent to produce software. Now we're in a situation where AI systems are producing software and these systems are getting better and better.

I think there's going to be an explosion in the amount of software written. And I think that's going to create all sorts of opportunities. But the other thing you have to understand about it is you have to think about the system. If you have a value chain where the constraint has been software developers and that constraint is no longer the key constraint, then value will shift. And what the market has been trying to work out is where are those pinch points going to be in the future. You know, which are the companies that benefited from that being the constraint and where are the new constraints. And I think that's the lens through which you've got to view the portfolio.

I think one thing we're doing is not having a single bet on AI, a single bet on one approach to AI. And so within AI, it's about having exposure across the value chain. It's about at the hardware layer having the exposure to Nvidia, to ASML, to TSMC, where there's an agnosticism that they benefit whatever model works. And then the infrastructure layer with Anthropic, with Amazon Web Services, and then finally the application layer where there are some companies for whom progress in AI is really helpful — Aurora, self-driving trucks — and there are some where it's not a make-or-break for the investment case, but it's just an option that makes it better. And I think it's about having that diversified exposure across the value chain that helps us navigate it. And then I think the other element is that Scottish Mortgage has exposure to a wide range of growth themes: iconic luxury, digitization in emerging markets, the energy transition. So I think it's about trying to construct a portfolio that has a multitude of drivers and it isn't a single approach in AI becomes the real beneficiary. Our expectation is that the build-out of infrastructure in AI will be bumpy, and we construct the portfolio on that basis.

I think it is genuinely plausible that AI is the biggest technological revolution of our lifetimes, and that creates a massive opportunity for value creation. At the same time, an opportunity of that size comes with a multitude of different risks that you need to be aware of. I think in some ways you could bucket it into three risks. The first risk is the risk we've talked about, that there's this huge build-out in infrastructure and the pathway is bumpy and that there are periods of difficulty and you have to navigate that. The second risk is the risk that Tom has been talking about. That's the risk of disruption as the market starts to question what does AI undermine? And the third risk is the exposure that we have to AI. And I think I'd frame it as much as, do we have enough exposure given the opportunity that exists here, and do we have that right type of exposure across the AI stack. So those are the three risks that I always kind of highlight when thinking about AI, but it is in the context of an opportunity that is incredibly large. And I think there is a health warning, and we've seen it with software stocks, of you cannot be invested in AI and still not be hurt by it. There aren't necessarily that many safe harbors where AI won't transform your industry.

Maybe I'll add to that. Think of it like a music festival. You used to go to a music festival, you used to pay for your tickets and you used to pay for your drinks when you were there. Now you go to the music festival, entry is free, drinks are free, but you have to pay to use the toilets. Now, you don't want to have exposure to the companies that sold the tickets. You do want to have exposure to the companies that own the toilets.

Yes. So, at this point we own two developers of large language models. There is Anthropic in the US and we also own MiniMax in China. I think that it comes slightly back to what Lawrence was talking about, that, you know, there's a lot of change going on. We want to invest across the different layers, whether it's infrastructure, whether it's chip developers, whether it's applications, but also at that intelligence layer, because what we've seen in recent months is more differentiation in strategy and capability of features of these systems. Now, we're not constrained to only investing in one of these things. You know, we have the luxury of being able to invest in several.

What we like about Anthropic is that they've had a very cautious approach to releasing these systems. They've thought a lot about safety. They've thought a lot about the constraints that they build around these systems, which I think has been good for their reputation, which has then enabled them to sell into enterprises who are very naturally very cautious about these types of issues. So we think they've had a sensible approach to the model release. We think they've shown pretty differentiated capabilities, particularly in software development. Again, back to this being a really core theme of what's going on. But does investing in Anthropic stop us investing in others? No.

Just to add to that, I think one of the appeals of Anthropic has been its focus on enterprise, whereas OpenAI and ChatGPT have been focused on consumer. And I think there's an attraction of AI in enterprise because as a consumer, I'm using it as an advanced search engine. I'm doing Q&A with it. I'm maybe getting it to spell-check and to bounce some ideas off of it, to generate some images. I'm not using it for the world's hardest problems. But as you shift to enterprises, your ability to monetize is far greater because you actually want to have the very best model, because if you don't, you risk being outcompeted by other companies. And so the willingness for enterprises to pay, I think, is more clear and larger than it is on consumers. And then I think there's an element of stickiness where even in my own use, I switch between ChatGPT and I switch between Gemini quite happily. But I think as you integrate AI as an enterprise with your proprietary data and it learns from it, that could be very, very sticky and have a very high switching cost to leaving. So I think there's an element to which the enterprise approach in AI could be particularly profitable and a particularly good business model.

Well, I think the important framing for us is we're long-only investors. That means we only own companies. We don't try to profit from companies doing badly. And even within looking at only being owners of businesses, we think that there's only a relatively small number of companies that matter. So we aren't really trying to do the job that is the premise of the question, you know, sorting the winners from the losers. We're trying to find a small number of exceptional businesses. Now, with that caveat, I think, you know, what are you looking for in a software company in this environment? And I think it's, you know, what is the data that you have? You know, how important is that to your customers, managing that data? What's the distribution that you have? You know, if the advantage ceases to be simply writing code, you know, then what are the other assets that you bring to that.

So let's take the example of Shopify. That's an operating system effectively for retail businesses, whether they're online or physical retail businesses. Now, they have a huge data set on, you know, what's being sold, who's buying it when, you know, how to operate a retail business, how to advertise, you know, all of these different aspects. And then they don't make money from just selling software to their customers. Actually, they make money by providing payment services. Now, let's imagine a world where AI has, you know, become an important surface for people to buy goods on. They're still going to have to pay for those somehow, and so they're still going to have to use Shopify's checkout. So, you know, it doesn't really matter to Shopify, you know, in this transition, because the mechanism through which they generate value will be untouched. And so I think that's the, you know, that's the model that you're looking for when you're evaluating software stocks.

The other thing to remember in this selloff is what's really happened is the market has decided to attach much greater uncertainty to the future cash flows from these companies. You know, software companies continue to trade very well. It's just that the market's saying we don't know how long into the future that's going to persist. And so, you know, back to this framing, you know, how secure do you think the franchise is? How do you think AI is going to change the model? And that's the way to think about the winners and the losers.

Not specifically about AI, how AI is affecting the investment landscape, but how AI is affecting the world of work. Is it going to take people's jobs? How is it going to affect them personally? So, I would encourage our viewers to go to scottishmortgage.com and read that if this is a topic of interest.

Sure. Well, I think that the scale of the opportunity that SpaceX is addressing continues to broaden dramatically. Yes. And I think people have just got to update their expectations of what this company can be. And the starting point for it all is bringing down the cost of access to space. That is the absolutely fundamental enabler of what comes next. But what comes next, I think, has got dramatically more exciting.

SpaceX now is in this position to potentially be the monopolistic provider of AI to the world. Now, why could that be the case? What's going on here? Well, the biggest constraint to the build-out of AI is power. Our use of AI is expanding dramatically, but the amount of power that we're generating on Earth is not. And so what this combination is about is, you know, if we move AI compute to space, you can be in a situation where you can dramatically grow the amount of power that you're using through using solar panels, through always being, you know, in daylight, through not needing to use batteries to store that energy. And if you look at SpaceX's planned launch pattern, you're going to be able to put huge amounts of power and huge amounts of AI compute into space in the coming years, and nobody else can do that. If the constraint is access and low-cost access, then how do you make money from that position? Well, the bet here is that you do it not by providing launch capabilities, but by providing AI.

Yeah. And so just to give you some sense of the scale of ambition here, what SpaceX are talking about is launching an amount of energy generation equivalent to the entire electricity grid of America every two years into space. And then the amount of AI that you can run from such a system is potentially huge, growing rapidly at a time where everybody else is constrained by, you know, power generation on Earth.

Well, if the question is, is it cheaper to run a data center on Earth or in space today, the answer is on Earth. But you have to look at the trend lines. And if you, you know, look at all of the big investments that have been winners for us over the past 20, 25 years, it's that they benefited from these exponential technology trends. And it was not about saying, well, what's happening today, but it's saying, if these trends continue, what happens in four or five years' time.

And so if you look at running data centers on Earth, you know, you have the cost of chips, and let's just assume for now the cost of chips on Earth and in space are the same. And then you have all the infrastructure that sits around those chips. There's things like electricity, cost of power generation, access to land, the cost of building these data centers, the costs of putting air conditioning and cooling into these data centers. You know, all of these things are on an upward trajectory. These things are getting more expensive. But if you talk about putting a satellite into space with AI compute on it, the cost of doing that is coming down on an exponential trajectory. The biggest component of that is getting the mass into orbit.

And the driver of that is SpaceX launch costs, and we expect those to come down dramatically, you know, 80, 90% over the coming years. But then, you know, if you start building these satellites in a factory, you then benefit from all the efficiencies and the cost declines that you see in that environment. And so you can see a situation where, you know, these lines cross. The cost of a data center on Earth goes up, but the cost of a data center in space comes down. And at the moment that they cross, then that becomes this huge opportunity which only one company can have access to, which is SpaceX.

The first thing I'd say, just to build on Tom's comments about AI and data centers in space, is that I think the beauty of the investment case of SpaceX is that that is one of the long-term opportunities. It's not the only long-term opportunity. You still have global broadband, you have orbital transportation, microgravity manufacturing, you have their position in missile defense ambitions that America has. So it's also the ability to have multiple large opportunities rather than one that really attracts in terms of the holding size. Because of its phenomenal success, it's grown to be a large holding in the fund. That is what we're trying to do, is that small number of outliers. It's grown to be a larger holding than we would usually have in a single stock.

Now, I think that has obviously pushed our private percentage above 30%. It doesn't make us full sellers of assets, but it does restrict for a temporary period our ability to make new investments. I think if you follow what SpaceX are currently saying in the media, they're looking at IPOing in the middle of the year or towards the end of the year, and that would potentially free up that unlisted capacity very quickly if that happened. And so that would change that dynamic very quickly.

I think the first caveat would be that the number of IPOs and the IPO environment depends on capital markets. If capital markets are strong, then yes, I think it's quite plausible. If they're not, then you may see less. I think the context would be that we have a number of companies that make up the bulk of our private company exposure that are cash-generating, large, sustainable businesses that are IPO-ready if the conditions are right and will value those businesses appropriately. SpaceX we've just talked about, but then one of our rather large holdings is ByteDance, the owner of TikTok. They've made significant progress this year in resolving what to do about TikTok US, and that's traditionally been seen as the block to ByteDance as a whole going public. And then as you go through the portfolio, there are a number of companies that I think, if the conditions are right, are in an IPO-ready situation, and that's how that private percentage number can potentially look very different in just a few months from now.

Well, it's an interesting framing. I mean, this is an individual who has created an enormous amount of shareholder value, both generally and specifically for Scottish Mortgage shareholders over the past 10 or 15 years. And so, is he a risk to these companies? I mean, he is a generator of value, and he has his own idiosyncrasies as every great entrepreneur has. But I think, providing the vision, the drive, you know, he has just been such an important figure for Scottish Mortgage shareholders. So, do I spend a lot of time thinking about Musk risk? No.

It's worth adding that over the last 18 months we've made large reductions to Tesla, which benefited enormously from the increase in value post the US elections. And that used to be one of our large holdings, and as Tom said, created huge amounts of shareholder value, but today it's a much, much smaller holding as well. So there's been a shift in where and how that Musk exposure is expressed within the portfolio.

Yeah, so I have sympathy and agreement with both of those points. If you look at what we've done, firstly, in the last few months we've taken two new holdings in China. The first is in a private company called RedNote, which is a digital community that doesn't really have a Western peer. It's a mix of search, Reddit, Instagram, and short-form video all rolled into one. 350 million users, growing very, very quickly, under-monetized, and backed by both Alibaba and Tencent, which is usually a good place to be in China's ecosystem. We've also taken a holding in the company that Tom mentioned earlier, MiniMax. We participated in their IPO. There you have a Chinese AI model company where the attraction is that you're able to offer 90% of the performance of some of the leading US models but at a fraction of the cost, which I think is really appealing for consumer use cases in the global south. And at the start of last year we took a holding in CATL, the leading battery manufacturer.

So we're finding those opportunities. It's more driven by, here are businesses that are trying to do something interesting and different in this part of the world, and at the same time you're able to buy them at relatively low valuations relative to the US. We don't have a target for what we want to get to in China. It's by individual companies where we see attractive risk-reward, and we'll build that up. At the same time, I would always say that we are cognizant that there are some shared risks of Chinese companies on the geopolitical side, on the domestic regulatory side. And so one overlay we have when thinking about China is what is our total allocation, and today it's around 12%. If you went back five years ago, it'd probably have been around 20% plus. We're unlikely to have that level of exposure in the future, but we're still finding really innovative Chinese businesses that are really worth backing and at attractive valuations.

So I would frame that successive American administrations have tried to reduce dependence on Taiwan, but not necessarily on TSMC. So what you've seen is that TSMC produces 90% of the world's most advanced chips. It is very difficult to build up that manufacturing efficiently without their participation. So what the Biden administration has done, and the Trump administration has somewhat continued, has been enticing TSMC to build manufacturing facilities in the US. At the moment, I think their current...

The plan is to build about six manufacturing facilities in Arizona. The US government have been so keen to ensure that they have provided about $6 billion in grants, $5.5 billion in loans. And that's a good thing for us as investors in TSMC because it is increasing the resilience of TSMC and reducing its Taiwan exposure as it develops manufacturing in the US, in Germany and Japan, and it's being helped in doing so. And the final point around that is that when you talk to the customers of TSMC, what they say is, well, today we also second source from some other foundries, but you know what? If we could get chips made in the US, we would pay a premium for that and we would do more of our business within TSMC.

Well, I think that we've seen a real push to increase defense spending across NATO. I think there the world does seem to have got a more uncertain place as we see great power competition between the US and China. But I'd be quite cautious about responding to the outbreak of war in Iran as pushing us into stocks. I think what you've seen is markets have reacted in a reasonably predictable way with the start of the hot war with Iran, which is that the dollar has gone up, the oil price has gone up and defense stocks have gone up. I don't think we should chase that. I think the defense angle that is more interesting to me is that you're seeing a real change in the way that wars are fought.

You've seen it in Ukraine. You're seeing it play out in the Middle East now, where you're seeing cheap disposable hardware largely in the form of drones, where the edge comes in software and you're actually using this very cheap hardware either to force an enemy to spend much more money on very expensive systems to shoot these drones down. And effectively what you're seeing is a shift you've seen elsewhere in the economy, right? The value moves to the software layer or away from the hardware layer. And so I think that's interesting. I think it presents new opportunities, something where we will keep an eye on. We haven't brought any direct exposure to defense, but if you want to link it to the previous conversation, we've also all seen the importance of space-based communication in some of these wars. And Star Shield specifically within SpaceX is obviously a really important technology in the current paradigm.