Charlotte de Kerpoisson:

Hello and welcome to this podcast by BNP Paribas Wealth Management. My name is Charlotte de Kerpoisson. Since the start of 2023, following the release of ChatGPT, global stocks have been increasingly dominated by one overarching growth megatrend: artificial intelligence. Since then, the Nasdaq 100 tech-heavy index has appreciated by 173% and an average return of 33% per year, with megacap tech companies leading the way higher. This year, the semiconductor sector has been the star AI-related performer, up 63%, and led by tripling prices of memory makers in particular. Are we in the midst of a technology investment bubble that will end a market sell-off as seen in the year 2000, or is this a multi-year megatrend that is still early in its development, with the AI revolution to deliver much more growth to come? 

To discuss this topic, I'm joined today by Edmund Shing, Global Chief Investment Officer. 

Hello, Edmund.

Edmund Shing:

Hello, Charlotte. 

Charlotte de Kerpoisson:

Edmund, do you believe that AI will fundamentally change the way we work and live, and if so, shouldn't we be invested in this technology revolution? 

Edmund Shing: 

Well, I think we need to separate these two things, Charlotte. First of all, yes, clearly AI is a technology revolution, and it is going to change the way we work and live in many different ways. Many ways in which we probably aren’t aware of yet, because the technology is developing so quickly that it's very difficult to predict where we're going to end up in the future. 

That being said, you have to differentiate between a disruptive technology, which can have very far-reaching consequences, like artificial intelligence, on the one hand, and the investment implications on the other. Precisely because we don't really know where the technology is going, it makes it very difficult from an investment point of view to know where to put the investments, where the profits and the sales growth is going to be made over the medium to longer term. And also we don't know to what extent the current generation of AI could be itself disrupted by an even better, newer form of technology. And that's all very important when it comes to thinking about the investment implications. For instance, we already see in the stock market today the underperformance since September of the so-called hyperscalers. These are the technology megacap companies that are investing very heavily in AI today, such as Alphabet, the parent company of Google, Meta, Amazon, Microsoft, and Oracle. 

These five companies are investing massive amounts of money in AI and data centers and compute. But we don't know how profitable that investment will be, and indeed the market has become more skeptical over the eventual return on investment of this massive capital expenditure, which is happening not only this year but projected to be even greater next year. And I think that Charlotte is the quandary that we're in. Even if we accept this is a revolutionary technology, how best an investor can profit from it without taking too much risk? That is the challenge. 

Charlotte de Kerpoisson:

So Edmund, you mentioned the hyperscalers. So, how concerned are you about the change in business model of these so-called hyperscalers—Google, Meta, Amazon, Microsoft, Oracle—away from high-margin, asset-light to uncertain profitability investment-heavy business models? 

Edmund Shing:

Yes, I am concerned because, as I said, Charlotte, let's take the example of Google. Google has made most of its money in an almost monopolistic fashion by dominating internet search, right? I don't think anyone uses Bing, which is Microsoft's internet search engine. No one remembers Alta Vista or many of the other internet search engines that existed in the past. Pretty much everyone, you know. Google has something like a 90% to 95% share of internet search today. So it's a quasi-monopoly, and this has allowed them to be one of the companies along with Meta that has dominated internet advertising, and this has been one of the key sources of Google's fantastic profitability. But as you said, Google today is investing in Gemini, an AI model, investing very heavily in AI, spending hundreds of billions of dollars this year and will do the same again next year. 

And I am not sure whether this huge investment will really pay off. In a sense, Alphabet or Google need to make this investment because, in a sense, they can see the risk that AI will eat into their monopolistic position in search today, and so they're obliged to invest, simply to try to protect their own market share or position. But will that succeed in a much more competitive marketplace where they no longer have a monopoly? I'm not so sure, and I think that is the question that many investors are now asking themselves. Even if AI is a revolutionary technology, even if the total addressable market is in principle huge, are the hyperscalers the right vehicle to profit from this? And if I think back to the technology bubble of the 2000s, I think that this was very much the dot.com bubble, all about the growth of the internet. Now, the telecom operators of the time, and indeed today, are the providers of this essential service. They provide broadband internet service to everyone, and everyone pays for it. But do they make the money? Are they the ones to make fantastic profits out of the internet? The answer is no. They effectively get a utility-like return. It's a reasonable return, but it's not that great because what they provide is essentially a commodity. If you don't go with one internet provider, you can go with another one who provides essentially the same service, and there's very little to differentiate between them. 

Where the value in the internet has been captured has been, of course, by companies like Google, Meta, Amazon, Microsoft, companies like Zoom with video conferencing. Those are the business models that profit from the internet. You might also argue streaming services like Netflix and Spotify profit from the internet, and these companies ultimately capture much more of the profit potential, generated by the internet, rather than the telecom operators. And you might argue that this could be the same today. The hyperscalers might get a utility-style return from offering large language models. But maybe no more than that. And maybe it is the companies that build software and applications on top of these large language models that could generate the excess profitability and growth rather than the LLM makers. And that's why, again, I'm a little bit skeptical of the hyperscalers today, and certainly I would be looking elsewhere for the bulk of my investment into the AI thematic today. 

Charlotte de Kerpoisson:

This year, the focus in AI has been on the picks and shovels hardware providers for AI, notably chip makers. But are we seeing an earnings bubble emerge in these companies? Yes, they are delivering very strong earnings growth today. But are these growth rates sustainable in the medium term, Edmund? 

Edmund Shing:

That's a very good question, and I'm not so sure that it is. I suspect that we're seeing somewhat of a not a valuation bubble so much, although there are hints that with certain companies they do look overvalued, but more of an earnings bubble when we come to the current mania around AI. Again we've got this massive investment boom. We have all these companies such as the memory chip makers, you rightly noted, like Micron, Sandisk, Samsung, SK Hynix, all generating fantastical earnings growth and making very high profit margins at the moment. But we also know that in the past, memory chipmaking has been a very cyclical industry with very big highs and lows, and profitability has also swung around from being very profitable to suddenly not being profitable. And so the suspicion is that maybe we're at a high point of the cycle. How long? We could maybe stay there for a while so these companies could enjoy excessively high profit, profitability and growth for a while. But like anything else, eventually you would expect competition to come in. It'll take time because, of course, you don't build memory chip fabrication plants overnight. It takes a lot of money. It takes a lot of skill, technology, but it can happen eventually. So there is a cycle, and so I do worry that maybe people are getting overoptimistic about the so-called picks and shovels, the semiconductor makers. Even though at the end of last year when we recommended and talked about the AI theme and how to play it, we certainly recommended investing in the picks and shovels. That’s certainly paid off, but I am not so sure I would be committing a lot more money to these semiconductor makers today at this point. 

Charlotte de Kerpoisson:

If you are an AI optimist, where would you invest today to gain exposure to this technology megatrend? 

Edmund Shing:

Well, I think there are two areas I would focus on. I think the first one would be where I see the bottlenecks around AI and the development, the growth in AI. And one of the bottlenecks is clearly energy or electricity supply. Clearly, you are seeing lots and lots of data centers being built in the US and elsewhere, but really the bottleneck here, particularly for these AI centric data centers, which require huge amounts of power, is exactly that supply of electricity. Electricity grids in the Western world today are not set up to provide the extra power that we're going to need to power these data centers on top of all of our other industrial and household needs for electricity. 

Remember, electricity demand tends to grow anyway, and now data center demand is turbocharging overall electricity demand, but we have not invested sufficiently in generation and transmission equipment to supply that extra demand. So I think that is the first area: the supply of electricity generation and transmission equipment, for instance, things like transformers, gas turbines. There, there is a key bottleneck. This has performed well, the segment, but I still think it's going to continue to perform very well going forwards. 

The second area I like is what we call physical AI, and this is the application of AI, for instance, to robotics, humanoid robots, for instance. So we get these multi-purpose robots that can accomplish many more tasks than perhaps in the past. Now, clearly, robotics has been around for a long time. You can't really see any car factory without seeing industrial robots. But remember, industrial robots are specifically designed to perform a specific task. Now, what we are talking about with the application of AI into robotics is taking the AI brain and putting it in the mechanical body to achieve a more general purpose robot that can accomplish a number of different tasks, perhaps not just in factories, but also in service industries. And I think that has huge potential. We've hardly scratched the surface of this potential for robotics at the moment. So absolutely, I would say, looking at robotics funds and ETFs is the second way I would invest in potentially the next wave of AI today. 

Charlotte de Kerpoisson:

Edmund, you mentioned recently that investors should not be all-in on AI in their stock portfolios and should look to diversify portfolios into non-tech sectors. So, which sectors and themes do you recommend in order to diversify away from AI and away from tech? 

Edmund Shing:

Well, the sort of sectors I'm thinking about away from tech are more value-oriented regions and sectors. So, when I think about that, let's take the example of the US first. So, in the US, as well as having technology exposure, I would recommend diversification away from just tech and looking to industrial companies, smaller-cap companies for different exposure, less tech-heavy, and in particular, buy into the thematic of the industrial renaissance or reindustrialization and reshoring of manufacturing in the US. And we have funds and ETFs that certainly do that. 

In Europe, again, we have a lot less tech, but we have a lot of other value areas that are performing well that we like, that we would put into a so-called barbell portfolio where at one end you have your AI and technology exposure, and at the other end more value exposure. And into that value segment, we would add things like European banks, which we still like, which are performing well, which but still offer good dividend yields, relatively low valuation, rising profitability amidst a very positive economic backdrop for banks. I think as well, if you think about healthcare, it is actually a bit of a value area these days. It has been neglected by investors in their headlong rush to buy up tech and AI exposure. But remember, there are long-term positive demographic factors in favour, of course, of healthcare and pharmaceutical companies in terms of long-term growth and demand. And we also think that AI is going to turbocharge both the diagnostics area and also the finding and treating of new drug candidates for new potential blockbuster drugs. So, I think AI is going to accelerate the research and development part here, make research and developments in healthcare a lot more productive and effective. So, I think that will be the second sector I would focus on today for the barbell approach. 

Charlotte de Kerpoisson:

Thank you, Edmund, and thank you to our audience for following this podcast. Please like, share and subscribe to our weekly podcast. Do not miss our mid-year investment themes that we recently published. For more information about these and for all our investment strategy research, please visit our website. Goodbye.