The AI arc: Sizing up the boom against history

Today I have the pleasure of speaking to our Head of Investment Strategy, John Normand, about all things AI and how we're thinking about it from an investment perspective.

Hi John, lovely to speak to you today on all things AI. It now represents a significant share of global listed equity markets. You’ve been doing a lot of research on the topic. Can you talk a little bit about your investment framework and how you approached the research process?

I set out a framework which I guess you could summarise as cycle analysis, recognising that AI is the latest innovation wave. Every innovation cycle has its life span to it, and if we can track these various indicators across the economy, across the equity sectors, across valuations, across leverage, we can come to some judgment as to whether or not the balance of these suggest that this cycle has many years to go in it, or whether it's much more mature and might end sooner.

Even though every innovation is different from the one before, from the previous decade, there are these common features across these different categories of indicators, and we can score these and we can come up with a balance of risk around them, and we can align the portfolio with where we think the balance of risk is and where we think it's going to head over the next year or two.

The question on everyone’s minds is where we are in the AI cycle, and I think your work touches to the AI arc?

So whether we're early, mid or late depends on the specific indicator. If I looked across a whole suite of them, maybe 10 or 12, I'd say we’re mid. So the ones that I think are quite encouraging because they look a little more early stage is adoption, and there are different surveys that measure this, but probably one of the best ones is produced by the US Census Bureau, and the message from that survey is, yes, it's been heavily adopted by corporates in the technology space. It hasn't been heavily adopted by corporates broadly in the US. That gives us some expectation that the diffusion process still has quite a bit to go. That's really encouraging.

Another indicator, which is still a bit early stage is the productivity improvement. Yes, there's measurable improvement in productivity amongst technology companies, but not so much in the economy broadly. That's also hopeful for thinking that the big economic gains are still to come.

Some of these indicators which are looking a little more late-cycle are around capex. Companies are spending extraordinary amounts of capex relative to their sales, and that dynamic, which we call capex intensity, is at levels that we saw in the TMT bubble in the late 1990s. So that's a little bit worrisome, but also, importantly, there's no indication that this AI boom is leading to overvaluation in the markets as a whole.

I think if you looked across the sweep of things, you would say this is probably a middle-aged investment dynamic, and therefore it's still an important anchor for how we position.

Thinking about some of the metrics you do track, are there any that have significantly changed over the last twelve months?

Earnings have been fairly consistently strong, and that's important. So despite all of the shocks that we've seen in the broader economy, some of which are geopolitical, it's pretty obvious from the earnings delivery that this theme maintains a lot of robustness and it's not so vulnerable, I guess, to some of these macroeconomic shocks.

The one that has been a little more worrisome is on the leverage side, and the degree of leverage that's being used varies quite considerably depending on whether we're talking about large, very cash-rich companies like hyperscalers, or smaller kind of second tier, newer entrants into the economy. That's an important distinction, because if the vulnerable companies tend to be on the mid-to-smaller size, it makes us slightly more comfortable with the dynamic in credit markets. But when I look at the sweep of indicators, when I look at the breadth of them across the whole economy, I'm still comfortable with the idea that this is a sustainable investment dynamic.

Are there any other particular risks that you’d be concerned about or any that you’re tracking as a part of the process?

More recently, the concern has shifted towards scarcity on the semiconductor side, and that scarcity is leading to moves in share prices for some of these producers which is pretty extraordinary. And this is having big impacts on an industry level. It's having a big impact in smaller equity markets like some in Asia.

That's just a reminder that there are bottlenecks in different areas and the bottleneck that many people have been focused on a year ago, which is in electricity, is not the bottleneck now. It's in semis. And so if you ask what my biggest concern is, it would be that these bottlenecks become an impediment to the profitability of AI. And when you think about how central banks respond to bottlenecks that create price pressures, they respond by tightening policy. So there are ways in which bottlenecks create price pressures, price pressures create a central bank response, the central bank response creates an interest rate dynamic which is unhelpful to the AI thematic. There's no innovation cycle which exists in a macro vacuum.

And so based on your framework today and weighing up all of the drivers and all of the risks, do you think it’s a bubble?

No, I don't think it's a bubble by any standard metric applied at the broad asset-class level. We're not so concerned about individual share prices, we're concerned about problems at the asset class level, and we don't see those yet. And I think it would take a reasonable passage of time for those sorts of problems to materialise on the leverage front or potentially on the valuation front.

It doesn't mean we haven't adjusted the portfolio in important ways. You can reach a judgment that AI is not a bubble, but still decide to risk manage the portfolio in a more conservative fashion because of the way some other indicators are moving. We wanted to diversify a bit more into non-US markets and non-tech sectors. So you don't need to be a buyer of the bubble view to be a good risk manager. I think there are ways that you can just look at some of the underlying indicators and adjust the portfolio in a way which is mindful of that without having to take a very extreme view that the market is in a bubble.

Thank you so much, John, for your time today. I’m sure members will have found that very valuable. I certainly did. Thank you.