AI and the Material World

[Music] Welcome to another episode of the Capital Cycle podcast. This is Edward Chancellor and I have with me Alex Duffy, Emerging Market's portfolio manager at Marathon. Hi Edward thanks very much for having me on. Alex we're living through what the Jeffrey strategist Chris Wood expert on bubbles calls the mother of all capex cycles and judging from your latest contribution to the global investment review it's been giving you nightmares. Yeah I think we're in that period of the cycle which you know Chris Wood I think is referred to as the melt-up phase of the of the AI cycle and the thing that really strikes one is how pervasive that's become so this is feels like one trade with a market that's fixated on one singular thematic and it's all in everywhere and so as a somewhat contrarian capital cyclist this is quite a painful experience to live through and know less fascinating as a consequence of that. And you actually cite a figure saying that US economic growth is 40% of the increase in the economic growth coming from AI capex I saw a piece put out by panmuolibrom saying that actually 100% of US growth is coming from AI capex you can take whichever number you want but it's pretty large. Yeah so I think this is actually the point about AI becoming an all and everything thematic it's driving so much of economic activity and has found its way into every area of a large economy such as the US so if you looked at capex outside of AI related sectors it's actually now negative year on year inside the US which tells you that there's very little going on in terms of a real world job creation environment and so this has become a very very pervasive thematic that we need to be aware of and I think I was talking to your colleague Charles Carter about how AI theme crowds out other sectors and stock market but it's also actually crowding out in the real economy too at least in the United States. Yeah and I think you see that if you just look at construction costs as they become more elevated if you look at the impact that this starts to have on commodity prices it's really an environment which has been driven by very large spending by price agnostic in this case the hyperscalers in the US as opposed to businesses and sectors and indeed individuals who are investing on a near-return returns based formula and so it's a very different driver of the rationale to invest which has implications far beyond the technology sector which is typically how it gets viewed and you think that people should be thinking more about the risks of an AI market dand turn rather than the opportunities for continued outperformance from anything related to AI. I think there's a two-pronged response to that that the first is this point around AI is driving so much of incremental economic activity that if there's any question in of the rationale for that capital investment or any air pocket in the time frame of which that capital is being spent then that could have quite material implications for the real economy that goes far beyond technology itself but it's also more around the impact that it's having on global markets equity markets as well as private equity markets where exposure to the thematic more broadly continues to ratchet up and that to my mind poses problems for asset allocators. So let's talk about that. I'm for instance Torsten Slock the chief economist of Apollo had a note out we're one of his daily charts out last week talking about exactly this the AI share of venture capital is needless to say 90% now dominated by the incredibly hungry likes of open AI and anthropic or lost making businesses we're concerned with the public markets Goldman's had an estimate out at the end of April that the US stop market the S&P 500 let's say is around 45% exposed to AI themes and as we all know the US market is more concentrated in a handful of names than it ever been in history and the US market is very large not quite at its peak but it's over 60% of the global market and you have added something which I wasn't aware of was the extent to which AI is now moving your own domain namely emerging markets. Yeah so the focus of the AI thematic is really in the last 12 months moved away from the US hyperscalers towards the picks and shovels so the tech supply chain predominantly in career in Taiwan Samsung electronics TSMC with the bellwethers initially but it's gone far deeper than that into some more esoteric companies that comprise those indexes so they're now stand at 45% or so of MSCI emerging markets to be clear this is the waiting of career in Taiwan correct in the MSCI emerging markets index and that is up from 25% and in other words a 20 point move in the course of a little more than a year which is a remarkable move you include a chart in your piece which shows that the range of career in Taiwan in the MSCI emerging index has really been around 25% 25 to 30% over the previous 15 years now you also have a chart showing the waiting of the semiconductors tech hardware and equipment as a share of emerging markets which has also increased substantially from around 18 to 20% where it was 18 months or so ago to around 38% today and that rate of increase has just accelerated continues to accelerate. So back to the earlier point this is one of the challenges for asset allocators from a diversification perspective if you now look at global portfolios global indices what you find is that there's increase in correlation to this singular trade the AI trade let's call it that not only represents the weightings of those indexes but also because of the rate of a cent of that part of the market is driving returns at a global level and so this is something that which we feel and we you know we believe our clients need to be quite mindful of is they're thinking about constructing a globally diversified equity portfolios and in the case of the emerging tech semiconductors businesses and memory makers and hardware and equipment makers you don't actually have an argument with the companies or their management no so we've had significant exposure to some of these businesses Taiwan semi delta electronics media tech they've comprised the portfolio over the last five to seven years and beyond in the case of TSMC management teams have been good stewards of capital that said we are at the stage of the cycle where the business risk which is predominantly what the management teams would be able to control and to focus on is being overrored by the valuation risk and the crowd like behavior of markets and so that I think is where our research is being more focused on so on the valuation risk the semiconductor businesses are highly sicken cool I don't know about your emerging names but I see the micron technology is trading on nine times earnings now those earnings are up you know seven or eightfolds in the last year but tell me what you see as the valuation risk for these businesses I think first and foremost stepping back we have seen some of these industry structures consolidate so we were quite positively disposed to them earlier on in the upcycle of technology spending but as we look at valuations today we feel that that is really starting to extrapolate current levels of profitability into the future over a time frame which starts to look very very optimistic about the duration of profits and ultimately the technology supply chain yes there are certain companies of which TSMC we would argue is one which dominates is end profit pool through a technological mode but there are others which are much more commoditized and one of the things that's really stood out to us is that it's no longer the performance of the leading bell weathers it's actually the deep commodity suppliers the components suppliers in the supply chain which are seen the most hyperbolic movements and share prices and to us these are time and capital businesses and at this juncture which I think we might go on to with valuations where they are capital is abundant and it's just a question of time for when that capital cycle will turn yeah and say you don't believe these bottlenecks will continue I read earlier today that in the case of memory makers that there is new supply coming on in late 27 early 28 yeah so I think that right now Now, there is clearly a very tight supply demand picture, and so that's an interesting point in the capital cycle, but that's priced by the fact that operating margins, in the case of Samsung Electronics, have gone from 15% to 90% gross margins in some instances, we wouldn't say that you can extrapolate that for any extent of time. And the point when I think about it is that the hyperscalers are in effect paying for these semiconductors and memory chips and so on. And these are the big US hyperscalers, and their returns on incremental capital are collapsing, their cash flow is collapsing, and their returns on incremental capital invested are also falling. And a panimo report I mentioned to you actually claims that based on analyst expectations, they're actually earning negative returns on invested capital. Now, my view is that even if you have a relatively tight capital cycle, which the semiconductors and memory stocks appear to be in, if the underlying business is not cash-generative or sound, then that's risk. And I was thinking, this goes back to something you'll know back source in your domain, which is if you remember the mining super cycle that went on from, I don't know, 2003 to roughly 2013. And again, you had fairly consolidated industries. You had massive cap ex, but the underlying economics of that mining of the commodity extraction was going to China where it was earning either low returns, say, I'm all going into steel-making businesses, or potentially negative returns by going into real estate. And I think my hunch is we may be seeing something similar today. I would completely concur with that. And I think that the mistake that markets make is to say this is not bubble territory because valuations are extrapolated as they were in the late 90s. And then the miners, if you remember, back in 2012, they were trading around bulk, or they trading at 10 times earnings, or they're about. They traded at seven to eight times earnings, but what actually happened is the EV to invested capital and price to book multiples inflated because the margins were so high. Trillions of dollars of investment. So absolutely, that's the parallel. Today, the tech supply chain, you'll hear quotes such as that it's attractively valued on 20 times earnings, but that ignores the fact that the price to book multiples have gone up by four, five, six folds. And so it's an earnings bubble. It's not a valuation bubble. Because that's what I call a fundamental bubble. When the bubble is in the balance sheet or in the earnings, I call that a fundamental bubble. And most investors, or people who think about markets, always look at just simple valuation metrics on the PE basis and say, look, my chrones on nine times earnings is cheap. Which misses the cyclicality of the business. And what happened in the mining cycle is not that the company's made terrible decisions at that point in time, but they were responding to a false price signal that was coming from over investment in China. And the analogy today is that the tech supply chain and the re-rating of the technology, the tech supply chain equity multiples is a response to a price signal that is coming from a degree of over investment on un-economic projects inside the US. And once that gets questioned, there will be a flow through the valuation multiples in Asia. And that I think is something that doesn't get discussed enough by broader market commentators. Yeah, I think so. And if you remember, one of the angstumps of the capital cycle investment approach is to focus on supply and not demand. Now, the trouble is if you focus on demand in the AI space. And then this is another point made by the Panium PANMEA report. The hyperscators are going to have to get up to five trillion dollars of revenue within the next two years to rationalize their cap expanding at the payment. Now, of course, if you're a bull and now it's reading the code to the Philippe Lefort tech outfit, they think the total addressable market for AI is a hundred trillion dollars. Well, good luck to them. If you think actually five trillion dollars within four years is going to be a hard call, then I think that there is a vulnerability to these emerging market tech plays. And I would agree with that. I mean, to come back to the point around the risk of a market drawdown, you can continue to allocate capital into these areas of the market. But one must understand that there's huge correlation in doing so. And yes, there might be further to run and some additional upside if you're right on what could be a low probability outcome. However, at these levels of valuations and this degree of market euphoria, we are at the point of the big IPOs that typically ring the bell on the top of a market cycle, such of that. If you're wrong on the positive outcome, well, the counterfactual is very, very challenging for equity investors because the drawdown will not be 10 to 15%, it'll be 50 to 60. And I think that upside given downside skew has probably moved against incremental capital coming into this sector at that point in time. And so we're kind of at peak tide that's about to start ebbin. And actually, it's the melt-up phase of a bubble is not just the IPOs, it's actually picks and shovels going through the roof. And what's interesting about what's going on now, at least in the States, is that the same names are involved today as the TMT bubble in 1999, 2000, namely, you know, Corning, Intel, Cisco, now the business of JDS, Unifase, which I think is now called Viva or something like that, they've all been going through the roof. So it's a similar picks and shovels trade as we saw back then. So tell me if you're trying to control your exposure to the Asia technology, what are you doing instead? I think this goes to the point that we touched on briefly earlier, which is that capital is rushing into this part of the global equity market. And there are other areas which are being neglected as a consequence of that. And so as we discussed last time around, we come back very much to first principles, which is what are the range of outcomes for a business? How do you think about accessing those return profiles at relatively attractive valuations, which provide protection if you're wrong? And so, you know, we're increasingly finding as a consequence of the crowding out of capital that there are lower half in our pollence, lower half balance sheet restructurins across global emerging markets, which are particularly interested. That is across a range of different sectors and geographies where our return thresholds can be met without making those bullish assumptions about future profitability and growth. And some of the areas where you're putting your money is actually in the bubble areas of yes, the year, namely Chinese real estate. I saw the other day chart showing that Chinese real estate prices are back at their 2010 level, which is just about when I started getting bearish on the sector and in basic materials that we're just talking about. But you've got other sectors as well. Yeah, I think the Chinese real estate is an area where we've seen a huge contraction, activity, a huge contraction of supply, rationalization of companies, and cleaning up of inventory that's been written off. And so stocks now trade material discounts to book value. And yet the book values might actually be worth, if not 100 cents in the dollar, at least something above 75. And so that to us is quite a compelling setup for incremental investing. There are other markets that have been totally neglected. The Philippines, which a decade ago was post a child for emerging market investing, rising consumerism, lower penetration and so forth, now trades back at 20 year lows on price to book and P multiples. And it's not just the fact that these markets have underperformed. It's what happens at the company level when you get that degree of multiple derating, is management teams change their approach to capital allocation. And so what we're witnessing in some of these other more neglected parts of global emerging markets is positive capital allocation decisions as a consequence of the derating, which results in a rationalization of supply, turning of the capital cycle. And we want to be early in those situations. They can always take longer to play out than one would hope. But actually the upside given downside because of your entry level and the actions that management teams are taking can be quite favorable for long term investors. That's plenty to think about. And Alex, thank you very much for another interesting discussion. Thank you. Thank you for your time today. I hope you will listen to the next edition of the capital cycle. This communication is provided for information purposes only. Please refer to Marathon's website and the Global Investment Reviews for further information, including important disclosures.