Could today's AI investment boom be following a familiar pattern?

Watch the video

Listen to the podcast

In this episode of Connected Investor, Brunner Investment Trust Portfolio Managers Julian Bishop and James Ashworth explore what past capital cycles can teach investors about today’s AI infrastructure boom.

Looking at historical examples such as the iron ore supercycle and US shale revolution, they discuss how surging demand can drive investment and new capacity, potentially affecting future returns. They also consider whether rapid growth in AI and data-centre infrastructure could lead to oversupply.

This is a marketing communication. Please refer to the key information document or KID before making any final investment decisions. Investing involves risk. The value of an investment and the income from it may fall as well as rise and investors might not get back the full amount invested. Past performance does not predict future returns. The mention of any particular security or strategy should not be considered as a recommendation. For further information on the Brunner Trust please go to www.brunner.co.uk.

Julian Bishop: Everyone's focusing on demand. Everyone's focusing on demand growth, and they're ignoring supply. As we know from these previous cycles, it's supply that's equally important as demand.

James Ashworth: But this new technology really changed everything. It changed the game and led to a really big increase in capacity.

 

JB: Hello and welcome to Connected Investor from the Brunner Investment Trust. I'm Julian Bishop.

JA: And I'm James Ashworth.

JB: Now we often hear that history doesn't repeat but that it rhymes. So, in this episode we wanted to look back at some past investment cycles and see if there are any general lessons that could be applied to today's AI investment boom.

JA: We've picked a few fairly recent examples where a hot industry attracted lots of investment, which ultimately brought the party to an end.

JB: So that's a good place to start. Is there a common pattern to the examples we'll discuss today?

JA: Yes, there is. The examples we'll talk about today are all ones with commodity products, where those products, by definition, are undifferentiated and where it's possible for new entrants to enter the market, or where it's possible for the existing players in the market to increase supply.

JA: So, if you think back to your GCSE or A-level economics, these are pure commodity markets. What happens in those markets is that the amount of supply in the market responds to the level of prices and the level of profitability, and we see what is called a capital cycle. So, if there is a disturbance, let's say an exogenous shock that leads to an increase in price levels, that increase in prices tends to lead to an increase in profitability for the people who are present in that industry. That level of elevated profitability draws in new firms or increases capacity from the existing firms in the market. That increase in supply, in turn, leads to a decline in prices and a decline in profitability. Ultimately, some capacity will look to exit the market. So, we end up with a cycle: high prices drawing capacity, which results in low prices, which leads to capacity exiting in a sort of perfect circle, if you like.

JB: I think you mentioned GCSE economics.

JB: I did A-level economics. And this is really a fundamental foundation of capitalism, isn't it? Just supply and demand. High prices attract competition and attract new supply. If you look back at economic history, there are numerous examples of cycles that follow this common pattern. I think one good example to start off with is iron ore. So, let's talk about the mining supercycle that occurred at the beginning of the last decade.

JA: That's a really interesting one. So, iron ore, obviously a commodity product, a key input into manufacturing steel. For decades, this had been a very sleepy commodity. I guess the price of iron ore before about 2005 had always been about $25 a tonne, give or take. But something really changed at that point. Maybe a little bit of a history lesson. What happened to the iron ore price, and what was the driver of that change?

JB: So really one word China.

JB: The industrialisation of China consumed a vast amount of steel. This really took off again and went up another leg following the financial crisis. If you think about the financial crisis of 2008-09, China's main export markets were in recession. China, in response, stimulated its economy with this huge programme of property building, infrastructure build-out, airports, bridges, railways and so on. Of course, all these consumed vast amounts of steel, which is ultimately made from iron ore.

JA: So, this huge demand, huge increase in demand for iron ore leads to an enormous increase in price. There's a huge increase in demand that drives prices up. And we see the iron ore price increase from what had always been roughly $25 a tonne to a peak of around $200 a tonne. presumably that's wonderful for the iron ore, companies, the iron ore miners. maybe talk us through how they respond to this, this environment, this huge increase in prices and presumably profitability.

JB: You're right. They're making huge amounts of money at these prices.

JB: These supernormal profits attract new supply. Now, most of the world's best iron ore resources are in Western Australia. If you think of north-west Australia, there's a region called the Pilbara where the ground is effectively made from iron ore. It's very, very high grade and requires minimal processing. At that point, because of these very high profits, pretty much every iron ore producer in the world looked to increase capacity. If you look at one of the biggest, Rio Tinto, the London-listed company expanded its operations in north-west Australia, in this area called the Pilbara, from about 220 million tonnes a year to about 290 million tonnes per year within just a few years. Their capital expenditure, the amount they have to spend on mines and railways to get it to port, went from about $5 billion in 2009 to about $17 billion in 2012. They even built their own port. Capital expenditure tripled in just three years.

JB: What does this mean, of course?

JA: A huge increase in supply naturally sows the seeds of the end of the cycle. I guess we see a decline in price, as you say, as that increase in investment was replicated across the industry. There was a good chart from Rio Tinto at one of their capital markets days that pointed out that while the whole industry had invested less than $15 billion a year in 2010, by 2012 that had roughly tripled to more than $40 billion. Industry-wide, a huge increase in supply naturally leads to a decline in prices. I think we've got a chart of the iron ore price. The iron ore price collapses from that peak of more than $200 a tonne to something much more sensible, I'll try to find my chart and my notes.

JB: I think it fell to about $50. I think history is that you went from $25 to $153, and at the very peak, $200.

JB: Then a few years later, once all this new supply came on, the price fell back to $50. I think the lesson here is that people tend to look at demand, but they forget supply. In markets without barriers to entry, supply is equally important to the economic equation as demand. With this, Rio itself sowed the seeds of its own destruction. I think at the peak of Rio's profitability in 2011, it earned operating profits of around $20 billion. But that fell back to $6 billion in 2015. Its response, of course, was then to cut CapEx. It cut what it was spending and didn't look to add new supply. Its CapEx went from about $17 billion in 2012 to just $5 billion a few years later.

JA: That tightness of capacity of investment was replicated across the industry. And that same chart from Rio Tinto shows that at an industry level, total investment dropped to think below 10 billion by 2017.

JA: We see that the low prices lead to a drop in investment and capacity effectively exiting the industry.

JB: With time, demand catches up and prices recover. This is just a story that repeats itself throughout economic history.

JA: Maybe that's a good time to move on to another example. I think the oil price boom and shale oil is another good example of a commodity product where a high level of prices and a good level of profitability draw in new capacity, which itself leads to a collapse in prices. We should talk a little bit about the backdrop here. The oil price coming out of the financial crisis was around $100 a barrel. Talk us through what was driving demand for oil at that point. Then we can talk a little bit about what might change on the supply side.

JB: After the financial crisis, the oil price was about $100 a barrel, which was very, very high by historical levels. Again, there was a lot of demand from industrialising countries like China. At $100, most oil producers were making pretty good money. What really changed in that period was the shale oil revolution. In the United States, oil production had been falling for decades. The US had become a big net oil importer. This is why it was so interested in the Middle East, thinking about that geopolitical angle. Then around the start of the last decade, oil companies in the United States effectively invented a new method to get oil out of rocks that were previously regarded as uneconomic. I don't want to go into a geology lesson here, but they're called shale rocks. They're found in spades in places like Texas and New Mexico. There's one area in particular called the Permian Basin. Shale rocks are effectively like siltstone, almost like clay or mud, very, very tight rock.

JB: It became possible to extract oil that was in that rock by fracturing it. This is called fracking. You would put an oil well down into the ground, combine water, chemicals and sand at very high pressure, and basically blast the rock open. That fracturing process would allow the oil to flow and be collected. Because of this, American oil production for the first time in decades started to increase by several million barrels a day. A material amount in the context of the overall world oil market.

JA: US oil production had peaked in the very early 1970s at 9 or 10 million barrels a day. But as those conventional oil wells were exhausted, I think by the time of the financial crisis the US was down to about 6 million barrels a day and had become, as you say, a net oil importer.

JA: But this new technology really changed everything. It changed the game and led to a really big increase in capacity, a really big increase in the supply of oil. I guess we're leading the witness here, but what happens with this huge increase in supply? What happens to oil prices as all this extra capacity comes online?

JB: It's exactly the same process, exactly the same trend. Production goes up, floods the market, prices come down and the oil price collapses from about $100 to $50. A lot of these shale producers were not economic. Over 100 shale producers went bankrupt. They just kept on drilling, kept on drilling, keeping the market oversupplied. Eventually the seeds of their own destruction were sown. We’ve got some figures here from one company that survived. It’s a company that's still flourishing today. It's called EOG. In 2010, their profits were $500 million.

JB: On the back of the shale revolution, that increased to $4.5 billion by 2014.

JA: So, a really big increase in a very short period of time.

JB: Huge boom, a huge boom. But then by 2016, they actually recorded a $6 billion loss.

JA: The decline in prices led to a drop in profitability, in their case significant losses. Lots of other people exit the industry. Industry supply gets curtailed. I guess one of the things that's very interesting about shale oil is that these oil wells that are being drilled have very short effective lives. They only operate for 12 to 24 months. This isn't like an offshore deepwater drilling project where it lasts 20 or 30 years. This capacity can exit the market very quickly when the oil price drops. These firms respond very nimbly to changes in the oil price. It's a good example of that ability of capacity and capital to exit an industry when prices are uneconomic.

JB: Yes, I think we can show a chart showing what's called the rig count. This is the number of drilling rigs in the United States. It shows how far they fell when the crash eventually happened. As you say, they're very short-lived wells. The supply can come out very, very quickly. Of course, when that happens, prices recover and then normalise.

JA: Those are two good examples from the past. Iron ore. Shale oil. What's the relevance of those to the AI investment boom, do you think?

JB: Well, I think the parallels to anyone who has studied these economic cycles are really, stark. We tend to think because it's technology, because it's AI, it's very, very complex. But we think that this has all the hallmarks of a classic capital cycle. What's driving this boom today? At the moment there are shortages of computing capacity. AI companies like Anthropic, OpenAI, and even Google with its Gemini product are reporting shortages of capacity, which are being created by strong levels of demand and are leading to very high prices for computing power.

JB: The issue is that this computing power itself, the provision of computing power, appears to us to be pretty commodity-like and more akin to heavy industry than something you would expect to see in the technology sector. We have numerous companies: Amazon, Microsoft, Google, Meta, which owns Instagram and Facebook, Oracle, and newcomers like CoreWeave and Nebius, all building data centres which basically provide computing power to AI companies like OpenAI and Anthropic. The numbers being spent are astronomical.

JA: The capacity that they're building is largely undifferentiated. Would you say a data centre from Amazon is different from a data centre built by CoreWeave? If you're Anthropic or OpenAI, do you have a preference, or do you think these are commodity products?

JB: I think there are slight differences in the semiconductors that they use, for example. But ultimately, I think this is a commodity product. I think that the buyer can switch between providers relatively easily. It seems like the level of differentiation is low.

JB: Our thinking is that this has all the hallmarks of being a commodity product. It's AI compute power, but effectively it's no different from iron ore.

JA: There's, as you say, a huge investment boom going on here. The current level of profitability is very good, and that is drawing in more investment and more spending. If we go back a few years, maybe about 2020 or so, the amount being spent on AI data centres would have been less than $50 billion a year.

JA: Just from pocket change. Yes, by last year. So, by about 2025, that had increased to $400 billion, which is an enormous amount of money. Yeah, but if we look at forecasts from various investment banks, they expect that number to increase to $1.4 trillion by 2028. So even having had this huge run to hundreds of billions of dollars a year, they still expect the amount of spending to triple or more in just the next couple of years.

JA: As you say, this investment, what they're spending the money on, is commoditised. Everyone's buying processors, everyone's buying memory, everyone's buying power. There's low differentiation between the providers here. Perhaps this sets up the next cycle, as a huge increase in capacity might lead to oversupply in due course.

JB: Exactly. Once again, I think everyone's focusing on demand. Everyone's focusing on demand growth and they're ignoring supply. As we know from these previous cycles, supply is equally important as demand. To us, it has all the hallmarks of a classic boom. We know from history that quite often these booms are followed by busts. This would imply that at some point all this capital spending will be cut. Of course, that would have devastating consequences for many of these companies, the picks-and-shovels players that are supplying semiconductors to companies like Amazon, Microsoft and others that are spending literally hundreds of billions of dollars today.

JA: Julian, if we come back to the central question, what do you think history teaches us about today's AI boom?

JB: My view is that there are many similarities with previous capital cycles that we need to watch. Supply is as important as demand. The amount being built today and the amount being spent today does risk creating oversupply. The economics on some of these investments may be very poor. Therefore, at some point, companies may start to cut back on their spending, and with that the boom turns to bust.

JA: Yeah, that’s a feature of commodity markets ultimately, that supply responds to levels of prices in ways that ultimately undermine the high profits that the industry has been making. I think that's a good observation and a good point to end the discussion. Thank you, Julian.

JB: Good chat. Thanks, James. thank you for watching and listening to Connected Investor. If you enjoyed this episode, do follow or subscribe and we will see you next time.

JB: Thank you.