Over the last three months, the UK stock market has been moving essentially sideways in a challenging and volatile manner. Each move up and down has been another tricky geopolitical opportunity/trap. It seems that, in the US, investors have now discounted the end of the war around twenty-five times. Most of these ups and downs, which have led essentially nowhere, have been driven by news flow (which often turns out to be unreliable) about the situation in the Persian Gulf. Rather than try and guess which headline is real and which is not, we have continued to act defensively and have chosen to wait the situation out.
In the meantime, there are a lot of other things going on. We will not dwell too long on AI and space, but it would be remiss not to note that the upcoming IPOs of SpaceX, OpenAI and Anthropic will, between them, raise more money than all the companies in the dot-com bubble combined. It is estimated that these IPOs will, between them, be equivalent to nearly 8% of the whole value of the S&P 500. While active managers can act ahead to make some space for these names, it will be hard for index funds to do the same. It is likely that there will have to be some notable selling of ‘everything else’. Also, it seems that SpaceX is going to come on a valuation of around 100x sales, and all three of these companies are losing epically large amounts of money. It would only be poetic if these issues mark the very top of a bull market that has made the dot-com era look relatively reasonable. With the overall US stock market trading at notably over 40x a cyclically adjusted price/earnings ratio, the echo of the year 2000 seems very real indeed.
While all of this is going on, we have been concentrating on bringing our stock portfolio into a defensive position and have also added a few new names. We cannot help but think that some of the early beneficiaries of AI will be in an area that we call the ‘paper shufflers’. These are businesses that have a strong bureaucratic aspect. Good examples of these sorts of businesses are banks and insurance companies. A recent example is Standard Chartered. Although Standard has not managed its PR very well at all, we could not help but notice that, at least partially because of AI adoption, the company has increased its medium-term return on equity target from 15% to 18%. At the moment, the company is running with a return on equity of around 10% and is trading on not much more than one times book value. If it can deploy technology to go from a return on equity of 10% to one of 18%, the potential for share price appreciation is clear. If Standard can do this, we suspect that other banks and insurance companies will be sure to follow. At least over the medium term, it does seem that there is potential for AI to increase corporate profitability, especially for companies in the service sector.
Longer term, we are becoming increasingly amazed by the true spending scale of the hyperscalers. According to the consultants McKinsey, by the end of the decade the hyperscalers may have cumulatively spent as much as US$7 trillion. To put this into context, there are around 8 billion people in the world, of whom just over 4 billion are defined as middle class. (It is worth pointing out, however, that the definition of middle class in some emerging economies can be as low as an income of US$20 a day.) Nevertheless, this would mean that the hyperscalers will have spent US$1,750 or so for every middle-class person on earth—or alternatively (assuming a family of two adults and two children), US$7,000 per family. The world has never seen anything like this. In real terms, the cost of the Second World War for the US is estimated to have been US$4.7 trillion. Almost certainly, in real terms, the hyperscale spend is going to be equivalent to the combined cost of WWI and WWII for the US. In addition, given that one of the claims of AI enthusiasts is that it is going to substantially replace middle-class work, we cannot help but wonder how this is all going to work. We are not sure what will be worse for markets: an AI that works or one that does not. Either way, this is almost certainly the biggest capital-spending gamble in world history.


