Back
I remember the week I got married. It can get quite tense. You have a lot of things to organise, you want everything to be perfect, there are other people involved. And the other people may care more about certain things (flowers, seating plans, music choices) than you think is rational. I can only imagine what it was like for Leo Aschenbrenner, whose hedge fund lost $35bn in July and who got married last weekend. I mention this not out of Schadenfreude. Leo will be fine: his fund still has around $10bn in private market positions (mostly Anthropic), he is 24 and most of the great hedge fund managers have a blow up early in their career. The point is more that now this unwind is done (Leo’s book was transferred to Citadel last Wednesday) some of the technical pressure in the market has eased.
This has then put fundamentals, and in particular Q2 earnings, front and centre. And, so far, US Q2 earnings growth has been pretty exceptional: up 26% year on year, the highest in recent history. And this is without another 21% in gains from marking up private investments (Including SpaceX and Anthropic):
Plenty of this growth comes from the AI/technology sector. But it is worth noting that even the median US company is tracking at 12% earnings growth year on year (vs a 9% consensus estimate). The US economy remains strong, AI investment is a tailwind and the technical picture (both in the US, thanks to Leo, and in Korea) is cleaner. This has all been enough for many markets – after a wobbly June and July – to take out all-time highs this week.
Which brings me back to the question I was asked again by a client this week: are we in an AI bubble? My answer remains (unhelpfully): I don’t know. But I would be very reluctant to bet against the longer-term trend of earnings accruing to tech businesses. This remains very much in place and was reinforced again in Q2. This is the chart I always have in my head when asked about all things US and technology:
That said, I also have a slight contrarian streak in me. So, I liked this quote from Marc Andreessen (who runs a16z, a Venture Capital business):
The twelve most dangerous words in finance are ‘The four most dangerous words in finance are “this time is different”’.
The take I have developed on this over the last few years is “every time is different”. Historical parallels can always give some guidance, but they can, I think, only take you so far. There has been plenty of good work (most recently here) comparing the AI infrastructure build out with previous investment booms (canals, railways and most recently the dotcom fibre build out). To cut to the chase, these previous booms did not work out well for the people making the investments (in this case Amazon, Microsoft, Google, Meta and Oracle) but longer term we all benefitted from the better infrastructure the world had. This would make you pretty bearish those five hyperscalers.
But canals, railways and fibre-optic cables are a pretty small sample size to compare against. And the demand for AI models today looks to be a lot higher than it was for fibre-optic cables in the early stages of the dotcom boom for example. Those cables were built out in the 1990s and 2000s in the expectation that there would be plenty of internet traffic to fill them. That did turn out to be the case, but traffic arrived too slowly for many of those cable investments to be profitable. Today, the demand for AI is clearly here. And in a “real businesses spending real money today” kind of way. Although both are private companies, it is estimated that OpenAI and Anthropic’s combined run rate revenues have grown from around $5bn combined at the end of 2024 to c. $140bn today. And this growth looks to have reaccelerated in July:
And this ignores the explosive growth in cheaper, open-source (often Chinese) models which look to be taking market share from the big two US model providers. To be bearish on the US technology complex (and the AI infrastructure build out) I think you need to see some sort of wobble on the demand side for the latest AI models. You just don’t see any sign of that, or anything like that, happening today.
Chris Brown, CIO
cbrown@ipscap.com
The value of investments may fall as well as rise and you may not get back all capital invested. Past Performance is not a guide to future performance and should not be relied upon. Nothing in this market commentary should be read as or constitutes investment advice.
