Weekend Reading #371

This is the three-hundredth-and-seveny-first weekly edition of our newsletter, Weekend Reading, sent out on Saturday 4th July 2026.

To receive a copy each week directly into your inbox, sign up here.

*****

What we're thinking.

We have had the worst 2-day momentum sell-off in markets since covid. AI names across the board have been smashed. Memory has led the bloodletting. The question once again is whether this is the end or whether it's just another trimming of the excesses of the past few months. As ever we don't know. The so-called rotation has been rather mild (so far, anyway) and is more led by selling of AI names than rallying in much else. It's more of a mini rotation. It can gain steam of course but that is for those who look for how the bones land to work out. What has been really interesting this week has been the contrasting strategies of the leading memory names. While Micron last week chose to sacrifice near term upside for longer term stability in its long-term contracts with customers, SK Hynix has decided to let it rip. In contracts with its customers, it has not signed take or pay and is riding on spot prices into the sunset. Once again, this quandary is reminiscent of the commodity super cycle in the 00s. Some commodity producers hedged, and others did not. Those that did not benefited for a full decade or so with high spot prices. If the cycle extends in this case, SK Hynix wins. If not, Micron will look smart in hindsight. But here is the thing. Markets prefer an unhedged way to express an opinion on something. If the cycle collapses, you don't want to be long anyway. And in reality, the only way the memory producers can impact their own multiple is to have oligopolistic supply discipline. That will extend the cycle and that may lead to better multiples but so far there isn't much evidence of supply discipline. They believe they are protecting their shareholders but in reality, their shareholders want to extract maximum from a potential super cycle. Time will tell what the correct play was, but this is a critical juncture in assessing the pecking order of these names. Elsewhere, Microstrategy announced a new corporate restructure which clearly protects their preference share vehicle. It protects from a near-term liquidity event and death spiral, but it really does bring into question the reason for it to exist at all.

What we're reading.

Semianalysis has become a preeminent analyst of the AI revolution and this week after Meta announced it will rent out its compute, they produced a note detailing how as opposed to the initial market reaction, this is not what it appears to be. In fact, they believe Meta will INCREASE its compute capex next year and they list their reasons. One of these reasons which interested me is the evolution of their own foundation model called Recsys, which is an internal advertising recommendation system which is apparently dramatically increasing outcomes for clients. I don't really know how it works to be honest but in principle this kind of thing is exactly what the promise of increased AI-driven productivity is about. Meta is getting better outcomes across its existing suite of products due to this system. Now imagine what that will be like for everyone else as AI-use proliferates across the broader economy. Interesting on many counts! DC

When it’s said that history doesn’t repeat but rhymes, I didn’t expect that it would rhyme to this extent – this post about a 1,500 page report published 100 years ago on the USA’s 150th anniversary entitled “Recent social trends in the United States” (recent for back then) is quite a read. Bear in mind that 100 years ago in 1926, the first world war had just ended, the great depression was around the corner (of course they didn’t know that then) and a revolution of mechanisation especially with cars and mobility was well under way. Yet the issues that bothered society were similar – it’s too much to truly summarise but I’ll extract this bit where the report tried to make some predictions about the future:
 
It may be that the world will find much use for talking books; school and college students may listen to lectures by long-running phonographs or talking pictures; moving pictures may be transmitted by wireless into houses; seeing with that new electric eye, the photo-electric cell, and recording what is seen, appear to have almost unlimited applications; new musical instruments different from any now in use may be given to us by electricity; the production of artificial climate may become widespread; an efficient storage battery of light weight and low cost might produce changes rivaling those of the internal combustion engine. And these are only a few of the myriad possibilities from new inventions in the future!
 
Seems they got most of it right – including the part where the critics worried that “modernity” transformed values and social structures, sometimes for the worse. So maybe history didn’t “repeat” or “rhyme” per se, and we’re just in the same long arc that was already in motion 100 years ago. Worth pondering. EL

What we're watching.

I was away this week so not much deep reading, writing or listening to report from me but this was a bit of fun. What is the only thing that can stop France from winning this world cup? Why it’s Menaldo of course. A short, animated clip in Japanese anime style combing the two super-footballers. DC

Everyone’s trying to get a bit excited about robotics now, but sometimes excitement needs a bit of moderation – first we had the Chinese trying to one-up the Japanese toilets with a toilet that comes to you, though hopefully the lessons from robot vacuum cleaners that don’t dock properly or have waste leaking have been learnt. Then we had a company called Weave Robotics releasing its slightly creepy household robot called Isaac (doesn’t have legs though, so presumably doesn’t work with staircases, unless it levitates like a Dalek in which case we’re in trouble), which I actually had to double check wasn’t an AI slop video prank – it’s real. And then there’s the actual AI-generated prank for a self-driving baby stroller which replicates the feeling of a car ride to supposedly lull babies to sleep which is thankfully all satire, but nonetheless makes a joke that isn’t entirely out of the realm of the possible. So, if things go wrong with a robot, whose fault is it though? EL