[Week 31 of 2026] Leverage and Labels

Welcome back to Price and Prejudice with a few musings from Week 31 of 2026.

Situational Awareness

This WSJ article opens with Leopold Aschenbrenner's wedding guests arriving in Carmel, complete with a pre-wedding colloquium of panels and breakout sessions, while his fund came apart behind him. I mentioned Aschenbrenner a couple of times in class as one of the many faces representing the AI revolution. His trajectory is quite impressive – he entered Columbia at 15 and graduated valedictorian at 19, spent a few months at OpenAI, and wrote the 2024 manifesto that made him famous. He turned that into Situational Awareness, a hedge fund that reached roughly $45 billion and was reportedly up 439% net in the first half of this year.

Unfortunately, by the end of July it was around $10 billion. The fund ran leverage of about four times, its three prime brokers called margin, and allegedly the entire public book went to Citadel in one negotiated sale at below-market prices. He was basically long AI infrastructure, names like SK Hynix and CoreWeave, and short application software. That looks hedged on a risk report, long some tech and short other tech, but reality is much messier than that.

If you listen to a lot of podcasts featuring fund managers, the single most thing they all emphasize is the importance of risk management. The idea and the thesis is usually the shortest part of the conversation, and most emphasis is on topics like position sizing, gross versus net exposure, what the correlation structure does in a drawdown, and what happens if you are right eventually but wrong for six months. This is not a coincidence, since in a levered book the path matters as much as the destination. The ability to hold leverage without people asking for their borrowed money back is an extreme privilege, and it's currently only allowed to a subset of investors.

I think this is also the reason the tech industry has never really taken over asset management, despite obviously having the talent to. The venture playbook is built for a world with no leverage, no daily marks, and ten-year lockups, where being wrong nine times out of ten is the design rather than a failure. Public markets with borrowed money, on the other hand, are a different game with a different loss function. Your positions get priced every afternoon, and your prime broker gets a vote on how long you are allowed to be right. Identifying the good opportunity is necessary and nowhere close to sufficient, and the gap between those two things is basically the entire profession.

Circling back to Leopold, the WSJ calls him the "Nostradamus of AI," but his manifesto was a confident extrapolation of scaling laws rather than a forecast that could have been falsified along the way. So the confidence inevitably reads as foresight when the underlying trend is going your way. Instead what was actually being priced by investors was aura. Apparently, Leopold was big into public speaking in his days in Germany, and so was Do Kwon of the Luna-Terra collapse fame. Competitive public speaking rewards holding a position and defending it fluently under pressure. However, it does not reward updating, and it certainly does not reward saying that you are less sure than you were last week. That is probably close to the opposite of the temperament risk management asks for, and a trait to look out for next time you ask someone else to manage your money.

Private Label

This piece profiles three RIAs that have launched or announced their own ETFs, and the trend behind it is large. It is worth being precise about why this makes sense for an RIA specifically, since the answer is not that these firms have suddenly discovered investment skill.

Every RIA of any size runs model portfolios and separately managed accounts, which means the allocation work is done and simply gets replicated across hundreds of individual accounts. Doing it that way is probably operationally miserable since a rebalance is not one trade, but one trade per account, each with its own tax lots and wash sale problems. As a result, the cost of running it scales with the number of clients rather than with assets. But put the same strategy in an ETF and it becomes a single portfolio, rebalanced once, with in-kind creation and redemption letting the fund shed low-basis lots without realizing gains. So there is some large cost saving from the RIA's perspective.

The second reason might be that listing a strategy converts a service business into a product business. By charging an expense ratio on the ETF, you can decouple growth from the headcount, which is the same reason a restaurant starts bottling its sauce. This also explains why the story runs on a site aimed at independent advisors. One of the interviewed managers in the article is explicit that a wirehouse would never let him list a proprietary strategy, so the option to become a manufacturer is now a real economic argument for breaking away.

Either way, it's worth remembering that private label works beautifully inside your own store (think Kirkland from Costco) but the moment you want assets from someone else's clients, you are no longer the shelf owner. So this strategy probably makes sense up to the size of your existing book, but it likely gets very hard immediately after.