Rates just screamed to 20 year highs and stocks have barely blinked. That looks to me like the mirror image of the mid-2010s, when Treasuries yielded 2%, the math said stocks should trade for 25x, and they sat in the mid-teens instead because the equity risk premium quietly widened from 4% to 6%. If the premium can widen when rates fall, why would it not widen again when rates rise? That is the double whammy running in reverse: earnings that got a decade of help from the Trump tax cuts and the AI boom, multiplied by a multiple heading the wrong way.

The other thing I cannot stop chewing on is what higher rates do to the AI data center buildout. These are 15 year leases where the NPV of the payments roughly covers the build cost, which means the developer is really underwriting the terminal value 15 to 25 years out. Move rates from 4% to 5% and you have to jack the lease rate up 5% to 10% just to stand still, and you discount that terminal value harder, right as the tenant credit gets scarier. If the AI trade cracks, you get hit twice: your tenant may not be around, and the release in year 15 goes from a $100m NOI lease to whatever the next best bidder pays. I do not think we are there yet, but finance 101 says investment gets crowded out eventually.

Then two management questions. UWMC and Cogent both ran capital allocation that looked designed for the CEO's personal balance sheet rather than for shareholders, and I want a way to spot that before the blowup rather than after. And a friend's text about a CEO everyone was calling the next Mark Leonard got me wondering how you would ever know, because a real compounder and one great bet with hidden leverage look identical for the first ten years.

I wrote the rates piece up this morning: https://www.yetanothervalueblog.com/p/rates-are-screaming-and-stocks-arent

The UWMC post: https://www.yetanothervalueblog.com/p/uwmc-lost-600m-hedging-a-deal-theyd

The Cogent episode with Aaron Chan: https://www.yetanothervalueblog.com/p/recurve-capitals-aaron-chan-on-cogent

This episode is sponsored by Trata: https://trata.com. Trata is two buy siders talking to each other about a name they both follow closely. Trata records it, anonymizes it, and publishes it. It is the fastest way I know to get up to speed on something new.

Chapters:

(00:00) What is on my mind this month

(01:07) Sponsor: Trata

(01:41) Rates screamed higher and stocks did not listen

(04:39) Should the equity risk premium rise with rates?

(06:29) Rising rates meet the AI data center buildout

(09:33) What a 15 year data center lease is really betting on

(13:03) Does higher for longer start crowding out AI capex?

(14:11) UWMC, Cogent, and CEOs who run capital allocation for themselves

(18:45) How would you know if someone is the next Mark Leonard?

(23:01) One great bet, or actual genius?

(24:37) Wrapping up

Links:

Yet Another Value Blog - https://www.yetanothervalueblog.com

See our legal disclaimer here: https://www.yetanothervalueblog.com/p/legal-and-disclaimer

Late August 2026 Random Ramblings

Yet Another Value Podcast

Late August 2026 Random Ramblings