Inner Mongolia had long stopped being remembered for Genghis Khan as much as open-pit coal mines. If building lots of AI data centers means that the Mongolian grasslands can be spared of stripping for coal it’d indeed be a great thing!

Everyone is speculating about when cheap debt stops for the AI infrastructure boom. It doesn't stop, it shifts from expensive to cheaper formats. The debate is not "build or don't build." Folks are moving to standardized formats less than 25MW. 50% cheaper and built in 120 days. 🧵

Chinese city govt answers to central government so it's unthinkable in China for a private company to host high level govt meetings in any city, but US federal govt can't order the city government to prepare and host high level meetings. Indeed a big contrast btw US and China.

Taylor Swift’s husband invested in a Ponzi scheme called “Phillip Morris Pakistan” operated by an investor by the name “Jawahar” and people still don’t take football concussion injuries seriously.

I said last night in response to this post that everything was technically a prediction market bc Arrow-Debreu contingent claims span the returns space. I want to correct myself. Everything is trivially a string of predictions but you couldn’t decompose a bond into a sequence of prediction contracts. Now, in reality, very few long-lived assets are prediction markets in any meaningful sense. I appreciate where Joe’s quip might come from. There’s a pedantic sense in which his statement is right. But ultimately he’s wrong in an interesting way that lies at the heart of financial/asset pricing theory. Joe famously likes to say that he doesn’t believe in the concept of bond “term (risk) permia” and instead believes that the returns of a bond, even a long maturity one, is simply the outcomes of a sequence of Fed short rate decisions. Viewed this way it’s tempting to think how any Treasury bond may be simply a sequence of properly lined-up FOMC decision prediction contracts. Except this is simply not true for exactly the reason why Joe doesn’t believe in the concept of “term permia”. There’s a name in economics for what Joe believes and it’s called the “expectations hypothesis”. Under the “strong version” of EH, a n-year bond yield is literally the market’s forecast of the average short rate over the next n years. There should be no term permia and long term bonds should earn no excess returns over short term bills. The expectations hypothesis has been repeatedly and resoundingly rejected in many different studies: Fama-Bliss, Shiller, Campbell-Shiller, Cochrane-Piazzesi etc. Whether it’s market-implied or survey-based, short rate expectations cannot explain long bond yields/returns. Bond prices move way too much to be explained by plausible changes in the expectations of future short rates. Bond yields are not simply forecasts of short rates. Bond term permia exist even if they aren’t measured well! You cannot line up a daisy-chain of prediction contracts on FOMC meetings to synthetically recreate the payoffs of a bond. Love you Joe but sorry bro.