Zuckerberg, profiled: how bad news reaches the top, and the wager under Meta's AI spending

Company strategy: the court around Zuckerberg, and the wager underneath Meta

"An Interview with Colossus EIC Jeremy Stern About Profiling Mark Zuckerberg" — Stratechery

1. The profile is organized around the information problem of holding absolute power. Stern's frame is the "dictator's dilemma": the more power a person accumulates, the more they depend on other people for information, and the harder those people find it to deliver bad news. Zuckerberg "frequently engages the direct-reports of his direct-reports" to reach past the courtiers standing between him and the work. An internal project his own team believed was failing carried the nickname M.M.H., short for "Make Mark Happy," because the people running it felt obliged to keep it alive. He also spends three mornings a week in a fighting studio at his Lake Tahoe house with his MMA coach Dave Camarillo and the former UFC bantamweight champion Merab Dvalishvili. Camarillo's account of why: "Mark filled the void with fighting. I think he does like the element of danger. If you go left when you should have went right, you get punched in the face." 3
2. Underneath Meta's AI spending, Stern lays out a wager with a winning branch on each side. If AI commoditizes, Anthropic and OpenAI go to zero and the value settles in what Meta already owns — distribution, attention, personalization and commerce; if capability stays scarce, Meta needs only to sit within about six months of the frontier, and it owns the data centers and chips its competitors rent. The numbers behind the bet: as much as $145 billion of spending this year, four data centers of a gigawatt or more with one planned to scale to five, and an install base of 3.6 billion people who, Stern writes, care little whether a model trails the frontier. Since June, he reports, Muse Spark has passed Google's Gemini and put Meta back in the frontier pack. The rival labs and investors he interviewed expect the spending to come to nothing, and they still refuse to bet that he fails. 3
3. Zuckerberg's own account of running Meta is about holding a team together, and he says outside opinion weighs on him less than on his peers. "I probably care about team cohesion more than most other people do," he tells Stern, and describes his method plainly: "My main tool in the world is sitting down, building a team, and methodically improving something over a long period of time. It's not a very sexy process." He contrasts himself with Sam Altman and Elon Musk, whose capital-intensive early companies, he says, "had to be memed into life," and describes a taste for being liked by strangers that he does not share. 3

Where to start