Meta's $135 Billion Question: Can AI Compute Turn CapEx Into A New Growth Engine?

My answer to the title: not yet, and maybe never the way bulls hope. Selling compute is unlikely to match the roughly 52% operating margin of Meta’s Family of Apps. Two things flipped it in July: a Bloomberg report on a Meta cloud business, and a paid model API. Suddenly, the capex looked rentable, not dead money. What I need on the 29th of July is proof that Meta Compute actually exists. Named customers would be very welcome. At 20x forward earnings, it's the cheapest of the big four, and 3.56 billion daily users mean Meta wins on distribution even if it never wins the frontier models.

Earlier this month, a Bloomberg report gave me a really good reason to revisit my thesis on Meta Platforms ( META ). According to the report, Meta may be building a cloud business to rent out.

Thematic. Top down. I often find the theme before I find the stock. My philosophy is that themes are often born quiet and die loud. I try to catch them while they're still finding their voice. When the music plays, I mainly chase pockets that rhyme with growth, momentum, perception shifts, and sometimes even the most absurd narratives (mostly AI-related). When the music slows and the tape deteriorates, I don't wait around. I raise cash/rotate out, and watch for the next setup. A parabolic run may trigger a similar move. During a bull run, you won't find much common ground between the deep value crowd and me. I liked the core ideas of deep value investors, and I briefly followed that philosophy. However, it demands patience, and the AI supercycle broke whatever patience I had left. The market changed, and so did I. My style is not set in stone. I’m mostly long when the music is playing. When it stops/slows down, I may dabble with shorts via put options, although it’s not my forte. My style i...