Don’t Drink The Kool Aid
From Julien at Macrostrategy:
I have calculated depreciation schedules for the Hyperscalers, based on current data plus Bloomberg capex forecasts to 2032. Under the most generous assumptions on chip depreciation (six years), and assuming 12% annual growth in underlying net income for the hyperscalers (not assuming any fresh AI income), depreciation reaches 98% of underlying net income by 2033. This means that the most successful businesses in the history of capitalism, have to build something that repeats all their past successes, and their extraordinary current market dominance, just to break even on the trade. They need to do much more than that to come out ahead.
Just to be clear, they are going to have to build the commercial equivalent of another Amazon digital marketplace, Amazon Web Services, Facebook, Instagram, WhatsApp, Google/Amazon/Microsoft Cloud, Google Search, YouTube, Google Play Appstore & Android OIS, Microsoft Office suite, SpaceX, Twitter and Oracle database solutions, from scratch, in seven years. And that’s just to break even on the LLM AI bet.
This note explains why they will never achieve that, or anything close. For the sake of your portfolios, ‘don’t drink the Kool Aid’.
Positions: Long AMZN VS GOOGL VS MSFT VS