Summary
Scott Galloway tackles a listener question about whether the AI data center building boom is repeating the speculative excess of the dot-com era. Drawing on Bloomberg estimates that a third to half of all US data centers planned for 2026 could be delayed or canceled, he explains how hyperscalers’ actual demand is often overstated by developers filing multiple grid-connection applications as a hedge.
The discussion digs into the financing behind the buildout, citing an estimated $1.7 trillion in hidden AI infrastructure debt, much of it pushed off balance sheets into shell entities, and a circular arrangement where Nvidia both sells chips and finances the customers buying them. Galloway argues ordinary ratepayers and overleveraged developers, not the hyperscalers themselves, will likely absorb the fallout, with a possible government bailout given how central AI valuations have become to the broader economy.
The episode also pivots to Apple’s subscription strategy, with Galloway detailing how recurring-revenue models command far higher valuation multiples than transactional retail businesses, using his own company L2’s acquisition as an example. Listeners get a grounded, skeptical take on AI infrastructure economics alongside broader lessons on subscription business design relevant to any AI-adjacent company weighing pricing models.
📺 Source: The Prof G Pod – Scott Galloway · Published September 21, 2026
🏷️ Format: Podcast
