You’ll Pay for the AI Bubble Even if You Never Use It | Office Hours

Professor Scott Galloway delves into three distinct but equally compelling topics: the potential bubble in data center construction, Apple's strategic shift towards a subscription-based model, and practical advice for career advancement within an organization.

The Data Center Building Boom: A Looming Bubble?

Galloway draws parallels between the current data center construction frenzy and the dot-com bubble of 1999. He highlights concerns raised by a listener, a UK Army engineer officer, who questions the sustainability of speculative data center builds for hyperscalers. The listener points to the risk of incomplete sites, sunk costs, and the strain on existing power grids.

Professor Galloway emphasizes that many announced data centers are not yet real, with utilities receiving five to ten times more interconnection requests than actual builds. This discrepancy is attributed to developers filing multiple applications to hedge their bets, often without securing financing, power deals, or equipment orders. He cites reports indicating that nearly half of US data centers planned for 2026 are either canceled or delayed due to precarious supply chains.

The financial underpinnings of this boom are also precarious. Wall Street's sentiment has shifted, with 45% of investors now calling the AI bubble the "biggest tail risk." The build-out is heavily reliant on borrowed money, with an estimated $1.7 trillion in hidden debt, much of it pushed off-book into shell companies. This debt is often backed by chips, whose value can plummet overnight with technological advancements. Galloway predicts a slow-rolling "thud" rather than a dramatic crash, where the weakest, most leveraged players will fall first, while big tech survives. He warns that ordinary people, through utility bills, will ultimately bear the cost if demand for these data centers doesn't materialize.