Bloomberg Tech Special Episode: Here's Why AI Costs Still Worry Investors
25 snips
Nov 22, 2025 Tom McKenzie, host of Bloomberg Tech Europe and a keen analyst on tech investing, dives into the complex world of AI data centers and the hefty costs associated with them. He discusses the investor worries surrounding operational expenses and rapid chip obsolescence, especially highlighted by Michael Burry’s cautionary thesis. McKenzie also explores the potential revenue gaps facing companies amid soaring infrastructure costs, while weighing the justifications from hyperscalers and forecasting the growing need for patience from investors by 2026.
AI Snips
Chapters
Transcript
Episode notes
Rapid Chip Depreciation Risk
- AI accelerators depreciate rapidly as newer, far more powerful chips arrive frequently.
- That rapid obsolescence risks stranded capital for companies that overinvest in today's hardware.
Michael Burry's Short-Position Example
- Michael Burry famously shorted Nvidia and Palantir citing concerns about asset depreciation.
- His position highlights investor skepticism about the long-term value of current AI hardware.
Dot‑Com Era Parallels
- Critics compare today's AI buildout to the dot-com era telecom overbuild that burned vast capital.
- The lesson is that infrastructure bets can fail if product-market returns lag expectations.