Better Offline AI Is Worse Than The Dot Com Bubble: Part Two
94 snips
Jan 28, 2026 A contrarian take comparing today’s AI investment frenzy to the dot‑com build‑out. Myth-busting on infinite compute demand and falling model costs. A deep dive into GPU economics, data center scale, rapid hardware depreciation, and Taiwan supply risks. Warnings about concentrated market exposure and potential financial contagion.
AI Snips
Chapters
Transcript
Episode notes
Demand Transparency Before Backing AI Funds
- Watch VC portfolio concentration in AI and demand transparency from funds about exposure and valuations.
- Avoid assuming perpetual demand for compute; require clear profit paths before investing.
Dot-Com M&A Mania Burned Cash Fast
- During the dot-com boom some startups made money from ads but still burned cash buying other companies and expanding wildly.
- Ed Zitron recalls many ill-advised mergers and acquisitions that marked that era's hubris.
Fiber Had Long-Term Value Unlike GPUs
- Laying millions of miles of fiber had lasting, useful effects for internet access despite the bubble.
- That long-lived infrastructure is not comparable to short-lived GPU-based AI data centers.