Better Offline The Reality of AI Economics With Paul Kedrosky
30 snips
Apr 7, 2026 
Paul Kedrosky, an economist who studies tech markets and data center investment, breaks down the economics behind massive data center land buys and the brittle NVIDIA GPU market. He discusses how data centers shaped recent GDP, powered-land speculation, long-term purchase agreements, and why inference efficiency and new chips could upend current finance models.
AI Snips
Chapters
Transcript
Episode notes
Track Fixed Investment Not Productivity For AI Signals
- Understand non-residential fixed investment as a category when tracking AI's economy.
- Monitor factories, equipment, and data center build-out instead of expecting immediate productivity gains in GDP metrics.
Data Center Buildouts Drove A New Debt Wave
- Tech companies moved from self-financing build-outs to heavy external financing as data center cash needs exceeded free cash flow.
- By 2025 data center debt became a major slice of US investment-grade and other debt issuance.
Debt Meets A Deflationary Revenue Commodity
- Lenders treated data centers like real estate, but the revenue commodity (tokens/inference) is highly deflationary.
- Fixed debt obligations clash with rapidly falling unit prices, creating a structural risk for financed projects.