Why AI Has No ROI with Paul Kedrosky
Summary
Host Ed Zitron and economist Paul Kudrosky critique the AI industry’s lack of ROI, highlighting GitHub Copilot’s shift to token-based pricing that exposed previously hidden costs. They discuss structural risks like duration mismatches between debt-funded data centers and deflating token prices, arguing that massive capital expenditures may become stranded assets. The conversation also debunks inflated GDP metrics and the "AI dark output" narrative, suggesting current AI adoption produces low-value "slop" rather than genuine economic productivity.
Topics discussed
Sponsorship and show introduction
Guest intro and the AI ROI debate
GitHub Copilot pricing and token costs
Accounting for AI operating expenses
Run rates and structural debt risks
Comparison to Amazon and dot-com era
Square sponsorship segment
UnitedHealthcare and ZepBound ads
Asset reuse and the dot-com fallacy
Measuring AI output and GDP distortion
Square sponsorship segment
UnitedHealthcare and ZepBound ads
Radiologist AI and labor market myths
Data center debt and regional economies
Google's equity raise and market circularity
Closing remarks and credits
ZepBound, Seafoam, and final ads
Listen ad-free on Castria