Noah Smith
· Noahpinion
· March 28, 2026 at 08:50
· ⏱ 26 min read
| Read on Substack ↗
Summary
The article argues that AI may not eliminate high-paying human jobs because of the economic principle of comparative advantage: even if AI is better at every task, its finite compute creates opportunity costs that keep humans employed in roles where they have a lower opportunity cost than AI. This implies that markets should not fear mass labor obsolescence, but the analysis is purely theoretical and offers no actionable trade ideas.
•Comparative advantage means everyone always has something they do relatively best, even if worse in absolute terms.
•Compute is a producer-specific constraint on AI, similar to time constraints for humans, preventing AI from doing everything.
•Example: a VC (Marc) pays a secretary to type even though Marc types faster, because Marc's time is more valuable elsewhere.
•If AI becomes ubiquitous, its opportunity cost (the next best use of its compute) will prevent it from being used for low-value tasks, keeping those tasks for humans.
•Horses were replaced by cars because they competed for energy/land, not because of comparative advantage failure.
•Energy is a shared resource between humans and AI, posing a risk if AI bids it up too high.