Both Baumol effect and Jevons paradox may apply to the AI supercycle, as services with AI productivity gains get cheaper, while those without AI become costlier
and gets pricier as wages rise (Baumol effect). Alex Danco / @alex_danco : The Baumol effect (I still like calling it “cost disease") is topical these days - we wrote about it in the newsletter, since you're going to see a lot of it in the next decade
Context & Ripple Effects
The AI cost story is already uneven: per-token prices have fallen while developers’ total costs rose as reasoning workloads consumed more tokens. That complicates any simple claim that AI efficiency automatically translates into cheaper delivered services.
This argument adds a distributional lens to the supercycle. Recent work finding that AI can handle longer software tasks faster but still faces a real-world “messiness tax” suggests gains will vary sharply by workflow rather than spread uniformly across service sectors.
First-order effects
- AI-enabled service providers can lower unit costs or expand output where automation materially reduces labor per task; providers whose work remains difficult to automate face relatively higher labor-cost pressure.
- Buyers gain a stronger incentive to separate services by automatable task content rather than treat “AI adoption” as a uniform cost-saving program.
Second-order effects
- Lower prices in AI-productive categories can stimulate additional usage, so efficiency may raise aggregate demand for AI-backed services even as each unit becomes cheaper—the Jevons side of the argument.
- Firms will compete to redesign workflows around the capabilities that are improving, while retaining human capacity for exception-heavy work; the gap between nominal model prices and delivered-service cost remains central.
Third-order effects
- If the pattern persists, AI’s economic effect may be relative-price divergence: scalable, AI-intensive services become more abundant while labor-intensive, hard-to-automate services consume a larger share of budgets.
- The outcome depends on whether productivity gains diffuse beyond clean, codifiable tasks. Persistent implementation friction would limit broad cost relief even as AI-intensive categories scale.
The trend: The AI supercycle is shifting competition from access to models toward the uneven ability to convert model efficiency into lower delivered-service costs and higher demand.