Agent Cost Economics: Token Costs and the $5 Trillion Bet#
Original (dev guide) | Original (macro) | Raw
Combined developer-level cost analysis (Cloudstar) and macro investment thesis (Vogel, Curious Compass). Fills the wiki’s identified gap on cost modeling.
Developer Costs (Post-Optimization)#
| Profile | Monthly Cost |
|---|---|
| Solo dev, daily Claude Code | $80-150 |
| Indie hacker + AI SaaS | $200-500 |
| Small team (3-5), production agents | $500-1,500 |
| Multi-agent builder | Variable ($50-100 per bad session) |
Without optimization: 2-3× these figures.
Five Waste Vectors (60-80% of tokens wasted)#
- File Reading Loops — 21K tokens for a 4-token fix
- Retry Loop Tax — 3× cost on three-attempt failures
- Over-Qualified Models — 60-70% of tasks work on cheap models (5-8× savings from routing)
- No Prompt Caching — 90% discount available, most don’t use it
- Context Contamination — 50K+ stale tokens resent per request in long sessions
Optimization Playbook#
Model routing, prompt caching (90% discount), RAG vs full context (60-80% reduction), session architecture (short + focused), scoped instructions (CLAUDE.md under 200 lines).
Macro Economics#
- $5T projected AI data center capex 2025-2030
- Token explosion: 1.3 quadrillion tokens/month (Google, Oct 2025), 8× increase in 8 months
- Per-token costs falling 85%, but total cost flat/increasing due to volume
- ROI scenarios: base case 3.2%, optimistic 14.6% by 2030
- Enterprise ARPU ($450-500/mo) vs consumer ($20-200) — industry viability depends on enterprise adoption
Three Analogies#
Metaverse (bearish: $5T white elephant), Railroads (likely: transforms economy, builders go bust), Airlines (nuanced: valuable but thin margins).