The latest 100 slides from SuperAI's Day 2 were dominated by one theme: how do you charge for AI?
The Numbers
- - 2 in 3 Forbes AI 50 companies now use usage-based pricing
- - Hybrid pricing models grew from 27% to 41%
- - AI companies are monetizing at breakneck speed — but costs are high, variable, and volatile
- - "Agents are on track to read more Stripe docs than humans by the end of 2026"
Stripe's Three-Stage Pricing Framework
Stage 1: Seat-based (traditional SaaS) Stage 2: Usage-based (pay per API call, token, action) Stage 3: Outcome-based (pay per result delivered)
Most AI companies are transitioning from Stage 1 to Stage 2. The leaders are already experimenting with Stage 3.
The Agentic Commerce Framework
Stripe presented their Agentic Commerce and Payments Framework. Key insight: agents aren't tools. They are hires. Each agent needs identity, payment capability, and accountability.
AI Interaction Levels
The spectrum from human-led to agent-led: 1. Human does everything, AI assists 2. AI does routine tasks, human approves 3. AI acts autonomously within boundaries 4. AI delegates to other AIs 5. Fully autonomous agent networks
Most companies are stuck between levels 2 and 3. The infrastructure for levels 4 and 5 is what's being built now.
What's Working
- - Cursor: $2B ARR on seat + usage hybrid
- - Anthropic: $47B valuation, enterprise contracts
- - Manus: $90M run rate in 4 months on waitlist-driven demand
AI revenue growth isn't slowing down. It's speeding up. But the pricing model that wins hasn't been invented yet.