Who:
- Chris Clark, COO of OpenRouter: runs the largest AI gateway, processing 28T tokens weekly.
What Happened:
- Agentic AI usage surpassed human token consumption on OpenRouter.
- Single agent tasks can cost 100x more than human chats due to heavy context loading.
- Most companies still budget AI costs based on simple chat interfaces.
Why It Matters:
- GTM teams must urgently revise AI cost forecasts to account for agentic usage.
- Companies relying on AI for sales or marketing will face unexpected infrastructure costs.
ARM Impact:
- SaaS Hoarders (Stage 2 (SaaS Hoarder)) will struggle with ballooning AI costs as they scale agentic workflows.
- AI Sprinklers (Stage 3 (AI Sprinkler)) must optimize token usage to maintain ROI on AI investments.
- ARM-native companies (Stage 4 (Autonomous Revenue Master)) will gain an edge by designing efficient agentic architectures.
What to Watch:
- Monitor OpenRouter's monthly token burn reports for usage trends.
- Expect AI vendors to introduce agent-specific pricing tiers by Q4 2024.
- Watch for startups offering token optimization tools as agentic AI adoption accelerates.