Stripe and IBM bet the model war is already over—routing is the

Stripe's $7B acquisition of OpenRouter signals a major shift: the economic value in AI is moving from the model layer to the orchestration and routing layer. Stripe is betting that controlling the tollbooth between users and models is more valuable than owning any single model. This mirrors IBM's strategy of becoming a neutral enterprise AI

integrator, partnering with both OpenAI and Anthropic rather than competing at the frontier. The panel argues that routing will eventually become commoditized, but for now, it's where the growth is. OpenRouter's token volume grew 5x in six months, and Stripe is buying that trajectory along with 8-10 million users. The real tension is that as

routing becomes crowded, the next battleground will shift to agent management platforms and governance layers. Ramp's AI Index data exposes a more sobering reality: AI adoption is deeply uneven. The top 1% of companies spend $7,400/employee/month on AI, approaching the cost of a full salary, while the median is just $12. CFOs are moving from

fear-driven experimentation to demanding measurable unit economics. The era of unmetered AI spend is ending. The panel's most practical takeaway is that AI cost is becoming a FinOps problem. Organizations lack the tools to budget, meter, and govern token spend across hundreds of agents and models. Without this, the total cost of ownership—including regulatory compliance