Model-routing startup OpenRouter reports surging usage

The routing layer grows

OpenRouter, which directs requests across many different AI models, said weekly usage had climbed to tens of trillions of tokens. The figure reflects how much traffic now flows through services that sit between applications and the underlying models.

A routing layer lets developers choose among models based on cost, capability or availability without rebuilding their integrations, which becomes more valuable as the number of models grows.

Investing one layer down

The growth fits a broader investor shift toward the control and orchestration layers around AI, rather than the models themselves. As models become more interchangeable, the infrastructure that manages, routes and governs their use takes on greater strategic importance.

This reflects a maturing market in which value is increasingly captured by the systems that make AI practical to deploy at scale, not only by the models.

Why orchestration matters

Businesses running AI in production care about reliability, cost control and flexibility. Being able to switch or combine models, manage spending and maintain consistent performance are real operational needs.

Rapid usage growth suggests these needs are widespread, positioning orchestration and routing as a durable part of the AI stack. The challenge for such companies is to remain essential as both models and the broader tooling ecosystem continue to evolve.