Stripe announced an agreement to acquire OpenRouter, the model gateway that routes and optimises token usage across more than 400 models from over 80 providers and is already used by NVIDIA, Zoom and Lovable. Stripe frames it as the other half of profitability: it already optimises revenue across payment methods, authorization rates and fraud, and since last year has worked on token cost with products like Token Billing.
The argument for routing as a product is about variance, not price. Deciding which model handles which task, at which speed and at what price is a live tradeoff, and the pace at which models are released and repriced keeps moving the answer; OpenRouter evaluates each request and routes on task complexity, price, speed and reliability. Alex Atallah's framing is that intelligence will be multi-model, so developers need a neutral layer to orchestrate them — and Patrick Collison's is that real-world economic potential depends on making good use of scarce compute.
Putting that layer inside a payments company makes token spend a billable, optimisable line rather than an engineering preference, which is the pressure already visible in model routing and pricing updates across GLM, DeepSeek, Grok and Qwen and in Gimlet treating inference as multi-silicon scheduling. The AI hub tracks the economics.
This is an acquisition announcement: terms, timing and integration plans are not published, and the neutrality claim is made by both parties to the deal. The condition to watch is whether routing stays neutral once the router's owner also bills for the requests, because a neutral layer with a stake in the transaction is a claim that needs evidence.