Stripe has entered its second era, one where payments infrastructure is converging with AI infrastructure. And it is expanding that footprint so aggressively that it is willing to pay more than six times OpenRouter’s $1.3 billion valuation from just a few months ago. 

The financial services and software-as-a-service company has agreed to acquire the AI model-routing platform for more than $8 billion, its largest acquisition to date. The companies have not disclosed the price; Reuters reported it at slightly above $8 billion.

OpenRouter gives developers a single interface to access and route requests across more than 400 AI models from more than 80 providers. It evaluates requests based on factors including task complexity, price, speed and reliability. Stripe says the combination will help businesses “maximise profitability” by routing requests intelligently and spending their tokens efficiently.

Since its founding in 2010, Stripe has largely focused on building the financial infrastructure that lets businesses operate on the internet, from accepting payments and managing subscriptions to billing customers and moving money.

But over the past year, Stripe has increasingly built infrastructure for the AI economy. It has developed tools to track token usage and help AI companies bill customers based on consumption, while also tackling rising and unpredictable model costs.

In January, Stripe acquired Metronome, a usage-based metering and billing platform. It then introduced Token Billing in preview, allowing AI companies to meter token consumption, apply their own pricing or markup, and bill customers based on usage.

At its April 2026 Sessions conference, Stripe went further, showing infrastructure for real-time metering and billing of AI usage, including individual tokens. The broader idea is that as AI software moves toward consumption-based pricing, Stripe wants to handle the infrastructure underneath that usage—measuring it, pricing it, billing for it and eventually moving the money.

OpenRouter adds a layer closer to the actual model request. Instead of simply billing after tokens have been consumed, Stripe can now potentially connect the billing system to the decision about where those tokens are spent.

Getting Rid of AI-Driven Billing Complexities

In an interaction with AIM, Varun Mishra, Principal Analyst at Counterpoint Research, pointed to the growing complexity of that problem.

“Billing for AI is meaningfully more complex,” Mishra says. “It involves a multitude of variables like which model is used, which provider serves it, and whether tokens are input, output, or cached, with each being priced differently.”

Those variables also change as providers launch models and revise their prices. 

Agentic workloads make accounting harder as one task can trigger dozens of model calls across different tools, users and projects, making it harder to attribute the underlying cost to the customer or workflow that generated it.

“And all this needs to be done at a scale of trillions of tokens per day, which is only going to increase in future,” Mishra adds. “Hence there is a need for a platform to measure and provide a reliable way to bill tokens accurately within complex workflows.”

Stripe has already been building towards that problem. Token Billing currently synchronises token prices for models from a select group of providers, including OpenAI, Anthropic and Google, automatically updating those prices when providers change them or introduce new models. 

Mishra said OpenRouter could expand that coverage substantially because it already sits across hundreds of models and dozens of providers.

“Bringing OpenRouter in-house should help close that gap directly, extending automatic price-tracking to dozens of additional providers OpenRouter already covers,” Mishra says, adding that OpenRouter could also give Stripe visibility into how developers are actually choosing and consuming models, allowing its billing infrastructure to evolve around real usage patterns.

Unifying Inference Bills and Revenue

This acquisition creates a more interesting possibility than simply finding the cheapest model for every request.

Most model-routing decisions today are based on performance per dollar. Companies generally send a request to the cheapest model that meets their quality and cost requirements. But whether that was the right decision for the business often becomes clear only later, when the inference bill and the revenue generated by that customer sit in separate systems.

“Stripe would be in a position to close that gap inside one system,” said Manish Choudhary, CEO and Co-founder of Flexprice, a billing platform for AI-native companies. “It would see what a request cost to serve through OpenRouter and what the developer charged their end customer through Stripe Billing and Metronome, on the same account, in the same week.”

Bringing those two numbers together could allow routing decisions to be made against a target margin rather than simply the price of a token. A lower-value customer could be routed to cheaper models, a workload that becomes unprofitable could be capped before the billing period ends, and pricing plans could eventually be based on the model mix customers actually consume rather than an estimate made months earlier.

Choudhary sees an even larger opportunity in financial services built around those economics. If Stripe can see both a company’s inference spending and the revenue generated by that usage, it could potentially underwrite credit based on observed gross margins, with repayments tied to the revenue generated by those AI workloads.

“Neither company could have built that alone, because neither one could see both halves of the ledger,” he says.

The Neutrality Question

But Choudhary cautions against assuming that the routing layer itself is the durable asset Stripe is acquiring. Routing is becoming increasingly competitive, with Vercel, Cloudflare and the major cloud providers offering their own ways to move workloads across models. 

That could make the basic routing function increasingly commoditised and put pressure on the fees OpenRouter can charge. OpenRouter currently charges the 5.5% fee when users purchase credits.

He also expects OpenRouter’s current 5.5% fee on credit purchases to eventually disappear, arguing that Stripe is more likely to use the platform to drive a much larger volume of AI transactions and monetise services around that flow than preserve a fee on inference usage. 

The harder asset to replicate, he says, is OpenRouter’s neutrality. The platform does not operate its own model and historically had no direct financial interest in which provider handled a request. Stripe’s ownership changes that dynamic because it will control the routing layer while also processing payments and revenue for many of the developers using it.

“That’s the thing I would watch rather than the price,” Choudhary says. “Would you keep production traffic on a gateway owned by the company that also settles your revenue?”