Stripe to acquire AI model router OpenRouter for reported $7.5 billion

Stripe agrees to buy OpenRouter

Stripe has agreed to acquire OpenRouter, the startup behind a single API that routes developer requests across more than 400 AI models. The deal was reported at $7.5 billion by The New York Times; the two companies confirmed the acquisition in a joint statement but did not disclose the price. According to the report, $1.5 billion of the consideration is earmarked for OpenRouter's founders, with the remainder going to investors.

The figure stands out against OpenRouter's recent fundraising. The company raised a $113 million Series B led by Alphabet less than three months before the deal, at a valuation of roughly $1.3 billion. Stripe is therefore paying close to six times a number set one quarter earlier — for a company that builds no AI models of its own.

What OpenRouter actually sells

Founded in 2023 by chief executive Alex Atallah, OpenRouter operates a routing layer: one endpoint that sends each request to whichever large language model suits the job, whether that is a frontier model from OpenAI, Anthropic or Google, or a cheaper open-weight alternative such as DeepSeek or Z.ai. According to the companies, the platform now processes more than 10 trillion tokens per day and serves upward of 10 million developers, with customers including Nvidia, Zoom and Lovable.

The strategic significance is that routing sits upstream of model choice. Many applications no longer select a specific model; they delegate the decision to OpenRouter. Whoever owns that layer influences which models receive traffic.

How it fits Stripe's AI stack

Stripe chief executive Patrick Collison framed the purchase as part of a broader push into AI payments infrastructure. "Stripe is building the economic infrastructure for AI, and together with OpenRouter we'll help businesses maximize profitability by routing their requests intelligently and spending their tokens efficiently," Collison said in the announcement.

OpenRouter joins a set of AI-focused products Stripe has rolled out over the past year: Metronome for real-time token usage tracking and billing, Tempo for micropayment settlement, and Privy, which gives AI agents their own stablecoin wallets. With OpenRouter in place, Stripe can observe, bill for and route token spend on both sides of an AI transaction, rather than only processing the payment at the end.

An unusual buyer

Notably, an AI lab did not win the bidding. OpenAI, Anthropic and Google all have reasons to want control of model routing, since it is the layer that helps determine which of their models actually gets called. According to TechCrunch, citing people familiar with the talks, Stripe had to outbid other suitors, including Databricks. A payments company outspending an AI infrastructure firm for an AI infrastructure asset indicates how central Stripe believes routing will become to the flow of money in AI applications.

The neutrality question

OpenRouter's appeal has rested on being neutral between model providers. Developers use it precisely because it has no preference among them; it simply finds whichever option is cheapest, fastest or best suited to the task. Under Stripe's ownership, that neutrality becomes harder to guarantee. Stripe earns money on payment volume and token billing rather than on which model a developer picks, but the company also has partnerships of its own to protect, including a newly expanded agreement with Google, announced at Stripe's Sessions conference, that lets businesses sell inside Google's AI Mode and Gemini app.

Neither company has said how OpenRouter's independence will be preserved after the acquisition. For developers who rely on the service to arbitrage between providers on cost and performance, the open questions are concrete: whether pricing changes, whether default routing shifts to favor certain providers, and whether a critical dependency now sits inside a company whose commercial incentives only partly align with theirs.

Why it matters

The deal is one of the clearest signs yet of consolidation in the AI infrastructure layer — the connective tissue between applications and models, rather than the models themselves. It also shows how payments and AI economics are converging: Stripe now aims to sit on both the routing and the billing side of every token a business spends. Finally, it exposes a tension the industry has not resolved: routing layers are only useful when they are trusted to be impartial, yet they are increasingly owned by companies with interests of their own. Whether Stripe can hold that trust will shape how thousands of AI applications choose their models — and how the revenue from those choices is distributed.