Foreign media reported that Stripe acquired the AI routing platform OpenRouter for over $7 billion—a deal significant not just for its price, but for bringing model invocation, usage billing, and payment settlement under one company. For Stripe, this represents a further expansion into AI infrastructure beyond payments.
What was purchased is the distribution entry point.
OpenRouter provides developers with a unified interface to connect to over 400 AI models. Developers can switch between different models using a single API key, without needing to integrate separately with multiple model providers. According to the article, OpenRouter does not own GPUs or train models; its primary revenue comes from a commission of approximately 5% on model inference expenses.
The article argues that what Stripe truly values is not the capability of any single model, but OpenRouter’s position—it sits between approximately 8 million developers and major model labs, controlling how model requests are routed and how fees are structured.
Previously, Stripe handled OpenRouter’s payments, invoicing, and tax processing. After the acquisition, model routing decisions and billing generation will be integrated into the same system, creating a closed loop from invocation to charging.
The acquisition continues the deployment of AI infrastructure.
The article mentions that Stripe has recently acquired several related companies, including the stablecoin company Bridge, the wallet infrastructure company Privy, and the pay-as-you-go company Metronome, the latter of which is already used by OpenAI and Anthropic.
In addition, Stripe is also involved in developing Tempo, whose Machine Payments Protocol aims to enable AI agents to initiate, authorize, and complete payments without human intervention.
In this context, OpenRouter’s value lies in extending Stripe’s existing capabilities further upstream. It doesn’t just process payments—it begins to engage with model selection, invocation pathways, and enterprise AI budget allocation.
Platform Neutrality and Pricing Pressure
The article argues that what developers are most likely to lose is a simplified version of platform neutrality. One of OpenRouter’s previous selling points was allowing users to choose among different models based on budget, latency, and performance, without the platform itself favoring any specific model provider.
However, after the acquisition, who sets the default routing and which models receive more traffic may become more prominent. The article also notes a shift: the share of U.S.-origin models in OpenRouter’s token usage has dropped from approximately 70% in mid-2025 to around 30% a year later, with some share absorbed by cheaper Chinese open-source models.
For model companies, the growth of routing platforms may also weaken their pricing power. As long as developers are willing to substitute cheaper models for expensive flagship models, platforms can capture more distribution value, leaving model providers with less bargaining power over end customers.
The article cites OpenRouter’s Fusion API, which can distribute a single prompt to multiple low-cost models simultaneously and then combine their outputs. In the DRACO benchmark, a model combination consisting of Gemini 3 Flash, Kimi K2.6, and DeepSeek V4 Pro achieved a score of 64.7%, outperforming GPT-5.5 alone at 60% and Claude Opus 4.8 alone at 58.8%. This indicates that the routing and aggregation layers themselves are becoming sources of value.
Additional information: As of publication, neither party has publicly confirmed the transaction, and the regulatory review timeline has not been disclosed. The article states that Stripe processed $1.9 trillion in payments in 2025 and was valued at $159 billion in a tender offer in February 2026. If the transaction amount exceeds $7 billion, it will become Stripe’s largest acquisition to date.