Stripe has agreed to acquire OpenRouter, an AI model-routing platform that gives developers access to hundreds of AI models through a single interface. The deal brings model selection and routing closer to Stripe’s existing efforts around AI usage tracking, token metering, and billing.
OpenRouter supports more than 400 models from over 80 providers, allowing developers to access different AI systems through one API rather than maintaining separate integrations for each provider.
Routing AI requests based on performance
OpenRouter’s platform does more than provide access to multiple models.
The system evaluates requests using factors such as task complexity, price, response speed, and reliability before determining which model is best suited to handle a particular request.
The platform can also make a second routing decision after selecting a model. If multiple providers offer the same model, OpenRouter can select an individual provider endpoint based on criteria such as price, throughput, or latency.
Customers can establish requirements such as maximum prices or minimum performance levels, allowing the system to select an endpoint that meets their specified conditions.
The same model can have different costs
Provider selection can have a significant impact on AI inference costs even when the underlying model remains identical.
OpenRouter previously showed Llama 3.3 70B input pricing ranging from $0.10 per million tokens through one provider to $1.04 through another. Output pricing also varied between providers.
That makes routing potentially valuable for companies operating AI applications at scale. Rather than permanently sending every request to one provider, an application can select an available endpoint according to current cost and performance requirements.
OpenRouter also monitors latency and throughput for individual model-provider combinations using rolling performance data.
Routing can provide automatic failover
Model routing can also help applications remain available when individual providers experience problems.
OpenRouter says its system can redirect requests to alternative providers or models when it encounters issues including outages, rate limits, context-length errors, or moderation refusals.
Data-handling requirements can also influence routing decisions.
Customers can restrict requests to providers offering Zero Data Retention and prevent requests from being routed to providers that collect data or use prompts for training. Enterprise customers can also request processing within specific regions, including the US or EU.
This means routing decisions can account for more than price. Model capability, provider availability, latency, throughput, processing location, and data-handling policies can all factor into which endpoint receives a request.
Companies are increasingly using multiple models
The move toward model routing reflects the growing use of multiple AI models within organizations.
F5’s 2026 State of Application Strategy report found that 52 percent of surveyed organizations were chaining or orchestrating multiple AI models, with respondents using an average of seven models. The survey included more than 1,100 IT decision-makers.
OpenRouter is not the only company developing technology for this environment.
Snowflake has introduced dynamic model routing for its Cortex AI Gateway, while Cloudflare offers Dynamic Routing through AI Gateway. AWS provides Intelligent Prompt Routing through Bedrock, and Microsoft Foundry offers routing profiles designed to balance factors including model quality and price.
These systems can allow organizations to send simpler workloads to smaller or less expensive models while reserving more capable models for requests that require greater reasoning or response quality.
More models create new operational challenges
Using multiple models can reduce costs and improve flexibility, but it also introduces additional complexity.
Microsoft’s Azure Architecture Center notes that dynamic model selection can make cost forecasting, debugging, and performance analysis more difficult because different requests may be processed by different models.
Organizations therefore need to understand not only which models they are using, but also how routing decisions affect performance, spending, and application behavior.
Stripe already worked with OpenRouter
Stripe and OpenRouter were already collaborating before the acquisition.
Stripe said developers could use OpenRouter to route AI model requests while Stripe tracked usage, applied pricing, and handled billing. The arrangement effectively paired OpenRouter’s model-routing infrastructure with Stripe’s payment and usage-management capabilities.
The acquisition brings these capabilities under the same company.
Token usage becomes increasingly important
Stripe has been developing infrastructure for businesses that need to bill customers based on AI consumption.
Its LLM token-billing service can measure usage based on model and token type, including input, output, and cached tokens where supported. Businesses can use the system to create per-token pricing, prepaid credits, fixed fees that include a certain amount of usage, or combinations of these approaches.
OpenRouter already generates much of the usage information required for these calculations.
Its API reports prompt, completion, reasoning, and cached token counts for individual requests, along with the associated cost. The platform also records the underlying inference cost charged by the provider separately from the amount charged to an OpenRouter account.
AI workloads are generating enormous token volumes
The importance of token management is growing as enterprise AI deployments expand.
A Deloitte survey of 515 US-based business and technology decision-makers found that 37 percent of respondents were consuming between one billion and 10 billion AI tokens each month. Another 30 percent reported consumption above 10 billion tokens per month.
By 2028, 61 percent of respondents expect their organizations to consume more than 10 billion tokens per month.
Deloitte also expects workloads exceeding 100 billion tokens per month to become significantly more common. However, the firm cautioned that greater token consumption does not necessarily mean an organization is using AI more effectively.
Oversized prompts, inefficient context management, and limited reuse can all increase token consumption without necessarily improving results.
OpenRouter is already operating at massive scale
OpenRouter says it processes more than 10 trillion tokens each day across a community of more than 10 million developers and companies.
The platform reported that its weekly token volume increased from five trillion to 25 trillion over a six-month period. At that point, the company said it was supporting more than eight million developers and providing access to more than 400 models.
OpenRouter was founded in 2023 and has attracted investment from firms including Menlo Ventures and Andreessen Horowitz.
Its Series B funding round raised $113 million and was led by CapitalG, Alphabet’s independent growth fund, with participation from investors including NVentures, ServiceNow Ventures, MongoDB Ventures, Snowflake Ventures, and Databricks Ventures.
A major acquisition in AI infrastructure
The financial terms of the acquisition have not been publicly disclosed by Stripe or OpenRouter.
Reuters reported that the transaction is valued at slightly more than $8 billion, citing a person familiar with the matter who was not authorized to publicly discuss the deal.
For Stripe, the acquisition extends its role in the rapidly developing economics of AI applications.
As companies increasingly use multiple models, the ability to determine which model and provider should handle each request can become just as important as the underlying AI model itself.
By combining OpenRouter’s routing infrastructure with Stripe’s existing capabilities around usage measurement and billing, the company is positioning itself closer to the infrastructure layer that manages how AI applications consume and pay for computing resources.
Source: https://www.artificialintelligence-news.com/news/stripe-openrouter-acquisition-ai-model-routing/
