Stripe Pilots Token-Based AI Billing
Stripe has launched an experimental private preview for LLM token billing, allowing developers to monetize AI features based on token consumption. The system mirrors pricing models from OpenAI and others, and even lets platforms add a markup. This signals a shift toward more granular, usage-based monetization for AI-powered APIs and SaaS products.
Stripe's new billing tool automatically tracks the API pricing for selected AI models, records token consumption for each customer, and applies a pre-set margin before generating an invoice. This automates what was previously a manual process of reconciling provider costs with customer billing. The system is designed to address the cost volatility that developers face when using large language models from providers like OpenAI, Google, and Anthropic. Without a way to pass on these fluctuating, per-token costs, companies risk eroding their margins as customers increase their use of AI features. This represents a shift from the common practice of monthly subscriptions with usage caps, where companies often have to absorb the cost of heavy users or charge overage fees. The new model allows for more direct, usage-based pricing that can be a managed revenue stream rather than just a variable expense. Developers can set a consistent margin, for example, 30%, on top of the raw token costs they pay to the LLM providers. Stripe's system monitors for price changes from the model providers and can be configured to automatically apply updated pricing to new or all customers. Beyond just billing, Stripe has also introduced its own AI gateway, allowing developers to access and switch between different models. The billing feature also integrates with popular third-party AI gateways such as Vercel and OpenRouter, providing flexibility for existing development stacks.
