Stripe Launches Token-Based Billing for AI Startups
Stripe just rolled out a token-based billing tool specifically for AI companies. The new feature allows startups to automatically track and monetize AI model usage, like LLM tokens, with built-in markups. It's a direct solution for the growing need to bill for computational use, not just flat subscriptions, positioning Stripe as core infrastructure for the AI economy.
Before Stripe's new tool, AI startups often relied on monthly subscriptions with usage caps, charging extra fees when customers exceeded their limits. This created a billing nightmare and risked unprofitability as a customer's API usage could surge unexpectedly, costing the startup significant money for the underlying models from providers like OpenAI, Google, and Anthropic. The new feature, currently in a preview available via a waitlist, allows a startup to set a consistent margin, for example 30%, on top of the token costs from a model provider. It automates the process by tracking the API pricing of the selected AI models, recording customer token usage, and then applying the pre-set margin to the bill. This turns a variable infrastructure cost into a direct revenue stream. This token-based billing is part of a larger push by Stripe into AI, which includes its own AI gateway to access multiple models. However, the billing feature also integrates with third-party gateways like Vercel and OpenRouter, acknowledging that many developers already use these tools. For comparison, OpenRouter charges a flat 5.5% markup on its entry-level plan for access to over 300 models. The move solidifies Stripe's strategy of embedding itself as essential financial infrastructure for AI-native businesses. More than 75% of the Forbes AI 50, including industry leaders like OpenAI and Anthropic, already use Stripe to monetize their products through various models like subscriptions and usage-based billing. The company has been actively building functionality for AI companies, a focus highlighted by co-founder Patrick Collison.
