On Monday, Stripe unveiled a preview of a new capability aimed at assisting AI startups (and other businesses) in addressing the challenge of transmitting the foundational costs of AI model utilization to their clients.

Nevertheless, Stripe’s functionality extends beyond merely passing on token costs. It enables startups to impose a markup percentage on token utilization. For example, a company could automatically add a 30% charge above the token costs that the startup pays to the model provider.

According to Stripe, “Imagine you’re developing an AI application: you require a steady 30% margin on raw LLM token expenses across various providers. Billing streamlines this process.”

This billing feature permits the startup to select the AI models they employ. It monitors the API costs of those models and subsequently logs the customers’ token usage, applying the profit-margin markup effortlessly.

As previously reported, AI startups utilize various methods for monetizing their offerings. Numerous ones adopt tiered monthly subscription models with usage-rate limits; once those thresholds are reached, subscribers may incur additional charges for exceeding the cap.

For example, Cursor modified the pricing on some tiers last year from unlimited access to rate-limited access, adding fees for extra usage.

In the absence of a usage threshold, users could incur substantial bills with the model providers, potentially pushing the startup into deficit. This scenario is particularly pressing for agentic startups. The more their clients engage with their agents, the more tokens they deplete from the foundational model provider, whether that’s OpenAI, Google Gemini, Anthropic, or others—rendering pricing and business model choices exceptionally pivotal.

Stripe has also launched its own AI gateway, a tool that provides users with access to various models, allowing them to select the most suitable one for the task. However, the billing tool is also compatible with third-party gateways that are already well-known, such as those from Vercel and OpenRouter, as mentioned in a tweet by a Stripe product manager.

Naturally, there are other startups providing AI model cost management functionalities through their own gateways. For instance, OpenRouter, which offers entry to over 300 models, applies a flat 5.5% markup on token fees for its initial tier, and it also includes budget controls.

According to Stripe’s product manager on X, Stripe is not currently imposing its own markup on the gateway. This feature is still in waitlist mode. Regardless, if Stripe can assist startups in effectively transforming the tracking and billing of this expense into a profit-generating opportunity, it might prove to be transformative. Stripe did not immediately respond to a request for clarification on the general availability of this feature.