Stripe wants to help your business claim back all those AI costs
New Stripe tool tackles variations in token usage by tracking pricing per tokenCompanies can apply a percentage markup that updates as token prices changeIt removes the stress and complexity of managing multi-model AI platforms
Stripe has launched a new billing feature designed to help startups and SMBs keep tabs on their AI bills at the token level.
At the moment, customers must pay model providers like OpenAI and Google a cost per token for the services they offer their customers, which means that high user activity results in high token bills.
Stripe says without proper controls, startups can risk operating at a loss. Agentic AI also changes token usage patterns slightly, making billing more unpredictable.
Stripe lets AI startups charge at the token level
With Stripe's new feature, startups can select AI models, track live API pricing, automatically record customer token usage and then apply their own custom market percentage on top of that to increase profit, making billing far more accurate and eliminating the chances of higher-than-anticipated bills payable to their model providers.
The tool works by displaying updated token pricing across major providers in one dashboard and notifying users when pricing changes. "Enter your desired markup... Click Submit and we’ll configure the required Usage-Based Billing resources – prices, meters and rate configuration," the company wrote in a support page. "You only need to set your margin percentage to start using Usage-Based Billing."
The payment processing platform hopes that the token-level transparency will help remove the pricing complexity that AI startups currently face, but having all of this information and control within one dashboard will also help for multi-model AI platforms.
And because users can set a percentage markup, this gets automatically adjusted as token pricing changes from the model providers, so they don't need to change anything to keep making a profit.
Stripe is currently looking for "developers willing to test the functionality and share their feedback," thus the feature is currently in private preview and is not yet generally available.
