Stripe Billing expands multiprocessor, invoicing, and hybrid pricing support

Stripe Billing, which now counts more than 300,000 users including Atlassian, Figma, Instacart, OpenAI, Anthropic, and NVIDIA, has shipped a wave of updates aimed at giving businesses more flexibility over how they monetize products—especially AI offerings. The platform was named a Leader in both the 2025 Gartner Magic Quadrant for Recurring Billing Applications and The Forrester Wave: Recurring Billing Solutions for Q1 2025.

Manage payments from any processor

Billing now lets you manage subscription payments made via off-Stripe processors across the full payments lifecycle, covering successful, failed, refunded, and canceled transactions. Previously limited to Stripe-native processing, this expansion means you can:

  • Report failed payment statuses for off-Stripe transactions and attach them to invoices
  • Schedule retries and apply dunning logic to off-Stripe payments
  • Display uploaded payment method logos and names in the customer portal for off-Stripe transactions
  • Use unified reporting and revenue recognition across Stripe and non-Stripe volume, available as an additional paid feature

More multiprocessor capabilities are planned for later this year, including Stripe Sigma advanced reporting, recording off-Stripe disputes, and allowing customers to update subscriptions in the portal using any processor.

More options for invoicing and payment terms

Billing now supports prebilling future subscription periods at any time. You can collect next month's payment today to close a deal, bill three months ahead, or prebill a specific subset of items. The feature handles arbitrary periods—like prebilling through December 2025 and then resuming normal billing—as well as partial periods.

Invoicing also gets new flexibility. Customers can pay in local currency across more than 150 countries using Adaptive Pricing, and pay over time via buy now, pay later methods such as Klarna, Affirm, and Afterpay/Clearpay on the Hosted Invoice Page. You can now accept partial payments, or unapply a payment from a paid invoice when a reconciliation needs correcting.

Pricing plans designed for hybrid AI revenue models

AI companies often need to mix flat subscriptions, usage-based fees, and credits to match their cost structures. Stripe's new pricing plans combine usage-based rates with dimensional pricing, recurring fees, and credits in a single construct.

The model is built around a consolidated plan: a customer gets a subscription, is issued recurring credits, and Stripe defines how those credits are consumed across products—all on one bill. Consumption tracking happens in real time, with dashboards via the Usage Analytics API showing usage and credit balances. Automatic top-ups can be scheduled when balances run low.

As an example of implementation speed, Lovable built its entire system—from design to launch—with just two engineers in under two weeks.

LLM proxy handles token-based pricing

Inference costs can shift 10%–40% within months, creating constant pricing pressure. If model costs drop 25% and prices don't follow, competitive advantage erodes; if costs rise, margins shrink fast.

Stripe's answer, in private preview, is an LLM proxy. Requests made to the proxy endpoint are tracked for token consumption on both query and response, and your preconfigured pricing is applied in that same call. Stripe tallies token usage, applies your markup, and issues unified invoices on your behalf—so when underlying costs change, margin maintenance doesn't require manual effort. Integrations with OpenRouter, Cloudflare, Vercel, and Helicone allow routing through third-party proxies while Stripe continues automatic token logging and event capture.