Two in three top AI companies now charge by usage, and Stripe says pricing pages haven't caught up

Two in three of Forbes' top AI companies now charge by usage instead of a flat subscription, up from under half a year ago, according to payments data Stripe disclosed at SaaStr's AI Summit this week — a shift that changes what belongs on a pricing page and how a marketing team explains its own product.

The numbers come from Maia Josebachvili, Stripe's chief revenue officer of AI, who walked through the company's own transaction data across its AI customer base, as first reported by SaaStr. Because Stripe sits underneath billing for a large share of today's fastest-growing AI vendors, it can see how those companies actually charge customers, as distinct from what their marketing pages say a plan costs. The company has separately built billing infrastructure for exactly this shift, letting vendors meter tokens, API calls and agent tasks and apply a markup on top of model costs, a move PYMNTS reported on when Stripe first shipped the tooling.

The flat fee is losing ground

For two decades, SaaS pricing pages sold a number of seats at a flat monthly rate. Stripe's data says that model is breaking down for AI products specifically because the cost of serving each customer now varies as much as the value they get. A support agent that runs a batch of automations overnight costs more to serve than someone who logs in once a week, and a flat seat price can't reflect that gap.

Most of the AI50 companies moving to usage pricing aren't dropping subscriptions outright. Replit is the case study Stripe pointed to: after building a decade of business on a flat developer-tools subscription, it added usage credits on top once agentic coding took off, and it's now targeting a $1B run rate on the hybrid model. The subscription still gives customers a predictable floor; the credits let the vendor capture more revenue as usage climbs, instead of leaving that value on the table at a fixed price.

This isn't only a story about AI-native startups. Any established SaaS company that has bolted an AI feature onto an existing seat-based plan is facing the same mismatch Stripe is describing, just one step removed: the AI feature has variable inference costs, but it usually ships as a flat add-on fee or a "included in Pro" line, which means the pricing page is already understating what heavy users cost to serve.

Pricing pages need a passport

The same talk included a second data point that's easy to miss if you're only tracking the pricing-model story: localization moves revenue on its own, independent of what the pricing model looks like. Stripe's data shows localized pricing drives 18% higher cross-border revenue, and adding a single local payment method adds another 7%-plus. Top AI companies now pull 48% of revenue from outside their home market, and reach 42 countries in their first year and 120 by year three — far faster than the old sequential playbook of winning the home market before opening a market in London or Dublin.

A pricing page that only accepts one currency and one card network is quietly taxing every customer outside your home market.

That's a marketing-owned problem as much as a finance one. The pricing page, the plan copy and the checkout flow are usually built and maintained by growth or marketing teams, and most of them were never designed with a second currency in mind.

What a software marketer should do differently

None of this requires ripping up an existing pricing model overnight, but it does mean the pricing page copy and the sales narrative need to catch up to how billing actually works underneath. Three changes matter most:

• Rewrite plan descriptions in the customer's own units — tokens or usage credits for developers, seats for enterprise buyers who still expect a per-seat number, and say plainly when a plan mixes both.
• Put a usage meter in the product, not just on the invoice. Stripe's own framing is that surprise bills drive churn, so real-time visibility into consumption has to ship before the pricing model does.
• Before writing "now available in Europe" copy, check whether the pricing page can actually charge in euros and accept a local card network — the revenue lift comes from the payment method, not the press release.

The risk in ignoring this isn't abstract. A pricing page that still reads like a 2020 SaaS seat calculator, next to a product that Stripe (and the finance team) now bills by consumption, creates a mismatch that shows up first as sales friction and then as churn when the invoice doesn't match what the page promised. Marketers don't set the billing logic, but they do set the story a buyer reads before they see the bill — and right now, for a fast-growing share of the market, that story is out of date.