OpenAI, the artificial intelligence research and deployment company behind ChatGPT, has begun letting some customers pay based on the results its models deliver rather than a flat subscription fee.
The arrangement ties charges to measurable outcomes, such as successful task completion, rather than access alone.
It marks a departure from the flat-rate licensing that has dominated software pricing for decades.
Slow pivot
The move follows earlier shifts this year toward metered, per-token pricing for OpenAI's application programming interface (API), the system that lets other businesses build products using its models.
The company has also trialled usage-based pricing for enterprise features.
Discussions at chief financial officer level have reportedly focused on licensing models that pay out specifically when artificial intelligence generates measurable business value.
Who bears the risk
Outcome-based billing lowers upfront costs for customers and links spending directly to results.
For OpenAI, the arrangement is riskier: revenue now depends on model accuracy, processing speed and how smoothly its systems integrate with a customer's operations.
If a model underperforms, OpenAI stands to earn less, rather than the customer absorbing the cost of a subscription regardless of results.
Compute costs add pressure
The pricing shift comes as OpenAI continues to publicise substantial commitments to computing power and infrastructure.
Predictable, performance-linked revenue is becoming more valuable to the company than fixed-fee contracts as those costs rise.
A limited test for now
Executives have described the rollout as a selective commercial trial rather than a full pricing overhaul.
The company could expand outcome-linked billing more broadly if the initial partnerships prove the economics and risk-sharing involved.
