OpenAI is cutting AI prices hard—and betting that a flood of new usage will more than compensate for thinner margins. Early figures suggest the gamble may be working, though the evidence is still fresh.

The move began with an 80% reduction in the price of OpenAI’s GPT-5.6 Luna model and a 20% cut for its mid-range Terra model. TD Cowen’s analysis of OpenRouter data found Luna’s effective price fell roughly tenfold, while consumption jumped about 14-fold; Terra became about three times cheaper and usage rose about fivefold.¹

That imbalance matters. In the seven days after the cuts, TD Cowen estimated Luna revenue rose about 34% and Terra revenue about 45%—an unusual result for products whose headline prices were slashed. The report casts the shift as a version of the old Jevons paradox: making a useful resource cheaper can unlock far more demand. As it put it, “The cheaper AI gets, the more of it we use.”¹

OpenAI’s case is not simply about stimulating demand. Cheaper models let companies automate work that had previously been too costly, from document analysis and customer support to software writing and multi-step agents. The broader market is also becoming more competitive: a separate account described OpenAI and Anthropic as being in a “price war” while Chinese AI rivals gain ground.²

That puts the early revenue bump in a sharper light. Falling computing costs may allow providers to keep cutting token prices, but they will also make defending margins harder. TD Cowen’s sample covers only about two weeks, and the analysts cautioned that they need to see whether the increase holds. For now, OpenAI is testing whether scale—not scarcity—will be the winning business model.¹