OpenAI chairman Bret Taylor said Monday that the artificial intelligence market will shift from charging for token usage to paying for outcomes [1].
This transition represents a fundamental change in how businesses calculate the return on investment for AI spending. As the industry matures, the focus is moving from the technical cost of processing data to the actual economic value delivered by the software.
During an interview on CNBC’s "Squawk Box" Monday, Taylor said the emerging challenges of AI tokenomics are evolving [2]. He said that companies are currently preoccupied with token efficiency — the amount of compute resources required to generate a response — but predicted that this concern will eventually fade [3].
"There will be a shift to paying for outcomes instead of AI tokens," Taylor said [4].
Taylor is also the co-founder of Sierra, an AI startup that has seen significant financial growth. According to different reports, the company initially raised $175 million at a $4.5 billion valuation [5]. More recent data indicates a larger funding round of $950 million, which valued the company at over $15 billion [6].
By moving toward outcome-based pricing, AI providers would essentially guarantee a specific result or efficiency gain rather than charging for the volume of text or data processed. Taylor said that this evolution is necessary for AI to be integrated deeply into corporate workflows, where the cost of a "token" is less important than the successful completion of a business task [2].
This shift would align the incentives of AI developers with the success of their clients. Instead of optimizing for the fewest tokens used, developers would optimize for the highest success rate of the intended outcome [3].
“"There will be a shift to paying for outcomes instead of AI tokens."”
The move toward outcome-based pricing signals a transition from the 'experimental' phase of generative AI to a 'utility' phase. By decoupling cost from token volume, AI companies are betting that their tools can provide predictable, quantifiable business value, which reduces the financial risk for enterprise adopters and forces AI labs to prove the actual efficacy of their models.
