OpenAI chairman Bret Taylor said companies will stop thinking about AI tokens within 12 months, as outside vendors take over the job of managing them.
Key Points:
- The chairman of OpenAI expects corporate anxiety over AI token costs to fade within about a year.
- He says vendors and sector-specific software firms will absorb token management for their customers.
- Taylor also questioned whether cheaper open-weight models actually cost less to run in practice.
Bret Taylor Says Token Costs Fade
Taylor, who also founded the customer service startup Sierra, made the forecast during a Monday appearance on CNBC's "Squawk Box." He blamed the current wave of complaints about tokenomics on an AI market that has not yet matured. Tokens are the units of text that models process, and providers use them to price usage and calculate bills.
Earlier this year, climbing token costs pushed finance chiefs at some of the largest technology companies to rein in AI spending and question whether the technology returned anything close to what they were paying.
Taylor expects other firms to carry that burden instead, and he pointed to Ramp's newly released spend tool as one early example. Sector-specific companies such as the legal AI startup Harvey already handle tokens on behalf of their customers, he said, and the market is moving toward paying for outcomes.
Also Read: Dow Jones Industrial Average Volatility Tests Bitcoin's Decoupling Story
Kimi K3 Tests Token Efficiency Claims
He compared the current moment to the early internet, when standing up a website cost far more than it does today. Taylor called the situation an opening for entrepreneurs to build products that spare businesses the accounting. Ramp launched its token spend dashboard on Jul. 16, reporting that token spending among its customers had climbed 20.7 times since June 2025.
Kimi K3, released last week by the Chinese startup Moonshot, drew attention across Silicon Valley by rivaling flagship systems from OpenAI and Anthropic at a lower advertised price. Taylor said the jury is still out on whether the model actually costs less to use in practice.
He downplayed the price advantage of open-weight models and argued that frontier labs hold a clear edge on token and compute efficiency.
Not every token is equal, he said, because model quality shapes how many a single task ends up consuming. Within a year, he predicted, information technology departments will run separate systems for marketing teams and for software engineers.
Taylor's Recent Public Remarks
The comments extend a run of appearances by Taylor, who told the same network on Jul. 16 that OpenAI had no update on its IPO plans while discussing enterprise budgets and an Apple lawsuit. He has also argued that connecting a chatbot to a knowledge base is now a weekend project, leaving the difficult work in guardrails, voice handling and regulated industries.
Read Next: Meteorite That Hit A Home In New Jersey Fuels AI-Driven Hunt For Life