US Markets
Home›US Markets›Sectors›AI providers face token cost uncertainty as agent use…
AI providers face token cost uncertainty as agent use expands
Firms are struggling to forecast future pricing because token consumption is rising much faster than unit token costs, leaving buyers unsure about what they will spend.
BBC Business highlights that many users start with free AI tools, while paid versions, such as offerings from Microsoft, Google, and Anthropic, add features for tasks like coding or billing. At the same time, third-party companies are packaging AI agent services built on large language models, which adds another layer of pricing and cost complexity for business customers.
The article explains that AI responses are broken into tokens, with prompts and outputs translated into chunks that models process. It says subtle changes in prompts, differences across models, and the use of multiple agents in agentic systems can all increase both token use and unpredictability in how much capacity a task will consume.
BBC Business also notes that while the cost of individual tokens or credits has fallen in recent years, overall token consumption has surged. Goldman Sachs analysis cited in the piece forecasts external token consumption will grow 24 times between 2026 and 2030 to 120 quadrillion tokens per month as companies shift to more AI agent usage.
The uncertainty has tangible consequences, the article says, because buyers may not know how many tokens they are burning through until bills arrive or services run out. It points to examples including Microsoft reportedly limiting some engineers' use of third-party coding tools and Uber using substantial amounts of AI resources.