Businesses face rising and less predictable AI bills as token use grows

In short: More companies are paying for AI based on “tokens”, and their monthly bills are rising and getting harder to predict.

What's going on

AI tools often look cheap or free to casual users, but business use can cost much more. Basic subscriptions for chatbots like Claude and ChatGPT start around $8 a month, and higher tiers can reach $100 or more. For heavier usage, many providers charge by “tokens”, which are small chunks of text the system reads and writes (like counting words and letters on a meter).

Token costs add up fast because AI charges for both what you send in and what it sends back. Output tokens usually cost more than input tokens, since generating text needs more computing power. More advanced models also cost more per token because they do deeper processing.

A big driver of new spending is “AI agents”, which are AI systems that can take a task and then use tools and look up information on their own (like sending an assistant off to do research, except it may keep going until it feels confident). Sara Hooker, CEO of Adaption Labs, told the Financial Times that agent workflows can become “long horizon”, which means far more tokens and bigger bills.

Data shared in the report shows how spending is climbing. Ramp said AI spending in the 18 months to June 2026 grew nearly sixfold. Median AI spend per employee per month rose from $2.86 in January 2024 to $11.95 in July 2026, and the top 1 per cent of spenders rose from $755 to $7,400.

What to watch

Companies are starting to control costs by choosing simpler models for simple tasks, shopping around providers, and using “open-weight” models that they can run on their own servers. Another key issue is whether upcoming investor pressure, including expected IPOs for major AI developers, leads to higher token prices.

Source: Financial Times