Chances are, when you’ve read about artificial intelligence (AI), the article included a reference to ‘tokens.’  As more companies (and consumers) begin using AI capabilities, they’ll begin to be charged for the services in the form of these token things.

But what are they, exactly?  

A token is a unit of measurement in the AI industry, basically standing for an amount of work.  Whenever you ask an AI system like Claude or ChatGPT a question, the model breaks the question down into smaller chunks of what might be called ‘deliberation’ or ‘analysis,’ and each of these are tokens.  All of your AI interactions are measured by the tokens that are consumed—and users (companies, primarily) are billed according to their overall token use.  You can think of the AI work as paying a human contractor by 15-minute increments of his/her time.

Tokens, in other words, are now functioning like utility meters or freelance expenses that pile up weekly or monthly on the cost side of the balance sheet.  This is a category of expense that did not exist on any company’s books three years ago, and companies are beginning to set token budgets to control the AI costs their employees may be running up.  Microsoft famously set controls in place when one employee racked up $28,000 in token costs in a single week.

Just like hiring contractors, not all tokens have equal value.  A token from the most advanced frontier AI models will deliver better judgment than a token from a less advanced model—and therefore their tokens cost more.  However, most of the costs of building these advanced models—which run in the billions of dollars—are generally in the past by the time they are rolled out commercially, and some observers see a time when the models are functioning roughly equally to each other.  That could conceivably set off a token price war, which would be terrific for companies, but terrible for the AI companies themselves.  By some estimates, the AI providers of agents need to generate $2 trillion in revenue per year to generate the returns needed to justify their infrastructure and development expenses totaling $5 trillion.

What does this mean for you?  The likely reality is that the world collectively will be increasing their AI usage over time, and the service will morph from free to (at first) relatively cheap tokens for consumers, increasingly expensive ones for corporations.  Be prepared for a new expense in your budget.

This article was written by an independent writer for Brewster Financial Planning LLC and is not intended as individualized legal or investment advice.