AI companies already rely on tokens as a basic unit for measuring how much a customer uses a model. That same unit is now drawing interest beyond billing, as economists look for practical ways to estimate how widely artificial intelligence is being used across the economy.
The idea is simple: if electricity has the kilowatt-hour, AI may end up with the token. In that comparison, tokens function like a metered measure of computing work, giving companies a way to charge for AI services and giving analysts a way to compare usage over time.
The discussion reflects a broader vision promoted by major AI firms, including OpenAI, that advanced AI could be consumed like a utility. In that model, businesses and individuals would not just buy software outright; they would pay for intelligence as they use it, with tokens serving as the unit that records consumption.
If that approach takes hold, token counts could become more than a technical detail buried in developer dashboards. They could emerge as a useful signal for understanding demand, pricing, and the pace at which generative AI moves from a niche tool into everyday economic activity.
