Tokens inferenced is a usage measure for artificial intelligence services: the number of tokens that models process when responding to requests, a stage known as inference. A token is the unit of text that a large language model reads and writes, roughly a word fragment, and counts normally cover both the input supplied and the output generated. Some cloud providers and model developers report the figure by day or by month. It has nothing to do with crypto tokens.
Investors follow the measure as an indicator of real demand for AI, as distinct from spending on training models. Rising token volume implies heavier use of a hyperscaler's capacity, of each graphics processing unit and of AI data center power, and most AI services are priced per token. The measure has limits: definitions vary between companies, newer reasoning models consume many more tokens per task, and falling prices per token mean that volume growth does not translate directly into revenue.