A new front is opening in the artificial intelligence boom, and it centers on something most users never see: the "token."
Tokens are the basic units of text that AI systems read and generate — the currency in which large language models are measured, priced, and rationed. As demand for AI surges, the cost and supply of tokens is becoming a pinch point, and a brief from Computerworld reports that companies are now racing to solve what it calls "the AI token problem."
The pressure is also reshaping who wants to be in the token business. According to Light Reading, the networking vendor Mavenir is urging telecom operators to sell AI tokens themselves rather than lean on outside providers such as OpenAI. The pitch reframes telcos — long seen as pipes that merely carry data — as potential sellers of AI capacity to their own customers.
Taken together, the two reports point to the same underlying shift. Tokens are no longer just a technical detail buried inside a chatbot; they are emerging as a resource to be supplied, priced, and competed over. That has set off a scramble among technology companies to ease the bottleneck, and a parallel scramble among other industries to claim a slice of the market rather than depend on a handful of dominant AI labs.
For businesses building AI features, the stakes are practical: the price and availability of tokens shape what is affordable to run at scale. For incumbents like telecom carriers, Mavenir's argument suggests an opportunity to move up the value chain instead of ceding it.
Why it matters: how the token squeeze gets resolved — and who ends up selling the tokens — will help decide whether AI's economics stay concentrated among a few providers or open up to many more players.