“Free AI is ending. Tokenomics is beginning.”
That’s the message from Citrini Research, the firm that sent the stock market into a tailspin with a report on the potentially dire economic impacts of AI earlier this year.
As AI power users focus their efforts on “tokenmaxxing,” or maximizing their use of AI tools to boost productivity and stay ahead on internal leaderboards, Citrini says the associated costs will reshape the market’s AI trade.
“Tokenomics” refers to the economics of AI usage as measured by tokens, the blocks of text or code being constantly processed by large language models. While tokenmaxxing has already become one of the hottest workplace trends of 2026, Citrini’s says the economics are unfavorable for companies.
The tech industry has been rocked by a surge in demand for tokens as companies push their employees to continuously scale their AI usage. The bill is already coming due, and in Citrini’s view, it marks a sea change in how companies will approach AI.
“The deepest pockets in the world — hyperscaler cash flow, venture capital, sovereign wealth, public credit, private credit, public equity — are footing most of the bill,” the author noted. “Eventually, customers have to start picking up the tab.”
So, what happens when they do?
Citrini maintains that the AI boom will continue to shift toward a more spending-conscious phase driven primarily by efficiency, as companies seek out ways to lower AI costs.
This could be accomplished by local inference, such as running AI models on a device like a PC, something that Nvidia has recently doubled down on helping users do through its new PC chip. This could create a new class of AI winners, Citrini said
“AI devices, running local models, will eventually be a thing,” it added. “This trend will not wipe out cloud computing or large, remote models — it’s both-and instead of either-or.”
Citrini sees a shift toward “edge AI” as the next phase of the boom. Importantly for investors, the new dynamic could create a range of opportunities.
“The first AI trade was solely about centralized compute. As that becomes more adequately priced in, we see more asymmetry available in distributed inference, the surrounding hardware, and the software required to secure and orchestrate it all.”
Edge AI refers to a computational model that runs AI either on or near a device without sending every task to a cloud or a data center.
In this case, edge AI is a broad ecosystem that distributes computer power across many different devices that can include not just powerful PCs, but also laptops and smartphones.
