Galaxy Securities said surging demand for AI inference computing is accelerating the commercialization of a so-called “token factory” model, under which pricing shifts from selling compute resources and time to selling token output and unit costs, according to Jiemian News. The brokerage said the model more tightly links upstream compute infrastructure with downstream large-model inference demand, potentially allowing infrastructure providers to participate more in value distribution and strengthening their pricing power.
Galaxy Securities added that compute leasing companies and large-model developers are exploring token-factory partnerships, evolving from a “sell resources” approach to a “sell output” approach. It said compute lessors may provide MaaS services and adopt token revenue-sharing models, which could improve profit margins and support higher valuation benchmarks, and it recommended watching leading compute leasing firms.