Zhou Hongyi: Tokens will never be like mobile data with unlimited monthly plans
Recently, Zhou Hongyi, the founder of 360 Group, shared his insights on the recent trend of large models and the operating model of the popular "lobster craze" (OpenClaw intelligent agent).
Zhou Hongyi stated that Tokens (the basic units of model computation) will never achieve "unlimited monthly usage" like mobile data. The essence of Tokens determines that they cannot be supplied at low cost and without restrictions like communication data.
Zhou Hongyi pointed out that each inference and generation of a large model comes with high computational costs, including computing power consumption, electricity costs, and infrastructure investment, which fundamentally differs from the pricing model of mobile data based on communication infrastructure;
Specifically, even though technological advancements and economies of scale may lead to a continuous decline in the unit cost of Tokens, the characteristics of Tokens determine that business models similar to "unlimited monthly usage" of mobile data are not applicable here.
He further explained that the future pricing model for Tokens will likely resemble water and electricity billing, charging fees based on actual usage; that is, the more you use, the higher the fees you pay;
This indicates that for large models to achieve widespread adoption, the key lies in continuously improving computational efficiency and constantly reducing computing power costs, rather than relying on launching "unlimited plans" to attract users and promote growth.
Overall, Zhou Hongyi's views have also prompted the industry to reconsider the commercialization path of AI. In the context of persistently high computing costs, how to balance user experience and enterprise costs has become one of the key challenges for the application of large models.