Token economy emerges as core metric for AI computing power

The emergence of the ‘Token economy’ marks a significant shift in the artificial intelligence industry, where the ‘Token’ is becoming the core unit of measurement and pricing for Model-as-a-Service (MaaS) sectors. As AI evolves, the industry is moving from measuring hardware capacity to measuring utility through standardized data throughput.

In China, the demand for tokens is projected to grow exponentially, with IDC predicting enterprise-level MaaS token calls will reach 400 trillion by 2026. This demand is driving the development of ‘Token factories’ in cities such as Beijing, Tianjin, and Wuxi, which encapsulate complex computing into elastic, pay-per-use services. This model allows enterprises to access intelligence without the sunk costs of building private computing clusters.

However, the rapid expansion faces structural risks. There is a notable mismatch between the high market valuations of AI infrastructure companies and their actual profitability. Reports indicate that approximately 95% of organizations have yet to see measurable financial returns from generative AI investments. This creates a risk where capital expenditure on hardware and maintenance outpaces the revenue generated from downstream applications, potentially leading to a market correction similar to the telecommunications bubble of the early 2000s.