[The Economic Account Behind the AI Boom: Industry Experts Break Down the True Cost of AI Computing Power] According to a report by Jinse Finance on June 6, citing CCTV Finance, with the rise of artificial intelligence applications, activities like 'raising lobsters' and 'raising horses' have sparked a nationwide craze, and the term 'Token' has become increasingly familiar to the public. In March 2026, the National Data Bureau officially designated the Chinese name for 'Token' as '词元' (Ciyuan). Also in March, the daily invocation volume of Chinese tokens surpassed 140 trillion times. The generation of each token relies on computing power, and computing power, in turn, depends on electricity. There is a claim that '60%–70% of AI computing power costs are electricity expenses.' In response, Peng Li, co-founder, CTO, and Chief Hardware Architect of Muxi Technology, stated during an interview on *Dialogue* that this claim refers to the proportion of operational maintenance costs during the operational phase, rather than the full lifecycle cost of an intelligent computing center. For a 10,000-card-level intelligent computing center with an investment of approximately 1.5 billion yuan, the proportion of electricity and maintenance costs in the total investment does not exceed 20%. However, during the operational phase, electricity is indeed the primary cost, making energy efficiency a critical factor. When discussing the criteria for selecting electricity partners, Peng Li mentioned two main considerations: first, cost; and second, 'general applicability, ease of use, and stability/reliability.'