A recent online seminar titled "Token Factory: A Deep Dive" was co-hosted by China Machine Press and Lenovo, focusing on the theme "Solving Four Key Challenges in AI Implementation and Understanding the New Trends of the Token Economy." During the roundtable discussion on "How Token Factories Reshape the Wealth Creation Paradigm of AI," industry experts including Huang Shan, Technical Strategy Director of Lenovo's China Infrastructure Business Group, Cao Hongwei, a technical expert at 01.AI, and Zhan Bingqiang, founder of AIGCLink, shared their insights on the topic.
A key consensus emerged from the seminar regarding a tiered delivery model that classifies tokens into consumer-grade, professional-grade, and frontier-grade categories. This corresponds to a three-level customer framework comprising L1 computing power services, L2 token services, and L3 scenario-based capability packages. This structure effectively addresses the supply-side question of "who is the production for" while simultaneously answering the demand-side inquiry of "how to purchase," creating a critical alignment point between the two perspectives.
This consensus reinforces the central thesis of the book Token Factory: advanced token factories determine the quality of advanced AI intelligence. The competitive advantages on the supply side must be continuously amplified through a flywheel effect, rather than relying on a one-time accumulation of hardware resources. The book distills this concept into a "value accounting" methodology: first, measure by attributing every token consumption to specific departments, projects, and business outcomes; second, settle by transforming AI departments from "cost centers" into "value centers"; third, reuse to ensure token investments generate compound returns; and finally, upgrade to mitigate depreciation caused by the "technology half-life."
Beyond solving the cost traceability challenge, the discussion emphasized the need to drive deeper business integration. Token consumption should expand beyond peripheral applications such as customer service interactions and document retrieval, moving into core business processes including production control, dynamic supply chain optimization, and real-time risk management. This progression enables a shift from "high-frequency, low-value" use cases to "high-frequency, high-value" applications.
The book Token Factory provides a clear timeline for industry transformation: three years to establish the market structure. The first year focuses on completing a pilot demonstration loop, the second year on platform expansion, and the third year on achieving ecosystem supply. Lenovo's Token Factory solution upgrades computing infrastructure into a "token production system," leveraging the system-level coordination capabilities of the Wanquan Heterogeneous Computing Platform V5.0. This approach aims to reduce the hardware cost per million high-quality tokens to below one yuan, while the Token Hub enables precise metering and transparent settlement, transforming the token economy from theoretical concept into industrial reality.