Currently, artificial intelligence (AI) has become a core focus of strategic competition among major nations. China's AI sector is undergoing profound structural transformations. Growing evidence suggests that the AI industry may be at a critical transitional window-evolving from a fluid state characterized by technological diversity toward a transitional phase where the emergence of a dominant design is becoming apparent. However, existing literature lacks a clear definition and examination of this phase. Such ambiguity constitutes a primary “puzzle” hindering the effective advancement of the AI industry and the precise formulation of relevant policies.

The Abernathy-Utterback(A-U) model, a classical theoretical framework for analyzing industrial innovation dynamics, divides the evolution of innovation into three stages: the fluid phase, the transitional phase, and the specific phase. It thus provides a powerful theoretical tool for understanding industrial transition dynamics. Based on this model, this study systematically constructs a theoretical analytical framework by examining the multidimensional characteristics of the industrial innovation process, clarifying the internal mechanism behind the evolution of the innovation process and the formation of a dominant design, and revealing the dynamic adaptation principles that influence strategy design.

Using this theoretical framework, this study first systematically examines the development stage of China’s AI industry. It finds that the industry is transitioning from the fluid phase to the transitional phase-a conclusion highly consistent with the evolutionary patterns of industrial innovation described in the A-U model. From a technological perspective, the industry exhibits accelerated convergence of technical pathways, a systematic shift of innovation focus toward performance optimization, and the crossing of critical thresholds in technological maturity. In terms of industrial structure, embryonic forms of a dominant design are emerging, process innovation is becoming the core of value creation, and the division of labor within the industry is being profoundly reconfigured. Regarding the competitive landscape, a notable coexistence of open-source and closed-source models is observed, characterized by complementarity and symbiosis.

Secondly, focusing on historical cases where dominant designs were successfully established during similar transitional windows-such as the Wintel alliance, the Android system, and the European GSM standard-this study summarizes general winning principles that can inform the current development of AI in China. These include: technological feasibility as a prerequisite for the formation of a dominant design; the critical importance of dynamically balancing market acceptance and production economics for its widespread adoption; the shift in competitive paradigm from zero-sum game to ecological collaboration as a hallmark of the transitional phase; and the pivotal role of rule-setting power in transforming tri-dimensional coupling outcomes into enduring competitive advantages.

Finally, addressing core challenges such as high computing costs, lagging standardization, and immature ecosystem governance mechanisms in the development of China's AI industry, this study proposes an adaptive strategy framework encompassing technological breakthroughs, ecosystem development, and rule-making engagement. Specifically, in technological advancement, efforts should focus on overcoming computational bottlenecks and pursuing forward-looking development of key technologies. In ecosystem building, the emphasis lies on creating an open yet controllable industrial ecosystem. Regarding rule-making, it is essential to strengthen standard setting and enhance international governance cooperation. These recommendations aim to overcome critical bottlenecks during the transition phase, foster leapfrog development of China’s AI industry, and build an independent, controllable, and globally influential AI ecosystem.