At WAIC 2026, among 130 generative AI exhibitors, 83 identified AI agents or intelligent applications as their primary label, while only 18 remained focused on foundation large language models. As the industry’s focus moves from “how smart is the model?” to “how much real value can it create?”, the basic unit of AI competition is shifting from a single model to a complete system—making the synchronized evolution of computing power, data, algorithms, talent, and organizational structure the...

A set of data from WAIC 2026 clearly reflects this shift in the AI industry: among 130 exhibitors in the large language model and generative AI track, 83 featured AI agents or intelligent applications as their primary label, leaving only 18 foundation model companies. There were fewer model launches and more agent demonstrations, and the industry conversation moved from “how smart the model is” to “how much real value it can create.”

ehind this change is a judgment that is rapidly becoming consensus: the basic unit of AI competition is shifting from “a model” to “a system.” In the past, discussions about computing power often focused on the performance of individual AI chips. But large model training and inference were never a single-card race; interconnect, memory sharing, and task scheduling determine whether a large-scale cluster can deliver effective computing power. The center of gravity for AI compute demand is also moving from training to inference. In this competition, computing