The AI toy sector experiences a qualitative change: limitations of model 'shells' are expected to be resolved in the next year or two, accelerating the development of intelligent agents

What happened

Honor (荣耀) this week unveiled two new robots — “闪电” and “元气仔” — and announced they will take part in the April 19 Beijing Yizhuang Marathon, a stunt meant to demonstrate advances in embodied AI and robotics engineering. It has been reported that Honor’s global chief marketing officer Guan Haitao (关海涛) made the announcement, and that Wang Bang (王班), president of sales and service at Honor Terminal Co., released imagery showing “闪电” in a red, sports-styled livery and the more humanoid, silver “元气仔” with finer joint detailing. Reportedly this is the first time a major consumer-device maker has presented robots as marathon challengers.

Why it matters

The move is being framed as evidence of a broader, qualitative shift in the “AI toy” and consumer-robot sector: hardware — the physical “shells” that house models and agents — has long been the bottleneck. It has been reported that industry insiders expect many of those limitations to be resolved within the next year or two, which would accelerate the migration from cloud-hosted chatbots to truly embodied intelligent agents capable of real-time perception and autonomous decision-making. What does this mean for Western observers? Simple: software progress alone isn’t enough without reliable, responsive hardware.

Geopolitical and market context

This progress comes as Chinese tech firms pour resources into domestic chip, sensor and actuation stacks amid strained U.S.–China tech relations and export controls that have reshaped supply chains. Observers say the push for self-reliant hardware integration is both a commercial imperative and a strategic response to those pressures. Whether the marathon is publicity or proof-of-concept, Honor’s public experiment signals that consumer robotics is moving from novelty toward practical demonstrations of embodied AI — and faster hardware fixes would make intelligent agents far more commonplace.