WAIC 2024: China's AI Ecosystem Comes of Age with a New 'Three-Piece Suit'
Each year, the World Artificial Intelligence Conference (WAIC) serves as a snapshot of the AI industry's evolution. Two years ago, the spotlight was on large models and their scale. Last year, it shifted to agents and humanoid robots. This year, however, marks a significant maturation, with the focus finally settling on an integrated trinity: models, computing power, and products, all working in concert.
A cornerstone of this new ecosystem is foundational infrastructure, exemplified by Huawei's Atlas 950 SuperPoD. The conversation around domestic computing power has shifted from the performance of a single chip to the stability and synergy of thousands of cards working as one cohesive unit. For China's trillion-parameter models, this system-level approach answers a critical question: where will these massive models run? It provides the stable, large-scale computational foundation necessary for their continued growth and iteration.
In parallel, domestic AI models are reaching new heights. The release of Moonshot AI's Kimi K3, a model with 2.8 trillion parameters, and Zhipu AI's GLM-5.2, demonstrates performance competitive with leading closed-source models. More importantly, their capabilities have expanded beyond chat and content generation to complex reasoning, code, and tool use. This signifies a shift where "openness" now refers to a new industrial division of labor: model companies provide foundational capabilities, cloud platforms handle deployment, and hardware manufacturers adapt them for specific use cases. The foundational pieces of AI—computing and models—are being put in place simultaneously.
This integrated approach is most tangible in the new wave of consumer devices. The competition in AI smartphones has entered a new phase, moving beyond simple AI-powered features. Devices like the StepX Neo and Nubia NaviX Ultra are pioneering a deep integration of AI with the operating system and hardware. The goal is to allow AI agents to understand user intent and directly execute tasks by calling on system-level functions, fundamentally changing how we interact with our phones.
The trend extends to other personal devices. Qwen, in partnership with Bose, introduced AI-powered smart earbuds designed for tasks like simultaneous translation and meeting summaries. This collaboration highlights a mature product strategy: embedding advanced AI into proven, wearable form factors like headphones and glasses. Instead of inventing new devices, companies are giving consumers new reasons to buy familiar ones by making them smarter and more capable, providing a natural, screen-free interface for AI interaction.
Embodied intelligence also captured significant attention, moving from the lab to the showroom floor. While humanoid robots were present, it was Unitree's GD01 manned mecha that drew the biggest crowds. This machine shifts the paradigm from robots acting autonomously to robots carrying and being controlled by humans. It serves as a powerful, tangible demonstration of how abstract concepts like control algorithms and mechanical engineering can come together, suggesting that the future of embodied AI may take many forms beyond the humanoid, tailored for specific tasks like engineering and heavy transport.
In previous years, advancements in computing, models, and products were often discussed in isolation. This year at WAIC, they have converged into a complete industrial portfolio. A solid foundation of computing power supports increasingly capable models, which are in turn being integrated into a diverse range of consumer products. The next stage of China's AI development will likely be driven not by a single breakthrough, but by the collective force of this integrated, three-part ecosystem.
