Overview
Held in Shanghai from October 12–16, the 26th China International Industry Fair made industrial AI and embodied intelligence its headline themes. CAE academician Zhou Ji argued that deep AI-advanced-manufacturing fusion is pushing manufacturing into a new digital-intelligent phase, with systems evolving from “program executors” to “autonomous intelligent agents.”
On the floor, compute chips and robot bodies raced side by side: Dongfang Suanxin showed its DF1000 chip and 128-card cluster; Axera launched the AX8860 (256 TOPS, native FP4); Tianshu Zhixin unveiled a 144-GPU single-rack node; Leju, KUKA and others demonstrated humanoids and intelligent platforms across scenarios.
Background & Interpretation
1. Technical Background and Evolution
Smart manufacturing spans digital, “Internet+”, and next-generation paradigms. A decade of digital/networked groundwork now enables AI-driven upgrade to “Smart Manufacturing 2.0,” whose core is autonomy—perceiving, deciding, executing and learning, optimizing continuously.
2. Core Drivers and Mechanisms
A layered path is emerging: upper-layer industrial Agents attack production decisions and close factory business loops; lower-layer edge chips and dev platforms tackle compute scarcity and AI-talent gaps. Zhou proposes two phases—during the 15th FYP parallel digital-network-intelligent integration targeting broad adoption by 2030; during the 16th FYP deepen intelligence toward baseline next-gen manufacturing by 2035.
3. Market Structure and Industry Impact
Embodied-model iteration is fast while the compute gap persists. Over half of China’s top embodied firms reportedly adopted a leading edge-robotics platform whose chips have shipped 8M+ units with 50%+ coverage in the field. Schneider executives note energy optimization and predictive maintenance are already the most mature AI commercializations. Short-term hype will clear; mid-long term, quantified value delivery decides survivors.
Implications & Outlook
Near-to-medium term, industrial AI shifts from “pilot trap” to value proof, with domestic chip substitution and embodied-robot scale deploying in parallel. Watch edge-compute cost curves, industrial-data-silo resolution, industrial-Agent ROI, and domestic-chip penetration in mass production.
Takeaways for Industry Participants
Industrial firms should import AI against business loops and quantified ROI, starting with mature scenes like energy optimization and predictive maintenance. Chip/platform vendors must confront the compute bottleneck with cost-effective edge solutions. Integrators should build cross-platform data connectivity to close the last mile of industrial-AI deployment.
This column compiles industry information and shares technical perspectives; it does not constitute investment advice.