From Showroom to Shop Floor: Automation Technology Moves From Demos to Real Work
A week after the China International Industry Fair closed in Shanghai, one change keeps coming up in industry conversations: the conceptual prototypes are almost gone from the robot halls. What replaced them are machines that arrive with production line data.

Image Source: 699pic.com
JAKA‘s wheeled humanoid robot spent months at a leading auto parts company’s headlight factory. Single-box cycle time dropped from 105 seconds to 28 seconds. New scenario deployment time shrank from 12 months to 57 days. Inovance showed a humanoid robot for the first time, equipped with a seven-axis bionic arm, coordinating with an arc welding robot on site to complete workpiece loading, unloading, and welding. Delta‘s embodied dual-arm system took photos and drew portraits in a chosen style, demonstrating autonomous operation from task learning to execution.
These scenes point in the same direction: automation technology is moving from executing preset programs to understanding environments and making judgments.
National data provides the backdrop. In the first three quarters, China’s industrial robot output grew 29.1 percent year-on-year, continuing the curve from 14.2 percent in 2024 and 28 percent in 2025. Export value grew 16.9 percent. Behind the output number is a change in demand structure. At the Wuhan Mengshi assembly workshop, domestic heavy-load robots handle the full process of gripping, transferring, and line feeding. At Geely‘s Yiwu welding shop, nearly 100 high-payload robots cover over 95 percent of mainstream processes, enabling multi-model switching without stopping the line.
Supply-side maturity is another pillar. Estun‘s industrial robot deliveries exceeded 10,000 units in the second quarter, setting a domestic shipment record. Topstar signed about 6,800 robot orders in the first half, with related revenue up 121.62 percent.
Siemens disclosed progress on another dimension this week. Its factory agent, built on NVIDIA’s FOX blueprint, is being integrated into the Intelligence Center X enterprise system to identify root causes and coordinate factory operations. At Siemens‘ Erlangen electronics plant, the approach has already cut calibration time in inspection processes by 40 percent, with an expected 20 percent cost saving. Agents are no longer just assist tools. They are entering the closed loop of production decisions.
Black Lake Technologies showed more than ten decision-making agents at the fair, covering quotation, engineering, scheduling, and order tracking. They share information around a single order, and downstream plans adjust automatically as conditions change. Founder Zhou Yuxiang‘s point is that industrial knowledge and production experience no longer depend solely on individual accumulation. They can be consolidated in systems and replicated quickly.
Shanghai is moving too. An action plan released this week to accelerate next-generation smart manufacturing includes four major projects and 19 key tasks aimed at deepening AI integration with manufacturing. The city has cultivated two national benchmark smart factories and 28 national excellence-level smart factories, the most among Chinese cities.
Overseas, progress is parallel. Schaeffler and Wandelbots announced a partnership this week to build Schaeffler’s own Physical AI platform on the NOVA platform, targeting up to 40 percent reduction in the lifecycle cost of automation solutions. NOVA decouples application logic from hardware, so industrial robots, mobile systems, and humanoid robots are orchestrated in the same software environment. In a completed robot-assisted packaging cell project, development time per application was cut by about two weeks.
Gaps remain. The industrial robot industry still faces challenges in high-end scenario adaptation and stability under complex conditions. Output growth is not the same as technology leadership. But from the CIIF show floor to Siemens‘ factory agents to Schaeffler’s cross-site deployment plan, automation technology is completing an identity shift: from a programmed tool to a trained, orchestrated, and trusted participant in production.

