Traditional industrial robots have considerable sales volume in the market, but some tasks are difficult for them to handle.
At present, the domestic industrial automation market is in the dilemma of “saturated stock and difficult growth”.
On the one hand, traditional fixed industrial robots have been replaced by domestic ones in core automotive welding and painting work stations, and their market share has historically surpassed that of foreign investment. On the other hand, a large number of unstructured flexible workstations on the production lines of car companies, such as loading, quality inspection, material transfer, etc., have never been covered by standardized equipment.
In 2025, China’s industrial robot sales will be about 280,000 units, and the market size will exceed 65 billion yuan. The automobile manufacturing scene will account for 32% of the total share, and the overall market share of domestic brands will climb to 55%, surpassing the “four major families” for the first time.
But the real issue isn’t sales, it’s coverage.
Traditional industrial robots are essentially “special equipment for fixed workstations” and rely on pre-programming to perform tasks. However, about 25% of the workstations on the automobile production line involve flexible requirements such as multi-station switching, dynamic environment adaptation, and cross-workstation transfers. Traditional solutions cannot be completed at all.
Take the loading and unloading and mobile quality inspection in the automobile production line as an example. Traditional solutions either require an equipment customization cycle of more than 6 months, or can only be completed manually.
The actual measured data of Phi-Bot X1 of Light Image Technology shows a huge gap: X1 can achieve 0.05mm end repeat positioning accuracy and 21.5 hours of error-free operation by relying only on body perception. By replacing the end effector, it can quickly switch between multiple work stations such as quality inspection, loading, and tightening, and the deployment cycle is compressed from months to weeks or even days.
The causes are divided into two aspects: internal and external.
The external factors are the structural changes in the automobile industry itself. The iteration of terminal models is accelerating, the personalized needs of users are increasing, and the traditional “one production line to build one car” model is gradually loosening. The demand for flexible production capacity of car companies continues to expand, and the response speed of traditional robots cannot keep up.
The real internal reason is that the product logic of traditional industrial robot manufacturers does not leave room for this kind of change. The traditional solution is to have one machine per site, with fixed installation, exclusive programming, and special tooling. Switching workstations means re-doing the entire software and hardware integration.
In the first quarter of 2026, the financing amount of the embodied intelligence track exceeded 30 billion yuan, a year-on-year increase of more than 100%. There are already 14 unicorn companies in the entire track with a valuation of over 10 billion.
The “Special Action for Humanoid Robots and Embodied Intelligence Real-Scene Training in 2026” issued by the Ministry of Industry and Information Technology and the National Assets Supervision and Administration Commission of the State Council clearly stated that “a 10,000-unit-scale implementation capability will be formed by the end of the year.” The dual support of policy and capital has verified the authenticity of the pattern level.
According to calculations in the “2026 Robot Industry Engine and Future Development Blue Book”, the penetration rate of industrial-grade embodied intelligence in automobile manufacturing scenarios will increase from less than 1% currently to 8% – 15% from 2027 to 2029. The corresponding annual replacement of traditional automation market size is about 12 billion – 21.6 billion yuan. The core replacement position is the flexible workstation that traditional industrial robots cannot cover.
The differentiated positioning of Guangxiang Technology determines that its impact is to “fill the gap” rather than “frontal force”. It targets work stations that are beyond the reach of traditional manufacturers, rather than fighting a war of attrition with Eston and ABB in mature fields such as welding and painting.
However, the speed of implementation of Phi-Bot X1 and the positive and open attitude of the car industry have sent a clear signal that traditional industrial automation manufacturers must supplement their embodied intelligence capability reserves as soon as possible.
Eston has incubated Kuzhuo Technology and released the wheeled humanoid robot “Panshi C05-L” for industrial scenarios in November 2025, clearly targeting 2026-2027 as the core iteration cycle for adapting to flexible automotive scenarios.
The impact of Guangxiang Technology on the market structure will not shake the fundamentals of traditional industrial robots in core processes such as welding and painting in the short term.
But from a longer perspective, when a mobile two-arm robot can cover the work of 3-4 traditional robot stations, and when car companies start to use “total ownership cost” instead of “single unit purchase price” to measure input and output, the deployment logic and pricing model of the entire industrial automation will be redefined.
The essence of this round of changes is not who replaces whom, but that the boundaries of automation in the automotive industry are being broadened by a new technological species, and once this boundary is released, it will never shrink back.
