Shanghai-based embodied intelligence robot company 3S Roboticshas officially announced the completion of a Series B financing round worth hundreds of millions of yuan. The round was led by Shanghai Semiconductor Industry Investment Fund and Jinqiao Fund, with continued participation from 01VC, Xinding Capital, Zhongguan Dinghua, and existing shareholder Weiguang Venture Capital. The proceeds will be primarily allocated to iterating and upgrading the company's welding-specific embodied AI brain and autonomous cerebellum motion control system, expanding the capacity of its self-owned smart manufacturing factory, growing its R&D and market service teams, and accelerating the large-scale deployment of its products across multiple industries throughout China.
Founded in 2020, 3S Robotics obtained National High-Tech Enterprise certification in 2023 and was named a Shanghai Specialized and Innovative SME and Shanghai Intelligent Robot Benchmark Enterprise in 2024. The completion of this Series B round marks the company's entry into a new phase of scaled, high-speed growth built on its three-tier, fully self-developed technology stack: an autonomous "brain," an autonomous "cerebellum," and self-developed hardware production. With mass-producible, self-evolving industrial-grade intelligent robots, the company is staking out a position in China's industrial embodied intelligence race, using welding as its entry beachhead.
Why welding? A trillion-yuan market colliding with structural labor shortage
According to founder Wang Dezhao, the company originally focused on general industrial robots, but concluded that the future of robotics lies in deep vertical-scene intelligence rather than generic platforms. Among all industrial processes, welding stands out as both ubiquitous and acutely underserved.
China currently has approximately 10 million registered welders, with an average age above 45. Young people are increasingly unwilling to enter a profession defined by hazardous conditions, high physical intensity, and high technical barriers. As a result, the shortage of welding engineers reached the millions by 2025 and continues to widen by roughly 100,000 retirees per year. At the same time, manufacturers are demanding ever-higher welding quality, efficiency, and consistency. Industry forecasts put the intelligent welding robot market at over one trillion yuan over the next decade.
Traditional welding automation—relying on pre-programming, drag teaching, or manual intervention—cannot adjust processes dynamically in response to workpiece variability and is largely restricted to standardized scenarios. It is poorly matched to the non-standard, high-mix, low-volume work that dominates China's steel structure, construction machinery, shipbuilding, rail transit, and automotive components sectors.
A self-developed "brain + cerebellum + body" stack
3S Robotics is one of the few Chinese companies to integrate AI algorithms, robot bodies, and welding process know-how in-house (a position it strengthened with the 2022 acquisition of Hagong Hyundai Robotics). Its architecture mirrors that of an embodied intelligence:
- The "Brain" is a welding-specific embodied intelligence model trained on real industrial conditions. Built on a multimodal large-model architecture, it is trained on tens of millions of real production-line welding data points, a comprehensive material-process knowledge base, and datasets of complex, irregular weld seams. The model can recognize workpiece geometry, material, and texture; reason autonomously about where and how to weld; and dynamically adjust current, voltage, travel speed, and weave parameters in response to plate thickness, groove geometry, and environmental conditions. A closed-loop self-learning pipeline continuously refines the process logic as new production data accumulates.
- The "Cerebellum" is a self-developed real-time motion control system, built on years of accumulated experience at subsidiary Hagong Hyundai in industrial robot bodies. Unlike third-party generic motion controllers—which suffer from response latency, poor trajectory tracking, and weak multi-axis coordination—the in-house cerebellum receives full process-decision output from the brain and performs dynamic trajectory correction and error compensation, effectively eliminating virtual welds, missed welds, and burn-through.
- The Body spans the HGXD-S300/600/900 series of embodied-intelligence welding robot systems (released June 2026), covering basic non-standard welding, mid-to-high-end precision smart manufacturing, and large heavy-industry production lines respectively.
Commercial traction and economics
A single 3S Robotics welding robot currently replaces an average of 1.5 to 2 human welders per unit time, with a payback period of just 1 to 1.5 years. The company has served more than 200 head customers including China Railway Group, Baosteel Steel Structures, and State Grid Corporation of China, with deployments across shipbuilding, bridge engineering, light and heavy steel structures, construction machinery, rail transit, and power equipment.
Go-to-market is led by channel distributors and integrators, with direct end-customer sales accounting for 10–20% of revenue (used primarily to build lighthouse cases). The company is also pioneering a "robot-as-labor" model that charges per welded meter, delivering a 2.88x increase in per-capita welding efficiency and a 32.5% reduction in average price versus manual alternatives.
Investor outlook
Shanghai Semiconductor Industry Investment Fund noted that AI's strength in non-standard perception and decision-making is a natural fit for welding, where automation rates have long been suppressed by high variability. Jinqiao Fund cited the company's rare integration of robot body, AI, and welding process, combined with in-house manufacturing barriers. 01VC highlighted breakthroughs in no-drawing welding, while Xinding Capital emphasized the advantages of intelligent welding robots in high-precision, high-efficiency, and harsh-environment operation.
3S Robotics is targeting revenue of hundreds of millions of yuan in 2026. Looking ahead, the company plans to use welding robots as a fulcrum to roll out intelligent robots across grinding, cutting, assembly, handling, and other industrial processes—eventually building a full embodied-AI product ecosystem spanning the entire factory floor.