The following article is from 36Kr Future Industry, author Yang Yuexin, editor A Zhi
36Kr has learned that3Srobotics, an embodied AI robot company focused on industrial manufacturing, officially announced the completion of a Series B financing round worth hundreds of millions of RMB. The round was led by Shanghai Semiconductor Industry Investment and Jinqiao Fund, with continued follow-on investment from Lingyi Venture Capital, Xinding Capital, Zhongguang Dinghua and existing shareholder Weiguang Venture Capital.The funds will mainly be used to iterate and upgrade the welding embodied AI brain and the autonomous cerebellum motion control system, expand the capacity of its own smart factory, expand R&D and market service teams, and accelerate the large-scale deployment of its products across industries nationwide.
Founded in 2020, 3Srobotics obtained National High-Tech Enterprise certification in 2023, and became a Shanghai “Specialized, Refined, Distinctive and Innovative” SME and a Shanghai Intelligent Robot Benchmark Enterprise in 2024. The completion of this round marks 3Srobotics’ entry into a new stage of large-scale high-speed development through its “autonomous brain + autonomous cerebellum + self-developed hardware mass production” three-tier full-stack self-developed system, entering the domestic industrial-manufacturing embodied AI track with mass-producible, autonomously evolving industrial-grade intelligent robot products.
Targeting Welding’s Rigid-Demand Scenarios to Create an “Intelligent Welder”
At its founding, 3Srobotics was mainly engaged in the R&D, production and sales of general-purpose industrial robots.As the business developed, founder Wang Dezhao gradually formed the judgment that robots would evolve from general-purpose toward deep intelligence in vertical scenarios.“So we positioned ourselves at the deep integration of AI algorithms and robot hardware, fully leveraging the value of intelligent robots in vertical scenarios.”
Among many industrial scenarios, 3Srobotics chose welding.Welding is a process involved in almost all industrial production, but due to its high danger, high intensity and high technical barriers, the labor shortage is worsening. Wang Dezhao noted that China currently has about ten million registered welders, with an average age above 45. As young people are generally unwilling to take up welding work, the welding engineer gap will reach millions by 2025. Meanwhile, manufacturing enterprises’ requirements for welding quality, efficiency and stability keep rising.
The convergence of these two pain points has created enormous market space for welding robots. According to estimates, the intelligent welding robot market will reach RMB 100 billion over the next decade. “Based on confidence in the welding market’s prospects, as well as the customer resources and scenario data we previously accumulated in the welding field, we ultimately chose to first deeply cultivate the vertical welding scenario,” Wang Dezhao said.
Most traditional system integrators’ welding automation solutions require pre-programming, drag-teaching or manual intervention, and cannot automatically adjust processes based on operating errors. They can only solve standardized problems within a limited range, cannot deeply understand industrial physical working conditions or autonomously handle non-standard complex scenarios, and have inherent shortcomings such as poor adaptability, high replaceability and no continuous evolution capability — making it hard to truly solve the core pain points of manufacturing.
By contrast, 3Srobotics’ robots can learn from mature welder operation data through AI algorithms, and through 3D real-time reconstruction and robot path planning, autonomously perceive and understand the physical world to make real-time decisions and dynamically adjust processes — better adapting to various complex working conditions and non-standard workpieces, while shortening product delivery and enterprise payback cycles. Wang Dezhao said that one 3Srobotics welding robot can replace an average of 1.5 to 2 welders per unit time, with an investment payback cycle of about 1–1.5 years.
Self-Developed Welding Embodied Autonomous “Brain” and “Cerebellum”
3Srobotics has self-developed a welding-specific embodied AI brain model, trained entirely on real industrial working conditions. The “brain” is an AI algorithm based on a multimodal foundation-model architecture. By feeding it tens of millions of frontline welding data points, a full-category material process knowledge base and complex irregular weld-seam datasets, the AI algorithm can — like an experienced welder — identify the spatial structure, material and texture of workpieces, and autonomously reason to generate process solutions for “where to weld” and “how to weld”, adapting to non-standard, multi-variety, small-batch core production scenarios in steel structures, construction machinery, shipbuilding, rail transit and automotive parts.
This autonomous brain has genuine industrial process cognition and autonomous decision-making: leveraging its self-developed high-precision 3D vision system, 3Srobotics’ “brain” model dynamically matches and adjusts full sets of core process parameters — current, voltage, travel speed and weaving amplitude — according to plate material, thickness, groove shape and on-site operating environment. Meanwhile, the model supports closed-loop self-learning of production-line data, continuously optimizing process logic and breaking the industry bottlenecks of fixed processes, heavy reliance on manual experience and limited scenario adaptation.
3Srobotics’ self-developed “cerebellum” capability — a self-developed real-time motion control system — originates from controlling subsidiary HGXD’s years of accumulation in industrial robot bodies, and is the key technical moat ensuring precise on-the-ground decisions and stable, high-standard operation.
Unlike the purchased general-purpose motion control solutions used by most manufacturers, which commonly suffer from response latency, insufficient trajectory tracking precision and poor multi-axis coordination, 3Srobotics’ autonomous “cerebellum” receives the full set of process-decision instructions from the “brain” and, faced with extreme complex conditions such as irregular curved surfaces, extra-large components, confined spaces and workpiece deformation, completes trajectory correction and dynamic error compensation in real time, ensuring uniform, stable weld formation and eliminating quality problems such as false welds, missed welds and burn-through. It can adapt to the full range of fixed and crawler-type mobile welding robots, achieving “one brain driving multiple devices for full-scenario flexible autonomous operation”.
3Srobotics’ massive accumulated process data serves as the bridge and learning fuel between the “brain” and “cerebellum”. As the practical application scale of the robots continues to grow, product iteration forms a positive flywheel of “the more they’re used, the more data there is, the smarter the model becomes”.
From Deep Welding Cultivation to Broadening Embodied AI Product Boundaries
Traditional welding automation solution providers, needing custom development and on-site adaptation for each specific project, mostly adopt a sales model of direct contracting with end customers. By contrast, 3Srobotics’ welding robot products are more standardized and intelligent, achieving “plug-and-play” for customers.
On this basis,3Srobotics’sales model relies mainly onchannel dealersand integration agents.Directly contracted end customers account for only 10%–20% of revenue, mainly used to build benchmark cases for newly developed application scenarios. In addition, 3Srobotics is exploring a “robot labor” business model, billing by welding meters. Compared with per-person welding, daily welding efficiency increases by 2.88× and average prices drop by 32.5%.
Currently,3Srobotics’customers mainlycoversteel structures, bridges, shipbuilding, rail transit, power equipment, construction machinery and otherindustrial manufacturing scenarios,with involvementin high-end manufacturing fields such as aerospace and new energy.Wang Dezhao noted that when using welding robots, customers care most about welding stability, coverage of weldable workpieces and per-unit-time efficiency, with a 2026 revenue target in the hundreds of millions. Next, 3Srobotics plans to use welding robots as its core product fulcrum and continue to launch a full range of industrial intelligent robot products for grinding, cutting, assembly and material handling, gradually penetrating the full production flow of manufacturing enterprises and upgrading from a “single-scenario hit product” to a “full-process embodied AI product ecosystem”.
Investor Perspectives
The Shanghai SemiconductorIndustry Investment team said:In industrial scenarios, the welding field has always faced an extremely high non-standard rate, which keeps its automation rate chronically low and forces continued reliance on manual operation. Yet the welding work environment itself is extremely harsh, meaning that the millions of welder positions of the past will face a severe labor gap in the future. What AI actually solves is precisely the judgment and execution problems in all kinds of non-standard scenarios across every industry — a property naturally matched to welding.
After overcoming multiple complex and interwoven problems that require long periods of accumulation — immature AI models, insufficient data and inadequate robot-body precision — 3Srobotics has finally become the first to give AI welding robots capabilities comparable to mature welders and achieve mass production and shipment. The welding industry is large enough and rigid enough in demand, and we are optimistic that 3Srobotics can grow into a much larger company with its scarce, leading products.
Jinqiao FundGeneral ManagerMa Wanli said:3Srobotics holds scarce value in the industrial embodied AI welding robot field. The company possesses a rare domestic “robot body + AI + welding process” three-in-one autonomous core capability, especially the mass-production capability and supply-chain self-controllability brought by its own robot-body production factory, building a solid moat that distinguishes it from pure software solution providers.
Welding is the most rigid-demand market in industrial manufacturing, with the largest labor gap and the most urgent need for intelligent replacement. 3Srobotics’ productized path of “autonomous brain + self-developed hardware” precisely fits the industry trend of manufacturing transforming from “solution empowerment” to “intelligent hardware deployment”. Jinqiao Fund will leverage its long-term positioning and ecosystem resources in hard technology and intelligent manufacturing to continuously empower the company’s scale-up, accelerating the deep deployment of welding robots in steel structures, shipbuilding, construction machinery and other fields.
Lingyi Venture Capital Managing Partner Yu Lu said:The “machine-replacing-human” trend is evident in non-standard manufacturing fields such as steel structures and shipbuilding, and 3Srobotics precisely targets this market gap. 3Srobotics has complete technical capabilities, and in software algorithms has cracked the industry challenge of drawing-free welding, enabling robots to understand working conditions and make autonomous decisions like a seasoned veteran craftsman.
We believe 3Srobotics demonstrates the technical depth, product thinking and global vision of a unicorn, and we are willing to accompany such excellent teams to embrace the wave of industrial intelligence together.
Xinding Capital Head Zhang Chi said:As core equipment in industrial automation, welding robots feature high precision and high efficiency. Through precise program control and multi-axis coordinated motion, they ensure consistent welding parameters and stable, uniform weld-seam quality, effectively resolving the quality fluctuation caused by fatigue or skill differences in manual welding. In efficiency, welding robots can work continuously 24 hours, greatly shortening production cycles and boosting capacity — especially suitable for large-batch standardized production. Meanwhile, with flexible programming and intelligent technologies such as visual recognition, they quickly adapt to small-batch, multi-variety and complex-shape workpiece welding, significantly shortening the preparation cycle for product changes. Moreover, welding robots can work stably over long periods in harsh environments with high temperatures and toxic dust, reducing worker labor intensity and safety risks.
