"In the first half of 2026, our welding robot shipments have already caught up with the full-year level of 2025, with sales nearly doubling," said the head of robot products at STEP (STEP) to GGII Robotics.
This growth is not an isolated case. Liang Rui, General Manager of Qianjiang Robot, revealed that sales of Qianjiang Robot bodies grew nearly 1x in the first half. Zhou Dasheng, founder of Jiuchuan Robot, stated that shipments of collaborative welding robots in the first half achieved several-fold growth, already exceeding last year's full-year level.
Through research, GGII Robotics has learned that the rapid growth of welding robots in H1 2026 is almost a common performance of outstanding enterprises in the industry. Stimulated by this boom in terminal demand and orders flying together, new welding robot products are emerging in endlessly.
According to GGII Robotics statistics, in the welding robot field, nearly 50 new products were released in H1 2026, far exceeding the same period in 2025 (20 products). New products cover industrial robots, collaborative robots, embodied intelligent robots, teaching-free welding systems, as well as supporting components such as 3D vision cameras and mobile tracks.
2026: The "First Year" of Embodied Intelligent Welding
The embodied intelligence boom that emerged at the end of 2024 has finally swept into the intelligent welding field.
Welding has always been the core process in fields such as automobile manufacturing, bridge construction steel structures, shipbuilding, port machinery and mining equipment, and metal machining. Traditional industrial welding robots can handle high-tempo, large-batch, standardized production operations, but lack flexible operation capabilities and are often helpless when facing mixed production changeovers and multi-variety small-batch applications.
The emergence of embodied intelligence technology brings dawn to breaking the "rigid operation" deadlock.
Facing workpiece shape and position, incoming material deviations, and weld seams, embodied intelligent robots can initially realize the "autonomous perception-decision-planning-execution" welding process through AI + 3D vision + intelligent welding control system + process library. After more than a year of technical polishing and POC scenario verification, some enterprises have launched the latest intelligent welding solutions.
In H1 2026, companies such as STEP, Visualbit Robotics, 3Srobotics (Shengshi Weisheng), Zhongxun Technology, CSCEC Steel Structure, Kepler Robotics, Kaierda, and Path Robotics all launched embodied intelligent robot welding solutions.
STEP showcased for the first time at the Essen Exhibition the "dual-machine collaborative" humanoid robot welding workstation: one wheeled robot completes workpiece loading/unloading and material sorting and transfer, while another upright humanoid robot performs weld detection and welding operations. The two units adopt an integrated multi-layer heterogeneous control system to achieve "dual-machine shared brain," achieving task synchronization at the millimeter level. Humanoid single-arm load is optional from 6kg-20kg, with operation repeatability of ±0.1mm, enabling flexible changeover within 10 minutes with the welding process package.
Visualbit Robotics and Kepler Robotics both demonstrated bipedal humanoid robot welding solutions this year, but with very different technical routes. Visualbit Robotics focuses on completing adaptive welding through "AI vision + teaching-free + process library parameter adjustment," meeting welding needs for ship special-shaped curved surfaces, large steel structures, and complex mechanical structural parts. Kepler Robotics' humanoid welding solution is based on VR low-latency teleoperation welding, with motion reproduction accuracy of 0.1mm, supporting real-time welding data collection, suitable for high-risk scene welding.
If humanoid robots can solve the "rigid constraints" of traditional welding station circulation, workpiece identification, and weld detection, then the emergence of quadruped welding robots further breaks the intelligent mobile boundary of real welding scenarios, making "zero modification" intelligent welding possible in production line workshops.
Companies such as 3Srobotics, Zhongxun Technology, Kaierda & Jingshi Technology, and Path Robotics have all launched quadruped robots for welding scenarios.
3Srobotics released for the first time this year the HGXD mobile quadruped robot welding workstation. This embodied welding station integrates four core technical modules: dynamic perception, accuracy compensation, decoupling function, and digital archive. It can complete workpiece identification, weld positioning, and welding path evaluation without CAD models, with full archiving of welding trajectory data, environmental data, etc.
Facing complex industrial scenarios such as shipbuilding, large steel structures, and offshore platforms, the robot uses a quadruped structure to cope with complex terrain crossing, stairs climbing and other mobile challenges, realizing zero-modification welding entry.
From the new products released by various enterprises, it is not difficult to see that the form of current welding robots has undergone transformation: from fixed industrial/collaborative robots, to composite robots, humanoid robots, and quadruped robot welding solutions. This greatly breaks the restrictions of traditional stations and welding poses, can better perceive workpiece, environment, and welding equipment status, and flexibly operate and adapt in complex working conditions and limited spaces such as steel structures, ship and marine engineering, and mining machinery.
Many enterprises still have doubts about humanoid/quadruped robot welding, such as operation trajectory accuracy, reliability, battery life, process richness, and scenario adaptability.
But many heads of enterprises specializing in welding robots told GGII Robotics that from teaching-free to embodied intelligent welding is the inevitable direction of future technology development, and most welding robot enterprises will transform into embodied intelligent enterprises in the future. The data flywheel of embodied intelligence "data collection-algorithm training-model deployment-field feedback" will be significantly amplified in the welding field, and as welding intelligence problems are solved, process and physical AI carrier problems will naturally be gradually overcome.
GGII Robotics noted that global shipbuilding giants such as Japan's Kawasaki Heavy Industries, South Korea's Hyundai Group, and Italy's Fincantieri have all taken action to develop human-like welding robots, quadruped welding robots and other products.
Perhaps, terminal enterprises are more eager to embrace embodied intelligence than robot manufacturers who are still waiting and watching.
Collaborative Welding Robots: The Year of Outbreak
Due to issues such as load weight, technical maturity, and body reliability, collaborative welding robots were not widely favored when they first entered the domestic market in 2020.
Fast-forward to 2026, three significant changes have occurred in the collaborative welding robot field:
- Greatly improved technical maturity and product form richness
- Increased market acceptance, preliminary reputation accumulation among terminal seed users
- Large load, lightweight, modular portable integration, and multi-scenario adaptability improvement
According to incomplete GGII Robotics statistics, nearly 15 new collaborative welding robot products were released in H1 2026, from FA Robotics, Qianjiang Robot, STEP, Huayan Robot, Jicui Intelligent Manufacturing, Jiuchuan Robot, CSCEC Steel Structure, Zhichang Group, Han's Laser, FANUC, Japan System LSI, Sweden ESAB, etc.
STEP, FA Robotics, Qianjiang Robot, Jiuchuan Robot, Jicui Intelligent Manufacturing, etc. have successively released lightweight collaborative welding robots. These machine bodies weigh between 11-22kg, can be moved and carried with only one hand, and cooperate with magnetic bases to achieve "weld as you go," responding to mobile welding demands such as ship segmentation and large steel structure workpieces.
To achieve intelligent and adaptive welding of workpieces of different sizes, domestic robot manufacturers have successively launched solutions such as multi-variety small workpiece mixed-flow welding, gantry welding, ground rail welding, positioner welding, mobile chassis welding carts, and multi-machine collaboration in the past two years, responding to diverse needs such as ship small and medium group segmentation, large steel structures, and small workpiece mixed production.
The root of all these changes mostly comes from the demand reversal of terminal enterprises: in the first half of the year, Chinese shipyards won 31 million tons of orders, accounting for 72% of the global total, an increase of 113% compared to the same period last year; steel structure manufacturers such as Honglu Steel Structure and Changjiang Jinggong Steel Structure all saw production and sales growth of over 12% year-on-year in the first half; the added value of heavy industry transportation equipment such as railways, ships, and aerospace exceeded 11%.
Terminal prosperity has brought numerous welding demand extensions, one of the distinctive signals is that collaborative robots are "accelerating" into shipyards, and teaching-free collaborative welding robots for ship segmentation manufacturing have become "hot cakes." At the just-concluded 29th Essen Exhibition in early July, FA Robotics, Qianjiang Robot, Huayan Robot, Zhongke Xinsong, STEP, Jiuchuan Robot, Yuejiang Robot, Elite Robot, etc. all launched collaborative welding robot solutions for the shipbuilding field, and many enterprises have already won substantial orders.
In addition to the shipbuilding industry where manufacturing and processing orders are scheduled until 2030, heavy industry fields such as bridge/building steel structures, mining/port machinery, metal machining, as well as subdivided fields such as environmental protection equipment, also show quite strong automation upgrade demands.
According to GGII Research Institute data, in 2025, China's collaborative welding robot sales were 5,912 units, a year-on-year increase of 93%. GGII predicts that benefiting from market penetration and declining application costs, this number will break through the 8,000-unit mark in 2026.
The robot industry is moving from "body competition" to "system capability competition." With the transmission of terminal demand applications, the intelligent and easy-to-use requirements of collaborative welding robot bodies are also forcing component and accessory manufacturers to innovate.
Although teaching-free technology has made rapid progress, drag teaching is still a necessary link for most collaborative robots to cope with complex working conditions and highly shaped workpieces. To adapt to the easy-to-use needs of collaborative welding robots, Huate CNC released a collaborative drag handle teach pendant in March this year.
The handle adopts ergonomic design, equipped with advanced six-dimensional force control assisted movement technology, enabling extremely smooth dragging experience even for large arm span mechanical arms. Through algorithm optimization, trajectory accuracy is significantly improved, minimizing post-debugging costs. It integrates RS485, Ethernet (100M), EtherCAT, and Wifi/Bluetooth multiple communication protocols to ensure seamless docking of equipment with various industrial scenarios.
From the blooming of collaborative welding robots at the Essen Exhibition to the arc flashing on the terminal production lines of heavy manufacturing industries, what is visible to the naked eye is that collaborative welding robots, with lighter postures and smarter "brains," are penetrating more subdivided scenarios, welding out an ascending curve of Chinese intelligent manufacturing among steel and sparks.
This transformation driven by demand, technology, and ecological collaboration has just kicked off, and the breakthrough of 8,000 collaborative welding robots is just the beginning.
Teaching-Free Technology Advances Again
The core composition of a teaching-free welding robot system cannot be separated from the full-chain collaboration of "core components + body + algorithm + process."
According to incomplete statistics, in the first half of the year, companies such as STEP, FA Robotics, Huayan Robot, Jicui Intelligent Manufacturing, Minyue Technology, Baichu Electronics, and Weimeng Co. all released or updated teaching-free welding system solutions; camera manufacturers such as Runtec, Weijing Technology, and Qianyi Technology also brought 3D vision cameras with strong light resistance, large depth of field, multi-mode integration, and adaptation to welding scenarios.
In the past few years, teaching-free technology has been extremely dependent on vision technology drive, such as imaging accuracy, anti-glare, and anti-arc light capabilities. With the development of embodied intelligence technology, camera point cloud and reverse modeling have gradually become standard for teaching-free systems. Intelligent welding software systems composed of welding processes, data parameters (weld seams, point clouds, trajectories, attitudes, speeds), AI deep learning capabilities, etc. are becoming new technical breakthrough objects.
Many enterprises combine embodied intelligence, visual servo algorithms with welding process knowledge bases to achieve multi-layer multi-pass welding gun attitude adaptation, weld dynamic positioning, automatic compensation for assembly errors and thermal deformation; AI models are iterating from pure data-driven to mechanism + data dual-driven, effectively improving problems of insufficient welding quality stability and weak generalization ability.
Jicui Intelligent Manufacturing released embodied intelligent teaching-free technology in March this year. Relying on the multimodal fusion capability of the embodied intelligent VLA model, robots can independently identify workpieces and generate welding trajectories, widely covering scenarios such as shipbuilding, port machinery, and steel structure processing, greatly improving welding accuracy and operation efficiency, enabling intelligent equipment to accurately respond to various operation requirements under complex working conditions.
Through the publicly released teaching-free system solutions in the industry in the first half of the year, the industry's technological evolution is slowly becoming clear: penetration from high-proportion scenarios of steel structures to welding in ships, heavy industry, and special fields; advancement from automatic identification of straight seams and simple curved welds to identification and cutting of long welds and special groove welds; transition from basic process packages to complex processes such as multi-layer and multi-pass welding, with continuously improving welding qualification rates.
According to data from the "2025 Blue Book on the Development of the Intelligent Welding Robot Industry" released by GGII, the accuracy of line laser weld seam tracking sensors has reached 0.01mm, and in pilot applications of multimodal fusion equipment in complex scenarios such as ships and steel structures, the welding qualification rate has reached over 98%, and one-time debugging can adapt to multi-variety workpiece changeovers, greatly reducing operation thresholds.
In addition to emerging industry forces such as embodied intelligence, collaborative robots, and teaching-free technology, it is undeniable that current arc welding robots still occupy the mainstream market of domestic welding robots. Domestic enterprises represented by Estun, Siasun Robot, Moka Robot, Canop, STEP, Kaierda, etc. are moving forward from "middle and low-end substitution" to "middle and high-end breakthroughs."
Muka Robot's new generation of high-speed welding robots MB10W-1500 and MB10W-2050 released at this year's Essen Exhibition successfully achieved product generational leaps. This transcendence not only maintains the inherent price advantage of domestic brands but also achieves leadership over international brands in key performance indicators.
The body of Muka's new generation of high-speed welding robots adopts self-developed aluminum-iron composite special alloy materials, reducing the overall weight by 30kg, 15% lower than foreign-funded brand products of the same specification; the core self-developed customized motor has higher torque, peak speed, and overload capacity, further improving operation tempo, 20% faster than foreign-funded benchmark brand machines of the same load, with a selling price of only 40% of similar foreign-funded machinery.
From hardware breakthroughs in visual perception, to data-driven upgrades of process algorithms, to dual efforts of traditional arc welding robots in performance and cost performance, teaching-free welding technology in H1 2026 is presenting a clear context of "both hard and soft, collaborative evolution."
Teaching-free is no longer just a concept demonstration on the exhibition stand, but has truly sunk into complex working conditions such as ship segmentation, steel structure workshops, and mining machinery, becoming a "smart partner" trusted by frontline workers. When AI vision and deep learning are deeply integrated with welding mechanisms, teaching-free technology is accelerating from "auxiliary tool" to "autonomous decision-making brain," paving the most solid foundation for the comprehensive implementation of intelligent welding.
Final Thoughts
The release of new welding robot products in the first half of 2026 clearly reveals that the industry is undergoing a profound transformation—from fixed bases to quadruped running and bipedal upright, from teach pendant programming to AI vision autonomous perception and decision-making, from single-station islands to multi-machine collaborative shared brains, the boundaries of welding robots are being pushed to more complex ship cabins, higher steel structures, and more dangerous mines.
At the same time, traditional arc welding robots have not exited the stage, but are accelerating domestic substitution driven by both performance and cost performance, forming a solid foundation for this transformation.
Three forces—morphological breakthroughs in embodied intelligence, category expansion of collaborative robots, and in-depth advancement of teaching-free technology—are forming synergy in 2026, jointly rewriting the industrial narrative of Chinese welding robots. 2026 is not only a "big year" for the outbreak of new welding robot products, but also a historical inflection point for embodied intelligent welding from concept to large-scale implementation. Under the arc light, a new path of intelligent manufacturing woven by steel and code is becoming increasingly clear.