Path Robotics raises $300M to automate industrial welding with AI
The startup secured Series E funding to develop AI models that adjust welding robots in real-time to account for metal warping and manufacturing inconsistencies.
The short version
- Path Robotics raised $300 million in Series E funding, reaching a post-money valuation of $2.45 billion.
- The investment was led by Georgian with participation from Yamaha Motor Co. and South Korean tech firm Navier Corp.
- The company's AI model, Obsidian, uses lasers and cameras to adjust welding paths in real-time based on metal distortion.
- The technology is being applied to diverse industries, including shipbuilding, vehicle manufacturing, and power generator production.
Why it matters
Traditional automation fails in welding because heat warps metal in unpredictable ways. By using AI to 'see' and react to these changes, Path Robotics is unlocking high-volume production for heavy industries that previously relied on manual labor.
Securing the capital for scale
Path Robotics has officially announced a $300 million funding round to accelerate its work in industrial welding automation. This Series E round was led by the investment firm Georgian and included a consortium of major tech players. The group of investors included the venture capital arms of Yamaha Motor Co. and Navier Corp., a prominent South Korean technology firm.
The new capital brings the startup's post-money valuation to $2.45 billion. This significant investment comes at a time when manufacturers are looking for ways to increase production capacity. According to Path Robotics co-founder and CEO Andy Lonsberry, there is accelerating demand from companies that cannot reach their goals using traditional automation methods.
The company intends to use the funds to scale its existing welding business alongside its current customers. Beyond immediate growth, the money will be directed toward the heavy lifting of AI development. This includes investing in the necessary data, compute power, and model refinement required to handle every step of the manufacturing process before and after the welding occurs.
Solving the warping problem
Automating the welding of metal parts has historically been a difficult engineering challenge due to a phenomenon known as heat-induced metal distortion. When a welding torch applies intense heat to join two pieces of metal, the parts often warp or bend. As the metal cools down, it can change shape again, creating unpredictable movements that defy static programming.
Standard industrial robots typically follow a pre-programmed path, which works well for consistent, rigid parts. However, because welding heat causes parts to shift, a static plan often results in poor-quality joints or errors. Additionally, mass-produced parts can have slight manufacturing inconsistencies that make it nearly impossible for traditional software to maintain precision across thousands of units.
Path Robotics addresses these variables by creating a system that can observe and react to the physical state of the metal in real-time. Their approach moves away from rigid instructions toward a dynamic workflow. By accounting for both the heat-induced warping and the minor variations in the raw parts, the system allows for a level of precision that was previously only achievable by human welders.
How Obsidian powers the torch
At the core of this technology is an AI model called Obsidian, which is specifically designed to guide industrial robots equipped with welding torches. Obsidian monitors the welding workflow by utilizing four lasers attached to the torch. These emitters illuminate the parts as they are joined, while a pair of cameras capture the reflected light to provide a constant stream of visual data.
The system does not just watch the welding happen in the moment; it also collects data before and after each specific run. This comprehensive data set is used to power what the company calls a weld world model. This model allows researchers to generate simulated welding projects, which serve as a training ground for the Obsidian AI to learn from various scenarios.
By training on these simulations, Obsidian becomes better at predicting how metal will behave under different conditions. This training workflow ensures that the robot can make micro-adjustments to its path instantly. The result is a robot that can navigate the complexities of a welding cell while maintaining the high standards required for industrial-grade components.
From welding cells to ships
While Path Robotics originally developed Obsidian for large robotic workstations known as welding cells, the technology is now expanding into more mobile applications. In April, the company introduced Rove, a four-legged robot designed to perform automated welding tasks in the shipbuilding industry. This move demonstrates the versatility of the underlying AI model when applied to complex, large-scale construction environments.
The software is not limited to just one type of manufacturing; it is being adapted for a wide variety of heavy equipment. Path Robotics notes that the technology can be used to manufacture parts for vehicles, power generators, and utility poles. These are all industries that require high-strength welds and can benefit from the precision of AI-guided automation.
As the company scales, the goal is to provide a solution for any manufacturer struggling with production bottlenecks. By moving beyond simple repetition and into adaptive automation, Path Robotics aims to take on the tasks that surround the welding process. This includes the preparation and finishing steps that are often the most difficult to automate in a traditional factory setting.
What to watch
- How Path Robotics plans to integrate 'pre-welding' and 'post-welding' tasks into a single automated workflow.
- The pace at which the 'weld world model' can simulate new types of metal alloys or complex geometries.
Try it
Look up the 'Rove' robot to see how four-legged designs are being used for heavy industrial construction like shipbuilding.
Sources and links
- ReportingSiliconANGLE: the article we sourced
First reported by SiliconANGLE. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.