AI Welding Robots: Why Path Robotics Just Raised $300 Million

Here’s a job most robots still struggle with: welding two pieces of metal together. Sounds simple, right? It isn’t. The metal bends while you work on it, and no two parts are ever quite the same.

That’s exactly the problem AI welding robots from Path Robotics are built to solve. And on November 10, investors put $300 million behind the idea. So what does this Ohio startup actually do, and why are people paying so much attention? Let’s get into it.

So what just happened?

Path Robotics announced a $300 million Series E round, led by the growth equity firm Georgian. The deal values the company at $2.45 billion after the new money comes in.

Plenty of familiar names joined. Existing backers Addition, Drive Capital, IronArc Ventures, Tiger Global and Yamaha Motor Ventures all came back. D. E. Shaw Ventures and NAVER Ventures signed on as new investors. Emily Walsh, a partner at Georgian, also gets a seat on Path’s board.

Here’s the eye-catching bit. Bloomberg reported that the new valuation is nearly four times the previous one, according to a Tech Times roundup. In total, Path has now raised more than $610 million.

So where does the cash go? CEO and co-founder Andy Lonsberry said it will help the company “scale our welding business alongside our customers.” He also plans to spend more on data, computing power and AI models for the tasks that happen before and after welding.

Why is welding so hard for robots?

Robotic welding torch joining two steel plates as heat makes the metal warp along the seam

Think about what welding actually does. You’re melting metal with intense heat. As SiliconANGLE explains, that heat warps parts as they warm up, and then again as they cool down. The technical term is heat-induced metal distortion.

The trouble is, those shifts are hard to predict. A classic factory robot follows a fixed path that someone programmed in advance. So if the metal moves even a little, the robot can end up welding in the wrong spot.

Then there’s a second headache. Parts made from the same blueprint still vary slightly. A skilled person can adapt on the spot. Old-school automation software, though, can’t easily make those tweaks on the fly.

That’s why welding has stayed so stubbornly manual in many shops. Path’s pitch is simple: let software do the noticing. In other words, the hard part isn’t moving the torch. It’s noticing what changed and reacting in real time, which is exactly where AI comes in.

How Path’s AI welding robots actually see the seam

Path’s secret sauce is an AI model called Obsidian. It watches how parts change shape during welding and adjusts the work as it goes. It also accounts for unexpected differences between parts.

To see what’s happening, the system uses four lasers on the torch to light up the metal. Then a pair of cameras captures the reflected light. That data comes in before, during and after each run.

Next comes the training. Path built what it calls a weld world model, which creates simulated welding projects for Obsidian to learn from. The company says that model draws on tens of millions of welded inches collected over eight years.

And yes, there’s a welding robot with legs. In April, Path introduced Rove, a four-legged robot that carries the same AI brain for shipbuilding work. Since it can walk, it can go to large or immovable workpieces instead of waiting for them to come to it. Lonsberry described it as “designed for flexibility over speed.” According to Path’s own page, only 50 units will ship in 2027.

Why this one’s a big deal

First, the growth numbers. Path says its annual recurring revenue grew 17 times over the last 18 months. That’s company-reported, and it hasn’t shared absolute revenue, so take it with a pinch of salt. Still, investors clearly liked what they saw.

Second, the customers. Path names shipbuilder HII, Lockheed Martin and Saronic, plus several mid-sized manufacturers. In August, HII announced performance-based deals with Path and GrayMatter Robotics worth up to $900 million combined over seven years. That’s a ceiling tied to milestones, though, not money in the bank.

Third, the ambition. “Our goal isn’t to automate one step, but to automate the entire factory,” Lonsberry said. Honestly, that’s a big leap. Welding is a narrow, well-defined task, and jumping to other factory jobs is a much harder problem.

We’ve seen this pattern before, from autonomous excavators to Physical Intelligence’s $20 billion valuation. Investors are betting that physical AI works best when it starts with one painful, specific job.

Okay, but what about the welders?

Fair question. Nobody likes the idea of a robot taking their job.

But here’s the twist. The American Welding Society projects a need for 320,500 new welders by 2029. So the real problem for many factories isn’t too many welders. It’s not enough of them.

Path goes even further. One of its videos is titled “America Is 600,000 Welders Short,” which is a bigger number than the AWS projection. Either way, the gap is real.

Lonsberry also told Bloomberg that none of Path’s customers has fired a welder because of a task robot. Path hasn’t published independent data on this, so we can’t verify it. Still, it fits the story of robots filling gaps instead of replacing people.

Our take? The likely future is a skilled welder running several robots, while also handling the tricky jobs the machines can’t. That’s a different job, and it probably needs new training.

What AI welding robots mean for Armenia and emerging markets

For countries building up their manufacturing base, this matters. According to SiliconANGLE, the software also applies to parts for vehicles, power generators, utility poles and other heavy equipment. Those are the kinds of products growing economies tend to make. So AI welding robots could help smaller factories take on precise metalwork without hiring a huge crew of specialists first.

It also points to where the jobs are heading. Someone has to set up these systems, keep them running and check the results. Those are exactly the hybrid skills, part trades and part tech, that training programmes in our region could start building now.

Of course, $300 million doesn’t guarantee anything. Path still has to prove that its AI can move beyond welding, just like Agile Robots has to prove its humanoids can scale. So here’s the question we’ll be watching. Will AI welding robots stay a niche tool for shipyards and defence contractors, or become standard kit in factories everywhere?

Scroll to Top