Path Robotics Brings Physical AI to Welding

Last week, Saronic hit a $9.25 billion valuation, courtesy of a fresh $1.75 billion in funding. Venture capital has poured in on the strength of the firm’s partnership with the U.S. Navy and the promise to deliver more than 20 autonomous ships by 2027. It’s an ambitious goal — enough to have more than doubled the company’s $4 billion valuation since early last year. 

As conflict has raged, demand for domestically produced warships has spiked. The proposed massive 2027 $1.5 trillion defense spending plan calls for Congress to approve the construction of 41 new ships, including 18-battle and 16-non-battle naval vessels. While Saronic has an ambitious plan to address these figures — including the construction of new facilities – it won’t be able to go it alone. 

In February, the company announced a partnership that will bring Path Robotics’ automated welding systems to its Franklin, Louisiana shipyard, the former home of the 100-acre Gulf Craft facility. Within days of that deal, Path announced a partnership with barge-builder LAD and a memorandum of understanding (MOU) with HII, the largest military shipbuilder in the U.S. (the latter, notably, was a promise to “explore the integration” of Path’s tech).  

Among the more remarkable bits of all this — at least from where Path stands–– is that this whole shipbuilding business is a fairly new undertaking for the Columbus, Ohio-based robotic welding firm. 

“The last 18 months have been extremely busy,” CEO Andy Lonsberry tells A3. He describes the startup’s early journey as one focusing primarily on small- and medium-size business, primarily operating in automotive. As 2024 turned into 2025, however, the Path begun looking at larger, more complex customers with even higher quality standards.  

“We opened up new verticals in defense, shipbuilding, AI data centers, energy infrastructure, and then also the classic kind of construction mining equipment in 2025, we've just seen massive market pull, massive market adoption.” 

In September, Path announced Obsidian — “Physical AI for Welding” — per the press material. Lonsberry tells me the company has collected north of 200,000 hours of welding data, so it did the only logical thing and built a model. “Model building’s an everything thing for all of us in robotics,” Lonsberry notes, adding that the trick is getting as diverse a data set as possible, while still keeping customers as happy as possible.   

Getting really good at welding across a range of different industries is an immediate goal, of course, but Lonsberry says Obsidian is the first step toward something even larger for Path.  

“That allows us to have this huge repository to really make this generalized model what people would call ‘task specific,’ which is welding,” he explains. “With welding there's 1,000 standards, and totally different material types from group to group, so it's allowed us to get to good generalization inside of the vertical welding. It’s allowed us to then take key learnings of how to build large models, distribute large models, fine tune models that we're now taking those key learnings to start to like pass welding into the verticals and the tasks that happened before and after welding.” 

Such examples include assembly, along with post-weld finishing. Lonsberry adds that, going forward, Path’s robotic embodiments are set to extend between the massive systems currently bolted to the ground at partner sites.  

“In our vision, in our future, we see all sorts of embodiments being useful, being productive, being powerful,” he says. “But there will absolutely be work cells as well as workers. We see the future of the embodiment should be purposed for the utilization of what [the customer] wants to be accomplished at that facility or those sets of tasks.”