Path Robotics raises $300M at a $2.45B valuation for welding AI

Path chose welding because it is one of the hardest factory jobs to automate. Traditional robots follow fixed paths, while Path's Obsidian model lets a cell identify weld joints, plan and execute welds and adjust as conditions change[1].

Georgian says Obsidian is trained inside a second model, the Weld World Model, a neural simulator built from tens of millions of welded inches of fleet and internal data. Data from every weld feeds back into it, and Path's internal tests show a first-pass yield above 97%, compared with roughly 70% to 85% for a human welder on the same complex, high-mix jobs[2].

"Our goal isn't to automate one step, but to automate the entire factory," CEO and co-founder Andy Lonsberry said[1].

Annual recurring revenue has grown 17-fold in 18 months, and headcount is up 97% year on year with a target of 400 staff by year end[1]. Customers include HII, Lockheed Martin and autonomous vessel maker Saronic. Earlier this year Path signed a US$600 million performance-based production agreement with HII covering US Navy shipbuilding programs, which Georgian says runs over seven years[1][2].

The money will take Path into new manufacturing tasks and robot form factors, along with more compute and model training[1]. It has already shown Rove, which puts Obsidian on a quadruped robot so welding can happen in the field at the part. Lonsberry told Bloomberg the company also plans to expand into defence and energy infrastructure[3].

Path is one of the clearest examples of physical AI earning real revenue on production lines rather than in demos. Its pitch, that every deployed weld improves the model for the whole fleet, is the same data flywheel humanoid and warehouse robot makers are chasing, applied to a skilled trade facing a chronic labour shortage.