Physical AI Eases Automation of High-Mix Manufacturing

Eric Chewning (left), executive vice president of maritime systems and corporate strategy at Huntington Ingalls Industries (HII) and Path Robotics CEO Andy Lonsberry (right) signed a memorandum of understanding (MOU) today to explore integrating Path’s AI-powered solution for automating welding into HII’s shipbuilding operations. Image provided by Path Robotics.

“We're in the midst of an American shipbuilding renaissance,” declares Eric Chewning, executive vice president of maritime systems and corporate strategy at America’s largest military shipbuilder, Huntington Ingalls Industries (HII). But despite the fact that the company hired 6,600 shipbuilders in 2025 and is on track to hire at least that many in 2026, this renaissance won’t be powered by people alone. According to data from the US Department of Labor, the shipbuilding industry may require about 200,000 to 250,000 additional maritime workers to satisfy demand over the next decade.

Automation can help address this deficit but has traditionally been difficult to implement in shipbuilding due to the size and complex nature of the ships being produced. According to Chewning, shipbuilding is effectively a high-mix, low-volume manufacturing application. Instead of producing one part 10,000 times, building one ship requires completing 10,000 tasks one time. Furthermore, much of the shipbuilding industry’s needs are for welders. Traditional automation has struggled to adapt to the wide variation in hand-cut workpieces that require welding. But new systems powered by AI can automate these and other similar high-mix, low-volume manufacturing applications.

Featured Content

Tool Presetters Reduce Downtime, Human Errors and Scrap READ MORE

When Will We See Humanoid Robots on the Manufacturing Floor? READ MORE

10 Tips for Dressing Your Grinding Wheel READ MORE

At an event held at Path Robotics’ headquarters in Columbus, Ohio, on February 17, 2026, the two companies signed a memorandum of understanding (MOU) to explore integrating Path’s AI-powered solution for automating welding into HII’s shipbuilding operations. They hope this partnership will accelerate throughput, strengthen the maritime industrial base and augment the shipbuilding workforce.

Path provides custom welding cells using a robotics-as-a-service model. Users rent or lease the cell, which includes fixturing, robotic arms, welding equipment, sensors, cameras, rails, support and regular maintenance. It also includes Obsidian, Path’s AI foundation model for automating welding. Image provided by Path Robotics.

A Path to Automated Welding

At the core of Path’s automated welding system is Obsidian, a foundational AI model that’s built specifically to handle welding tasks. Over the past eight years, the company has gathered data from “real-world welding episodes” and used them to train a neural network via reinforcement learning, says CEO Andy Lonsberry.

Path combines Obsidian with robots and welding equipment from outside suppliers, along with a vision system, microphones and other sensors that feed data into the system. The system begins by scanning the part. Obsidian compares data from the part scan to the part’s CAD model, identifies the differences between the two and adjusts the offsets as needed. This enables the system to weld the part accurately, regardless of how much it varies from the CAD model.

But, “It's not just about the robot and the welder,” explains Path’s Chief Information Security Officer and VP of Engineering Matt Randall. “This is a complete package. This is all designed with the customer in mind.” The company builds each automation cell according to customer needs, including customized fixturing, one or more robotic arms, welding equipment, sensors, cameras, rails and more.

Notably, instead of selling systems outright, Path uses a robotics-as-a-service (RaaS) business model. This means the company leases or rents the entire system, including equipment, software, monitoring and maintenance to users. Lonsberry says this model helps ensure user success. “What we saw in automation, traditionally, was you sold a piece of equipment to an end user and said, ‘Here it is! Good luck,’” he explains. “And we see this time and time again, where we go to a customer site and there is a traditional robot, and it's no longer on the line. It's sitting in the corner collecting dust.”

Support, monitoring and maintenance are particularly critical to making RaaS work for users. “We have a huge support staff on this to make sure that these robots show up and deliver,” Lonsberry says. The company provides 24-hour support on weekdays and 12-hour support on weekends, during which Path’s support team can monitor diagnostics and remote into any cell that’s down to resolve issues. Path also performs preventive maintenance on cells every three months, minimizing downtime and ensuring they are performing at the highest level.

Beyond shipbuilding, Path’s solutions are used in other high-mix end markets, such as AI infrastructure and utility poles (like the one seen here). Photo by MMS. 

Beyond Defense

Lonsberry describes the MOU between Path and HII as marking “day zero of the new era of defense manufacturing.” But other end markets, such as energy infrastructure, are experiencing the same bottlenecks and also seeing success with Path’s technology. “Large utility poles, lighting poles, traffic poles,” he says. “Basically, if you walk down the street, you'll most likely see something welded by us.” While all of these poles might look the same at first glance, Lonsberry explains it’s actually more of a high-mix, low-volume market. Utility pole designs differ based on municipality, location within the municipality, civil engineer, utility needs, weather, weight and more. “Every pole is different,” he says.

But perhaps the largest market for physical AI is AI itself. Lonsberry says AI infrastructure such as data centers is Path’s fastest growing market. “It was probably over half of our growth in 2025, and we only see this continuing to grow very aggressively.” This market isn’t challenging because it’s a high-mix, low-volume application — its needs are both high mix and high volume. “They have an immense amount of need for continuing to produce and produce that extreme quantity,” he explains.

Beyond end markets, Path sees applications for automation powered by physical AI in other areas of manufacturing. “Physical AI for welding is just the first step forward for us,” Lonsberry says. “We look at it as the heartbeat of the company, but every task before welding and every task after welding we ultimately look at as a task we can solve with physical AI.”