How Physical AI Will Rebuild America's Maritime Industrial Base
The first thing that stands out when you walk through a modern shipyard isn't the technology – it's the sheer scale of the work and the size of the parts going into these massive ships the size of city blocks. I recently visited a Korean shipyard watching teams of workers join two 50-foot-tall, 100,000-pound hull sections into place. This particular shipyard does this every day, building up to 40 ships a year.
Seeing that kind of scale firsthand puts into perspective just how much coordination, labor, and industrial capacity it takes to keep a yard operating at full speed. Huge steel structures stretch across the yard while teams of people tackle hundreds of tasks simultaneously. Every weld represents years of experience, and every section under construction is part of a much larger effort to keep production moving. What also becomes clear is that there is no shortage of work.
Across the maritime industry, demand is accelerating. Naval modernization, commercial vessel construction, autonomous maritime systems and critical infrastructure projects are all expanding at the same time. Policymakers increasingly recognize shipbuilding as both an economic priority and a national security imperative.
The urgency is hard to ignore. The U.S. has gone from building 5% of the world’s ocean-going commercial ships in the 1970s to building about 0.2% today, according to McKinsey. Beijing now operates the world's largest navy, fielding more than 370 warships compared with roughly 296 in the U.S. fleet, according to CSIS.
America already knows how to design world-class ships. The pressing challenge is increasing the industrial capacity to build them faster and at a scale we’ve never seen before.
Capacity Has Become the Limiting Factor
Order books continue to grow across both naval and commercial programs, hitting 191 million tons in Q1 2026, according to the Baltic and International Maritime Council. But U.S. production capacity has become the industry's greatest constraint.
The U.S. Navy expansion to a larger Golden Fleet is expected to require roughly 250,000 new workers across the maritime industrial base over the next decade. At the same time, the American Welding Society projects demand for more than 320,000 new welding professionals by 2030.
Shipbuilders are also competing for talent with defense manufacturers, energy infrastructure projects, data center construction, mining equipment manufacturers and heavy industry. Every one of those sectors depends on the same skilled trades.
Experienced welders continue to retire faster than new workers enter the profession, tightening an already strained labor market.
The industry's challenges are becoming increasingly visible. Programs like the Navy's Constellation-class frigate have experienced significant schedule delays and cost growth despite substantial investment, underscoring just how difficult it has become to translate demand into production.
Growing the workforce remains essential, yet even the most successful recruiting and training efforts will take years to meaningfully increase output. Adding workers alone won’t close that gap fast enough; the industry needs to rethink how the work itself gets done.
That makes improving productivity just as important as adding people.
Why Shipbuilding Has Been So Difficult to Automate
Shipbuilding has never resembled a traditional assembly line. Steel shifts during fabrication. Heat from welding changes geometry. Components rarely arrive perfectly aligned. Conditions on the shop floor evolve throughout production. Experienced welders make countless small adjustments based on what they see in front of them. You have one shot to get it right and risk thousands in material costs and rework if you don’t
Traditional industrial robots were never designed for this kind of work. They perform exceptionally well when every part is identical and every movement can be programmed in advance – producing one part exactly the same 10,000 times. When variation enters the equation, their effectiveness drops quickly.
The next generation of automation has to do much more than repeat programmed motions. It needs to perceive changing conditions, understand what is happening, and adjust while work is underway rather than simply repeating a fixed motion. Those capabilities have been missing for years.
Physical AI Expands What Can Be Automated
Physical AI changes that. Traditional automation follows pre-programmed instructions. Physical AI continuously observes its surroundings, interprets what’s actually in front of it and decides in real time what it needs to do to achieve the desired outcome.
The key difference is the model behind it. Machine learning is only as good as the data it's trained on, and physical AI is built on a massive multimodal dataset of real-world welding data that teaches the system what good welding looks like across countless scenarios. That trained model becomes the foundation for decision-making, allowing the robot to reason through variation rather than simply execute a fixed program. Combined with computer vision, planning and motion control, it can identify weld joints, compensate for variation, adapt to heat distortion and modify weld paths as production evolves.
For shipbuilding, that adaptability changes the equation. Tasks that were previously considered too variable for automation become practical. More weld hours become addressable. More production can move forward without waiting for scarce skilled labor.
These capabilities are also becoming increasingly mobile. Intelligent systems are no longer confined to fixed robotic cells. They can move throughout shipyards, fabrication facilities and other dynamic work environments, bringing advanced manufacturing closer to where the work is actually happening.
Expanding Capacity Alongside Skilled Labor
This shift is already gaining momentum across the maritime industry. Recent collaborations between shipbuilders and AI technology developers, including Huntington Ingalls Industries’ High-Yield Production Robotics (HYPR) initiative, reflect a growing recognition that meeting future production goals will require new manufacturing approaches alongside continued investment in people.
Physical AI extends what experienced welders can accomplish by taking on increased welding operations while keeping human expertise at the center of production. Skilled tradespeople remain responsible for the most complex fabrication, quality oversight and judgment-intensive work, while AI-powered systems increase the amount of productive work the entire team can complete.
Every additional welder matters. But the broader objective is increasing completed vessels, and that requires expanding productive capacity alongside the workforce. The significance extends well beyond any individual technology deployment. Increasing welding throughput, shortening production schedules, enabling more distributed manufacturing and strengthening resilience across the defense industrial base all contribute to expanding national shipbuilding capacity.
Building the Maritime Industrial Base America Needs
America's maritime future depends on sustained investment in people, facilities and manufacturing technologies that allow all of the above to work more effectively together.
Workforce development will always remain foundational, and no technology replaces the craftsmanship and experience that skilled shipbuilders bring to every vessel. But expanding the workforce alone will not close today's production gap quickly enough.
Physical AI represents one of the most promising tools for increasing industrial capacity. The shipyards making those investments today are positioning themselves to deliver more vessels, strengthen the defense industrial base and improve America's long-term maritime competitiveness. As demand continues to accelerate, the ability to consistently turn designs into completed ships will become one of the industry's defining competitive advantages.
About the Author: Andy Lonsberry is the CEO and co-founder of Path Robotics, the global leader in physical AI for manufacturing, where he leads the company’s strategy and operations with a focus on applying physical AI to longstanding challenges in manufacturing.
