Latest / Minds, Bodies, and Terawatts / HII Partners with GrayMatter Robotics to Deploy AI in Shipbuilding

Transcript

HII Partners with GrayMatter Robotics to Deploy AI in Shipbuilding · Minds, Bodies, and Terawatts

• 0:00
  [Host] This is Minds, Bodies, and Terawatts, where the welders are retiring and the robots are clocking in. Marine News Magazine reported on April 7th that Huntington Ingalls Industries, the Pentagon's largest shipbuilder, is partnering with GrayMatter Robotics to put autonomous AI robots on shipyard production floors.
• 0:13
  [Guest] HII builds nuclear submarines and aircraft carriers for the U.S. Navy. They reported 14 percent throughput growth in 2025 and they're targeting another 15 percent in 2026. The robots aren't a productivity experiment. They're a response to a workforce that's physically disappearing.
• 0:25
  [Host] When you say disappearing, what's the actual problem? Are people quitting, retiring, or just not showing up in the first place?
• 0:32
  [Guest] All three. Skilled shipyard trades like welding, grinding, and surface prep require years of apprenticeship. The experienced workforce is aging out, and the pipeline of replacements never filled back up after decades of offshoring hollowed out American manufacturing training. HII can't build Virginia-class submarines on schedule if they can't staff the yard.
• 0:47
  [Host] So this isn't a company choosing robots over people. It's a company that literally cannot find enough people.
• 0:52
  [Guest] And the defense urgency makes it different from a warehouse operator saving on labor costs. The Navy's shipbuilding backlog is a known problem. The Government Accountability Office has flagged schedule delays on the Columbia-class ballistic missile submarine program, which the Navy calls its number one acquisition priority. When your nuclear deterrent is running behind schedule because you can't hire enough welders, the calculus changes fast.
• 1:11
  [Host] What does GrayMatter actually do? Are we talking about humanoid robots walking around the shipyard?
• 1:16
  [Guest] No, GrayMatter specializes in autonomous robotic systems for surface finishing, grinding, sanding, polishing. Their AI uses 3D sensing and machine learning to adapt to irregular surfaces in real time. The hull of a submarine is curved, complex, and every section is slightly different. Traditional industrial robots need precise programming for each piece. GrayMatter's systems figure it out on the fly.
• 1:34
  [Host] So robotic arms with AI brains, not humanoid figures carrying welding torches.
• 1:37
  [Guest] Correct. Industrial automation tuned for tasks that are physically grueling and increasingly hard to staff. A skilled human working on a ship hull is doing demanding, repetitive work in confined spaces, often in extreme heat or cold. The robots handle the dull and dangerous portions, and the remaining workers focus on higher-judgment tasks like inspection, fitting, and complex assembly.
• 1:55
  [Host] Now, HII says 14 percent throughput growth last year. I want to pause on that. That's a self-reported number from a company press release. Shipbuilding has historically improved at low single-digit percentages annually. Is there any independent verification, or are we taking their PR at face value?
• 2:08
  [Guest] Fair challenge. HII is a publicly traded company, so throughput metrics feed into earnings reports that get scrutinized by analysts and auditors. But you're right that the 14 percent figure isn't independently benchmarked against other yards.
• 2:19
  [Host] So how do we check it against reality?
• 2:21
  [Guest] The Navy tracks delivery milestones, and HII's submarine delivery timelines have been public record through congressional testimony. If they were inflating throughput claims while still missing delivery windows, that gap would show up fast. So far, the numbers are directionally consistent with what the Navy is reporting to Congress.
• 2:36
  [Host] Fair enough. But even if the number is real, compounding 14 and 15 percent over two years gets you nearly 30 percent throughput improvement. For a shipyard building nuclear-powered warships, that would be unprecedented. I'm skeptical that robots grinding hulls account for all of that.
• 2:49
  [Guest] They don't. HII has also invested in digital shipbuilding tools, workforce scheduling optimization, and facility upgrades. The GrayMatter partnership is one piece. But it's the piece that signals where the next decade of gains will come from, because the other improvements eventually plateau. Automation doesn't.
• 3:02
  [Host] Who else is watching this? If HII gets a real throughput advantage from robots, that changes the competitive equation for every defense shipyard in the country.
• 3:10
  [Guest] Every defense prime. General Dynamics runs the Electric Boat shipyard in Groton, Connecticut, which co-produces Virginia-class subs with HII. If HII gains a throughput advantage through robotics, Electric Boat faces pressure to follow. And beyond defense, commercial shipbuilders in South Korea and Japan have been integrating welding automation for years. The U.S. has been behind on this.
• 3:27
  [Host] South Korea and Japan didn't wait for a labor crisis to automate their yards.
• 3:31
  [Guest] They had their own demographic pressures pushing them earlier. Japan's shipbuilding workforce has been shrinking for two decades. But the U.S. situation is fundamentally different because American defense shipyards are constrained by security requirements. You can't outsource submarine construction to Hyundai.
• 3:43
  [Host] So when the domestic labor pool dries up, there's no overseas backup plan.
• 3:47
  [Guest] None. The work has to happen domestically, with cleared workers, in secure facilities. You can't import the solution. You have to build it.
• 3:54
  [Host] And you can't exactly buy shipyard robots from Chinese manufacturers either. Not when they're going inside a facility with classified nuclear propulsion systems.
• 4:01
  [Guest] That's the subtext. TechCrunch reported in February that Chinese firms controlled roughly 90 percent of global humanoid robot shipments in 2025. For commercial applications, that market dominance is a trade concern. For defense manufacturing, it's a non-starter.
• 4:11
  [Host] So the GrayMatter deal is partly a supply chain play.
• 4:14
  [Guest] Absolutely. HII partnering with GrayMatter, a California-based company, is a supply chain decision as much as an operational one. The robots that build American warships need to be American robots running American software.
• 4:24
  [Host] So what happens to the shipyard workers? HII has framed this as augmenting, not replacing. And I'll be honest, I bought that framing for the first half of this conversation. But then we described a self-reinforcing loop where automation removes the incentive to train replacements, and now I'm not sure the "augmenting" story holds for more than about five years. Which is it?
• 4:43
  [Guest] Both, on different timescales. Right now, the Bureau of Labor Statistics shows welding employment has been relatively flat nationally while demand from defense, infrastructure, and energy has climbed. HII isn't cutting welders to save money today. They're deploying robots to fill shifts they cannot staff. The existing workers are being repositioned toward tasks the robots can't handle yet.
• 5:00
  [Host] "Yet" is doing some heavy lifting in that sentence.
• 5:03
  [Guest] It always does. And here's where I'd actually push back on your self-reinforcing loop framing. You're describing it as if the robots broke the apprenticeship pipeline. But the pipeline was already broken.
• 5:12
  [Host] Broken how? Welding is still a recognized trade with real training programs.
• 5:16
  [Guest] The American Welding Society has reported that the average age of a skilled welder in U.S. shipbuilding is north of fifty. Trade school enrollment in welding has been declining since the early 2000s, long before anyone deployed an AI-powered grinding robot. The robots didn't kill the pipeline. They arrived at the funeral.
• 5:31
  [Host] So your argument is that automation is filling a hole that was never going to be filled by humans regardless.
• 5:37
  [Guest] For this specific workforce, in this specific decade, yes. The question isn't whether robots replaced welders. It's whether anyone was going to show up to replace the welders who retired. The answer, based on twenty years of enrollment data, is no.
• 5:49
  [Host] But once the robots are in place, does anyone bother training human replacements? That's the part that worries me.
• 5:54
  [Guest] That's the real risk. If HII can meet throughput targets with fewer human welders supplemented by robots, the economic case for investing in five-year apprenticeship programs weakens. You could end up in a cycle where the shortage that justified automation becomes permanent because automation removed the last reason to fix it.
• 6:09
  [Host] So we agree on the destination. We just disagree on what caused the car to start rolling.
• 6:15
  [Guest] Right. And that distinction matters for policy. If you think robots broke the pipeline, you regulate the robots. If you think the pipeline was already broken and robots are a response, you ask a different question: what happens to the communities that depended on those jobs, and are we building anything to replace that economic function?
• 6:31
  [Host] A research paper from the National Bureau of Economic Research, "Robots and Jobs: Evidence from US Labor Markets" by Acemoglu and Restrepo, found that each additional industrial robot per thousand workers reduced the employment-to-population ratio by 0.2 percentage points in affected commuting zones. The displacement effect was concentrated in manufacturing, and the workers most affected were those without college degrees doing routine manual tasks. That's a precise description of a shipyard welder.
• 6:52
  [Guest] And it maps directly onto what we're watching at HII. The Acemoglu and Restrepo findings showed that new job creation in other sectors didn't fully offset the losses in robot-adopting industries. The net effect was negative for the specific communities where automation landed.
• 7:05
  [Host] And Newport News is exactly that kind of community.
• 7:08
  [Guest] It is. HII's shipyard there employs over 25,000 people. If the ratio of humans to robots shifts meaningfully over the next decade, the local economic impact extends well beyond the gates of the shipyard.
• 7:18
  [Host] Patrick Day-glon calls this the labor cliff in Unscarcity, and the labor cliff article on unscarcity dot ai tracks this exact pattern: the moment when workforce attrition meets automation capability, and the old equilibrium never comes back. Once machines fill the gap, who builds the bridge for the workers on the other side?
• 7:33
  [Guest] And at a shipyard building nuclear submarines, that question carries national security weight that most industries don't have to reckon with.
• 7:39
  [Host] For more on the labor cliff, defense automation, and what comes next, check out the articles and forum at unscarcity dot ai. The book is on Amazon, and the audiobook is on Spotify. This has been Minds, Bodies, and Terawatts.