← Back to Field Notes

April 10, 2026 · The Bloomfield Team

Physical AI Is Here. American Shipyards Are Running the First Test.

In April 2026, Huntington Ingalls Industries, the largest naval shipbuilder in the United States, signed a memorandum of understanding with GrayMatter Robotics to deploy physical AI across its shipbuilding operations. The partnership targets autonomous surface preparation and automated coating inspection as part of HII’s High-Yield Production Robotics initiative.

The numbers in the announcement are specific. HII achieved a 14% throughput increase in 2025. The target for 2026 is an additional 15%. GrayMatter claims its systems deliver up to 12 times the output of skilled manual workers, with a 95% reduction in rework rates.

The proposed FY 2027 defense budget allocates $65.8 billion for shipbuilding, the highest figure since 1962. Eric Chewning, HII’s executive vice president, described the goal in plain terms: “By working with new partners like GMR, we can further augment our workforce and speed up U.S. Navy shipbuilding production.”

What Physical AI Actually Is

Physical AI is not software deployed on a screen. It is robotics guided by AI perception, able to read and adapt to real-world conditions in ways that traditional automation cannot.

A standard welding robot follows a programmed path. A physical AI system reads the actual workpiece, identifies variation in geometry or surface condition, and adjusts. In shipbuilding, that matters because no two surfaces are identical. Corrosion patterns vary. Welds vary. The geometry of a hull section is not the same from one ship to the next. Traditional automation handles repetitive, identical work. Physical AI handles variable work that has historically required skilled human judgment.

Ariyan Kabir, GrayMatter’s CEO, described the pressure driving adoption: “We have to build the essential components for our war fighters, and we have to do this very quickly.” Speed is the constraint. Physical AI is one answer to it.

The Adoption Pattern That Follows a Test This Scale

The HII deployment is about shipbuilding, but the technology does not stay in shipyards. Manufacturers who have been watching physical AI from a distance are watching the test run now, at one of the most demanding production environments in the world, under the scrutiny of a $65.8 billion budget commitment.

In 2015, automated inspection systems were primarily used in aerospace. By 2022, they were standard in medical device manufacturing, automotive, and precision machining. The lead time from proof of concept at scale to industry standard has compressed significantly over the past decade.

The 12x output figure and 95% rework reduction are GrayMatter’s claims, stated under a partnership agreement with a company whose throughput targets are public. The numbers will be tested in production. If they hold at HII’s scale, the adoption timeline for adjacent industries will move faster than most manufacturers expect.

The Question for Your Operation

The shop floor question is not whether physical AI will reach your sector. It is when, and whether you are building the operational infrastructure to absorb it when it arrives.

Physical AI systems generate data: workpiece measurements, surface condition readings, process parameters, rework events. That data has value only if the operation has a system to capture, structure, and act on it. The manufacturers who build the information layer now will be able to integrate physical AI into their existing operations when the technology reaches them. The ones who wait will be building two things at once.

HII is running the test. The results will be public. American manufacturing has always moved fastest when the proof of concept is visible. Pay attention to this one.