Marine veteran Chris Albert works as the general manager of Fortress Marine Anchors in Florida. With only four machinists, three warehouse associates and a reputation for quality, the shop churns out lightweight aluminum fluke anchors for customers who run the gamut from recreational boaters to the US Navy. He says a “lean manufacturing” philosophy is foundational to the business.
Albert is no stranger to the “physical AI” buzz that has hit a fever pitch in the past six months — with believers prophesying an ultra-productive future where AI-operated factories and agentic robot workers enable the reshoring of American manufacturing. Nvidia’s Jensen Huang, who popularized the buzzword, claimed that soon “every industrial company will become a robotics company.”
For now, Albert doesn’t buy it.
“As far as the shop floor, particularly our shop floor, I don’t see robots walking around here,” he said. “Part of our ethos is not wasting time or money.”
While manufacturers are generally optimistic about the productivity boosts physical AI could bring, many see the more extreme futuristic visions as removed from reality — and say a rollout of new tech to the smaller firms that make up the bulk of the industry won’t be as simple as a software update.
“With this much capital flowing into the space, I think by definition we will see an impact,” said Kyle Crum, president of Wisconsin-based precision manufacturing firm Infinity EDM. “Will it be an industrial revolution? I don’t know.”
The fourth industrial revolution?
“Physical AI” is an umbrella term for the marriage of AI and hardware to create products that interpret and act on the physical world, just as LLMs can navigate all that’s digital. The goal is technology that can identify and solve problems, unscripted, in a closed-loop format — adding capabilities to the existing “smart factory” model, which already deploys automation, robotics, and integrated layers of sensors and software.
Venture capital has poured into the space in recent years, and tech giants are throwing their hats in the ring: Jeff Bezos recently raised $100 billion to buy up and automate factories with AI, and Elon Musk has made humanoid robot development Tesla’s new primary goal.
In many ways, the conditions are ripe for Silicon Valley’s mission. Unlike white collar America, manufacturing struggles with a labor shortage. China has added existential fuel to the fire with aggressive progress on robotics and automation across their entire manufacturing ecosystem. And with Trump in office, an appetite for onshoring buoys the landscape.
But for small and mid-sized firms, upfront cost will pose a barrier.
New technology has never been adopted uniformly across the sector, said Brett Conner, chief manufacturing officer at the Society of Manufacturing Engineers.
“The small and medium businesses are always lagging,” Conner said. Those businesses are also foundational to tiered supply chains, he explained, often supplying materials and parts to larger companies.
Policies to boost “mom and pop” manufacturers’ access to capital for upgrades will be critical, according to Groundwork Collaborative’s Alex Jacquez, who previously served as President Biden’s industrial strategy advisor. Otherwise those businesses will lose out as prices from increasingly-automated overseas manufacturers drop.
Solutions and problems
Price isn’t the only issue: manufacturers want certainty that any change they’re implementing will improve their bottom line, not cause more problems.
“I live in the real world of pragmatism. So, you know, I actually have to do stuff,” Crum said. “How do I serve my customer, at pace, with quality product at a reasonable price?”
Crum says he’s seen an often-mismatched approach between the industry and Silicon Valley, with developers promising “solutions” disconnected from the problems manufacturers have.
Complex, general-purpose humanoid robots —still a work in progress, but lavished with tech-world funding and excitement — embody this cultural difference. They’re often advertised for use in factories, but manufacturers generally aren’t convinced, preferring robots purpose-built to address specific tasks in the production process.
“There has to be this melding of the coasts with the industrial Midwest,” Crum said, explaining that the tech world’s “bias for change and action” and manufacturers’ tendency for pragmatism should be balanced.
That melding is arguably in progress, if you peer through the hype. Although Nvidia has a penchant for humanoids, it has rolled out partnerships with the “big four” in manufacturing robotics — FANUC, ABB, KUKA and Yaskawa.
Some startups, too, are designing physical AI products that slot more directly into manufacturer’s workflows. Trener Robotics is one: the company has developed an AI software layer, compatible with the big four robots, that automates much of a machinist’s workload. The product allows robots to understand tasks in plain English rather than code.
“The pull that we feel from the market is immense,” CEO Asad Tirmizi said, explaining that his customers include small and mid-sized metal fabricators.
Path Robotics is another — the startup boasts continuously learning AI-enabled welding machines, which can see and adapt to variations across parts. Huntington Ingalls Industries, the largest military shipbuilder in the US, is among their newest partners.
“There’s a certain set of technologies that are really good at doing one task 10,000 times,” an HII representative said at a press conference. Path’s AI welding, he said, “will enable us to get at those use cases where I need 10,000 things done just once.”
What about workers?
Physical AI companies say that they’ll automate away “dull, dirty and dangerous” work. Depending on your definition, that could be the majority of jobs in the sector today.
Economists and industry leaders argue that the extent of the labor shortage means anyone who wants a job in manufacturing can likely keep one — though it may involve overseeing a fleet of robots. The boosts to productivity would see factories expand and multiply, they argue, creating enough of those roles to accommodate the current workforce.
But that will be a narrow tightrope to walk, according to Jacquez. In the near-term, he said, policies that actually guarantee upskilling programs for workers will be crucial. If physical AI reaches its full potential, the conversation will need to shift.
“That future is certainly envisioned — one where you can replace workers one-to-one,” he said. “How we take care of those workers who lose their jobs is really important.”