American manufacturers are rebuilding production while making entry harder for the workers they need.

At Swift Coat, a seven-person Phoenix startup that makes antireflective coated glass for satellites, three shop-floor technicians inspect samples. One has an associate degree. Another spent time in college. CEO Peter Firth says experienced operators remain necessary even when the equipment is functioning perfectly. Their work is to interpret sensor and data readings, watch for deviations, prevent faults, and diagnose what robotic handling cannot explain by itself.

That is the technical distinction the employment figures often blur. Formal credentials and tacit experience are not substitutes. A degree signals that someone has completed a prescribed course of study; it does not certify that the person can recognise a failing line before it fails, infer what an abnormal reading means, or diagnose a fault that the manual did not anticipate. Swift Coat values both. So, in a less precise account, do many other manufacturers.

Swift Coat plans to hire two or three workers by year’s end, and every open position requires an engineering degree. Its seven-person full-time staff includes a physics Ph.D., two people with master’s degrees in mechanical engineering and one with a bachelor’s degree in chemical engineering.

The company is installing a highly automated assembly line in which robots will load and unload samples while one or two workers monitor the process. It might hire one technician and two engineers instead of five technicians. “We’re not necessarily spending less on people,” Firth said. The statement does not say payroll will fall; it says the work will be distributed differently among fewer bodies with different skills.

Swift Coat’s degree requirement is a company rule, not a rule of the sector. Experienced technicians without engineering degrees remain valuable there, and elsewhere. What the company demonstrates more generally is that the new factory can require formal credentials at the entrance while depending upon tacit knowledge once workers are inside. The two are compatible. They are not the same thing.

American manufacturing is hiring again. Head count has increased by 72,000 since December after years of declining factory employment. That sounds like a revival, and in one sense it is. Between 2015 and 2025, however, manufacturers shed 244,000 production-occupation jobs, including assemblers and machine operators, while adding 358,000 jobs in management, business and financial operations. The data do not prove that a particular robot caused a particular layoff, but they do establish a profound change in the occupational composition of factory work.

It is true that higher-skill manufacturing jobs tend to pay better, create greater profits for employers and support economic growth. Binil Starly, director of Arizona State University’s School of Manufacturing Systems and Networks, puts the change plainly: “The kind of manufacturing that is returning back to the U.S. is not the same as what we used to have in the ’70s or ’80s or ’90s.” The trouble is that a better job and an accessible route to that job are separate facts. One may benefit the worker who gets the position while closing the doorway behind everyone else.

Susan Helper, an economist at Case Western Reserve University’s Weatherhead School of Management, put the concern more directly: “What I worry about is the access to good jobs.” The era when a person could leave high school, enter a factory and build a career through experience is increasingly over. The operative word is increasingly. The old route is becoming increasingly inaccessible, not universally closed, and experienced workers still enter and advance through it. The point is that access is becoming less automatic, more dependent upon credentials and training that the prospective worker may have to find and finance before an employer will give them the chance to learn.

The credential arrives before the opening.

The divergence is visible across industries. Older sectors such as furniture manufacturing are shrinking, while knowledge-intensive sectors such as aerospace, computers, electrical equipment and heating and cooling systems are adding workers. More than half of workers in airplane manufacturing have at least a bachelor’s degree, according to the Census Bureau, and semiconductor manufacturers are more likely than other manufacturers to hire college graduates. These figures show a changing job mix; they do not prove that every factory now excludes workers without degrees. They do show why the label “manufacturing job” can conceal entirely different economic bargains.

Jeannine Kunz, CEO of SME, described the demand from the other side: “Technology is now outpacing our people. There’s not a manufacturing sector that’s not starving for talent.” Her claim concerns the shortage of talent, not proof that every person without a degree has had a reasonable opportunity to become qualified. A shortage and an inaccessible training pipeline can exist at the same time. In fact, that is precisely when the shortage becomes more acute and more convenient for employers to describe as inevitable.

Companies are building pipelines, and the work is real. SME’s workforce-development division employs around 125 people and sends equipment such as drone-coding kits and 3-D printers to middle schools to spark interest in factory work. Arizona State University is training around 170 people as equipment technicians for TSMC’s Arizona plant and plans to increase that number to 800 next year. It also has around 370 students pursuing bachelor’s, master’s and doctoral degrees.

Bigger commitments are following. Big business and philanthropies are pouring $400 million into skilled-trades training. The money is necessary because the labour market is not an abstraction: manufacturing pipelines are physical systems, built plant by plant and trade by trade.

Yet training and investment do not run on the same clock by default. Technology moves at the speed of capital investment. Training programmes move at the speed of committee meetings. That is not an argument against the programmes; it is an argument for starting them before the machinery is installed and waiting for a shortage to supply the urgency.

Intel has announced two factories east of Columbus with a combined 3,000 workers. Aspyr Workforce Innovation CEO Lisa Patt-McDaniel said local programmes can keep up with current demand, but added: “Call me in two years and we’ll see where we’re at.” Against a U.S. labor force facing historic contraction as demographic pressure raises the stakes for AI adoption, two years is not a distant footnote. Her answer is better evidence than a confident prediction: present capacity appears adequate; future capacity remains uncertain.

The answer cannot be merely to wait for universities to produce more graduates. The documented skills gap supports a more deliberate set of responses: paid apprenticeships that combine workplace instruction with earned credit; stackable technical credentials that allow workers to accumulate qualifications instead of pursuing one all-or-nothing degree; greater community-college capacity; training for incumbent workers whose skills need updating; and pathways that recognise practical experience alongside formal education. Employers receiving public incentives should be required to build the labour pipeline they claim to need. Smaller manufacturers need shared training infrastructure rather than being expected to recruit workers from one another.

These are evidence-based responses to documented gaps, not remedies already demonstrated to work at national scale. That distinction matters. It is also why the 72,000-job increase should be celebrated without pretending that a good job becomes broadly shared merely because the position exists.

None of this requires pretending that the manufacturing recovery is bad. The increase is evidence that American production can grow again. It is also evidence that the bargain is changing: fewer factory jobs now offer the on-the-job route into the skilled trades, while employers ask for more credentials, more specialised knowledge and more capacity to diagnose machines that their own systems cannot fully explain.

Robots can load and unload the samples. Engineers can design the line. Skilled technicians can keep it from breaking. Communities need the chance to train for those roles before the jobs arrive.

A recovery that cannot admit the people needed to sustain it is not a full recovery; it is a gated community with an efficient production line. The factory jobs came back. The ladder got higher.