The Robot Skills Gap Isn't About Headcount. It's About Portability.

Picture the version of this story that plays out on factory floors across Southeast Asia every year, not as one company's case study but as the pattern itself. A plant buys a collaborative arm. A technician spends two weeks in a certified course, passes an exam, and becomes genuinely good at keeping that one robot working. The plant expands and buys a different brand for the new line, because that vendor had the better lead time, or the better price, or the better local distributor. The certification the technician already holds is worth nothing on the new machine. He starts over.

That is not a hypothetical. It is the default condition of industrial automation right now, and it is the part of the "robot skills gap" story that keeps getting skipped.

The headlines say there are not enough trained people. The more accurate statement is that there is no such thing as a trained person, only a person trained on one vendor's equipment, and that distinction is where the real shortage lives.

The scale problem everyone already agrees on

Start with the number that is not in dispute. The International Federation of Robotics reported in September that the world's factories installed more than 600,000 new industrial robots in 2025, up 11 percent, pushing the global operational stock to a record 5 million units. China alone installed 354,000 of them, 59 percent of global deployment and up 20 percent year over year. That is the fastest-growing slice of the installed base anywhere, concentrated in exactly the region, Asia, where this column's own beat already shows training infrastructure struggling to keep pace with procurement.

On the demand-for-people side, a 2024 study from Deloitte and the Manufacturing Institute projected that the United States alone would need as many as 3.8 million new manufacturing workers by 2033, with up to 1.9 million of those roles at risk of going unfilled if the skills pipeline does not change. Line up the two data sets and the conventional read writes itself: robots are multiplying faster than the workforce that can run them, so the fix is more training, more apprenticeships, more classroom seats.

That read is not wrong. It is just incomplete in a way that matters for anyone actually buying robots.

Two training pipelines, two different workers

FANUC's education arm, CERT, now reaches more than 1,700 schools across the Americas, issuing NOCTI-backed credentials at two levels each for operators and technicians. It is a serious, well-built program. It also only teaches FANUC equipment. Universal Robots runs a parallel system: more than 200,000 registered users across 130 countries, roughly 120 certified training centers, free e-learning plus paid in-person classes. Also serious. Also only transferable to UR cobots.

Put those two numbers next to each other and the shape of the problem comes into focus. These are not two on-ramps into one profession. They are two separate professions that happen to look alike from across the shop floor, each gated by a credential the other vendor does not recognize and has no reason to. ABB runs its own university. Yaskawa runs its own academy. Every major arm maker has built a parallel pipeline, and none of them connect.

A journeyman electrician's license travels. A robot technician's certification does not, and that is a design choice as much as it is an educational gap. Training a buyer's workforce for free, or near enough to free, is one of the most effective retention tools a hardware vendor has: the more a customer's technicians know only your interface, your fault codes, your teach pendant, the more expensive it becomes for that customer to ever switch brands.

Call it what it is. A certification program is also a switching cost, wearing a safety vest.

The industry already noticed, quietly

Here is the detail that confirms this is not a coincidence of market structure. A3, the trade association that represents most of these same vendors, launched its first vendor-neutral training product only this spring: a two-and-a-half-hour, on-demand "Introduction to Industrial Robotics" course, explicitly built to not favor any single brand's equipment. A trade group does not build a vendor-neutral product unless its members have already concluded, on their own, that the brand-specific pipelines were never going to solve the labor shortfall by themselves.

A two-and-a-half-hour primer is not a counterweight to a 1,700-school certification network. It is an acknowledgment, arriving late and sized small, that the industry's own training model has a hole in it that no single vendor is incentivized to fill. Nobody is lying about the skills gap. They are each solving a narrower version of it than the one they describe in press releases.

What a vendor-neutral public system can and cannot fix

I read procurement decisions from a desk in the Philippines, where the public vocational track runs in the opposite direction from the vendor academies. TESDA's Mechatronics Servicing qualification, levels two through four, teaches PLC programming, motion control, and HMI troubleshooting as general skills, deliberately not tied to one brand of hardware. TESDA is also building a dedicated Regional Training and Innovation Center in Taguig specifically for mechatronics and industrial automation, a real, under-construction answer to exactly the gap this column is describing.

It is the right instinct, and it will not by itself close the gap that matters to a buyer evaluating vendors this year. A technician who finishes a general mechatronics qualification still cannot walk onto a factory floor and safely run a specific robot cell without the manufacturer's own credential, because the manufacturer, not the state, controls the interface. Public vocational training builds the foundation. It does not grant the keys. A buyer who assumes a generically trained local labor pool solves the maintenance question for whichever brand lands on the floor is solving the wrong half of the problem, and will find that out during the first unplanned downtime, not before.

What the taxonomy on my own desk already shows

Running a robotics data platform means spending most days reconciling which company made which machine, under which brand, with which control software, because almost nothing in this industry assumes you already know. The database behind this site currently tracks 1,662 companies and more than 8,100 robots, and the single hardest part of maintaining it has never been finding new robots. It is mapping which of those companies' machines share anything at the interface level with any other company's machines. The honest answer, most of the time, is nothing. If a data platform built specifically to track interoperability keeps finding none, a factory floor trying to staff around it should not expect to find more.

None of this means the vendor academies are a bad idea, or that FANUC and Universal Robots should be faulted for training their own customers well. A company has every reason to make its own product easiest to run. The failure sits one level up, with an industry that keeps describing a structural, deliberate fragmentation as an accidental, temporary shortage, and with buyers who keep budgeting for "training" as a line item instead of asking, before signing, which brand that training locks them into.

The number worth watching from here is not another installation count. It is whether A3's vendor-neutral course gets built out into something an employer can actually hire against, a credential that moves between shop floors the way the robots increasingly do not, or whether it stays a two-and-a-half-hour gesture while the real pipelines stay exactly as closed as they are today.

This analysis reflects the author's own reading of the International Federation of Robotics' September 24, 2026 World Robotics press release, the 2024 Deloitte and Manufacturing Institute workforce study, FANUC America's and Universal Robots' own published training-program figures, the Association for Advancing Automation's 2026 course announcement, public information on TESDA's Mechatronics Servicing qualification and its Taguig training center, and the author's own operational data from maintaining RobotAIGeek's company and robot database. It is provided for general information purposes only and does not constitute investment, procurement, or professional advice.

Hero image: a student training on a FANUC CRX-10iA/L collaborative robot through FANUC's CERT education program. Photo: FANUC America.