Cobots: the robots already working beside people, with no cage
While everyone argues about whether a humanoid can fold a T-shirt, there is a category of robot that has spent years moving into real factories and workshops, with no cage, no fencing, and a person standing a metre and a half away. They are called cobots, collaborative robots, and they are by a wide margin the part of robotics that has changed most in the last decade. They also generate the fewest headlines, precisely because they work.
What a cobot actually is
A classic industrial robot sits inside a metal cage for a simple reason: it moves fast, with a lot of force, and it does not know you are there. Walk in while it is working and it will hurt you. A cobot is designed the other way round: limited power and speed, rounded edges, force sensing in every joint, and control logic that stops the machine the moment it detects an unexpected contact. That is what lets you remove the cage and put the robot literally on the workbench, next to the operator.
The number that places the category
The International Federation of Robotics, the serious source here, put collaborative robots at 10.5% of the 541,302 industrial robots installed in 2023. One in ten. For a category that barely existed commercially fifteen years ago, that share is enormous, and it keeps growing faster than the market as a whole. Next time you read that industrial robotics has stalled, remember this is the segment that is growing.
What they actually do
The applications the IFR identifies as leading are fairly prosaic: welding, machine tending, bin picking (pulling disordered parts out of a crate) and end-of-line palletising. None of them look good in a viral video. All of them are repetitive, physically awkward tasks that, in a small or mid-sized firm, do not justify building a full robotic cell with fencing, a PLC and bespoke engineering. The cobot fits where the classic robot did not fit, neither physically nor economically.
Why they matter more than humanoids
A cobot costs a fraction of a humanoid, is programmed by physically moving the arm to the desired position and saving the waypoints, pays for itself in months, and has had established safety standards for years. A general-purpose humanoid has none of those four things yet. If somebody wants to automate something today, in 2026, the correct answer is almost always a collaborative arm, not a bipedal machine. This is not an argument against humanoids, it is a question of timing.
Where they are being installed
Automotive and electronics remain the biggest buyers, but the interesting growth is in aerospace, consumer goods, pharmaceuticals and warehouse logistics. In that last one the cobot usually works as a fixed station, packing, labelling or preparing orders, while mobile robots bring material to it and take it away. A fixed collaborative arm plus an autonomous mobile robot is the standard automated-warehouse architecture today, and it works without a single humanoid.
What nobody tells you about cobots
Two real limitations. The first is payload: a typical cobot moves between three and twenty kilos; above that you are back to a caged industrial robot. The second is that collaborative does not mean safe by default: the standards require a risk assessment of the complete application, because a cobot holding a knife or a sharp part is still dangerous even if the arm stops on contact. The robot is collaborative; the installation has to prove that it is.
What to watch over the next two years
Convergence. The humanoids now reaching factory floors, Walker, Kepler and Fourier among them, are in fact competing for this same budget, and they will have to justify why an expensive biped does a task better than a cheap arm bolted to a table. The honest answer is that the humanoid only wins when the task requires moving through a space designed for people. At a fixed station, the cobot still wins by a mile, and will keep doing so for a good while.
