The installed robot base, not the humanoid, will remodel the factory first

The humanoid robot has become the marketing face of factory automation, but the machines actually doing the work are fixed arms bolted to the floor ^[1]. That gap matters because the realistic route to a more flexible factory over the next few years runs through improving equipment manufacturers already own, not deploying human-shaped machines.

The numbers set the terms. According to the International Federation of Robotics, factories installed 542,000 industrial robots in 2024, and the number operating worldwide reached roughly 4.66 million ^[2][3]. Annual installations have topped 500,000 for four consecutive years ^[4], which means at least two million units, or more than 40 percent of the running base, went in over that span ^[5]. Last year's installations alone equal about 12 percent of everything in service ^[6].

That scale is not just hardware. It is an ecosystem of integrators, software vendors, component suppliers, safety specialists and maintenance staff, plus a costed understanding of the return on an arm used for welding, painting, palletizing or assembly ^[7]. A fixed arm also has a mechanical advantage: it does not walk, balance or understand the whole plant, it repeats a defined motion quickly and accurately thousands of times a day ^[8]. When a task rarely changes, specialization beats versatility ^[9].

The one argument humanoids had left, adaptability, is also eroding. Collaborative robots, which work closer to people and can be reprogrammed, made up 10.5 percent of industrial robot installations in 2023 ^[10].

Where humanoids fit is the work that does not justify rebuilding a line. Plants are full of stairs, carts, shelves and workstations built for the human body, and a humanoid can in principle enter them without new conveyors or cages ^[11]. The early deployments are narrower than that pitch. Mercedes-Benz is testing Apptronik's Apollo for moving components and initial quality checks, and is starting with repetitive intralogistics rather than core assembly ^[12]. GXO signed a multiyear deal with Agility Robotics to run its Digit robot, which at a Spanx facility moves totes from mobile robots onto conveyors ^[13]. These are controlled, repetitive material-handling jobs, not roaming machines switching between dozens of operations ^[14].

Economics, not resemblance, decides this. Legs consume energy and add fall risk, dexterous hands add complexity, and the autonomy has to stay dependable across a shift ^[15]. Agility Robotics said in 2025 that upgrades extended Digit's runtime to up to four hours, which still implies charging, battery swaps or extra robots for continuous work ^[16].

What to watch: whether humanoid runtime and reliability improve enough to hold a station against a cobot on cost per task, and whether the pilots at Mercedes-Benz and GXO move from intralogistics into work that a fixed arm cannot already do more cheaply.