Toyota has put a number on how much automation its factories could need. Its estimate is about 400,000 robots from 2028, at roughly 1 trillion yen, or $6.4 billion, a year. The figure covers Toyota, its group companies and its suppliers, and Toyota has not committed to spending that much.
The expensive part of a factory robot was the teaching
A conventional factory robot is cheap to run and costly to set up. Someone has to program every movement, and that work was only worth paying for when the same part was made in very large numbers. So robots filled the welding and painting lines of car plants and stayed away from small jobs, like a worker tipping parts from a bin into a box.
Toyota's research arm has been building robots that skip the programming. The robot watches a person do the task, and a few minutes of demonstration can teach it a skill such as pouring a liquid or using a tool. Toyota's own humanoid, Eley, rolls on wheels, weighs 50 kg, and is aimed at jobs like moving parts between boxes. Teaching one more small task now costs a demonstration instead of a programming project, and work that never justified a robot starts to.
The suppliers are where this changes the most
About 250,000 of the robots are earmarked for group companies and their suppliers. These are mostly smaller firms that make many different parts in short runs, which is exactly the work the old economics shut out.
Picture a supplier that makes brackets in batches of a few hundred and changes to a different bracket every couple of days. Reprogramming a robot at every change never made sense. A shift lead who shows the robot the new box layout at the start of the shift does make sense. Each robot then takes over one person's repetitive handling on a shift, in an industry that is short of workers for exactly that kind of job.
Toyota's own plants come first, because they are easier to control. Its research arm has already spun out Walden Robotics, which launched in July with $300 million to put the same kind of learning robots to work in other industries. The supplier is the second step, and a warehouse or packing room that moves boxes all day is a short step after that.
The person who teaches becomes the asset
If a robot learns from demonstrations, its quality depends on the quality of those demonstrations. The operator who loads a box fastest, with the fewest dropped parts, is no longer only doing a job. That person's way of doing it gets copied onto every robot the company owns.
That raises a question factories have not had to answer. Tesla's workers have already balked at training robots that could replace them, and a company that wants good demonstrations will have to give its best people a reason to give them.