Back in 2004, Toyota showed off a robot that could play the trumpet. Cute, but it didn’t build any cars. Now the company’s back with a far bigger plan. According to Financial Times reporting picked up by several outlets on Monday, 26 October, Toyota wants to put 400,000 robots to work, and Toyota humanoid robots are part of the mix. So is this a real shift or another shiny promise?
So what did Toyota actually say?
Here’s the plan in numbers. Toyota could spend about $6.4 billion a year on factory automation, starting in 2028. That money would cover roughly 400,000 robots across Toyota’s own plants, its group companies and its major suppliers. Group companies and suppliers would get about 250,000 of them.
Two quick reality checks, though. First, this isn’t 400,000 humanoids. The figure covers robots of all kinds, from classic industrial robots to machines that move materials around, and it includes replacements for old equipment. Humanoids are one part of the plan.
Second, Toyota hasn’t actually committed to spending the money. It first discussed the estimate with investors in September, as Assembly Magazine reported at the time. It also hasn’t said how many years the spending would last.
So why’s everyone talking about it now? Because the new reporting fills in the strategy behind the number. According to Humanoid.guide’s summary of the FT report, Toyota will put a dedicated robotics team under its chief production officer. It’s also working on its own robot brains, which we’ll get to in a minute.
Meet Eley, the robot that learns by watching
The star of the show is Eley, short for Embodied Learning Robot for Enhanced Yield. The Toyota Research Institute developed it. It weighs 50 kg, rolls on wheels and has two-fingered hands. You can run it on batteries or plug-in power.
The clever part is how it learns. Instead of needing someone to code every move, Eley can pick up a physical skill from a few minutes of watching a human. It uses diffusion-based AI, which is the same broad family of tech behind many AI image generators.
And Toyota has a secret weapon here: its takumi. These are 18,000 veteran master craftspeople. While they do routine jobs like handling interior trim and peeling off protective film, they wear finger-shaped jigs modelled on the robot’s hands. That way, the robot learns from the best hands in the building.
In one demo at Toyota’s European headquarters in Brussels, a robot learned to fold T-shirts after 1,500 practice sessions over two weeks, as autoevolution reported. Then, once one robot masters a task, Toyota shares that know-how with its robots worldwide. Learn once, use everywhere. Honestly, that’s the most interesting idea in the whole plan.
Why Toyota humanoid robots start at home
Toyota isn’t rushing to sell robots to other companies. Tomohiro Nomi, who heads its humanoid robotics research, said Toyota is “giving the highest priority to deploying robots in our own factories.”
That makes sense. A Toyota executive explained that its own plants are easier to control, so they’re a safer place to start. You know the layout, the parts and the people. Robots tend to struggle with surprises, so you start where there are fewer of them.
Then there’s the labour problem. Japan’s population is ageing, and robots could help offset worker shortages. Toyota’s Executive Vice President Hiroki Nakajima put it this way: “We aim for a world where robots coexist with humans, rather than replacing them.”
Toyota also isn’t starting from scratch. Its AI robotics spinout, Walden Robotics, came out of stealth in July with $300 million in funding at a $1.1 billion valuation. Its general-purpose robots already work in Toyota factories in North America.
Tesla, for comparison, already uses early Optimus robots to sort and load heavy battery cell packs. The difference is that Toyota is talking about its suppliers as well, which could spread its robots much further.
The software is the real prize
Here’s where it gets interesting. The Toyota Research Institute in California has developed a foundation model, a big general-purpose AI, that turns high-level decisions into physical actions and real-time motor control. According to the FT, Boston Dynamics’ Atlas humanoid uses that model. So some of the smarts in that famous robot come from Toyota.
That’s why analysts aren’t obsessing over hardware. Bernstein’s Masahiro Akita told the FT that Toyota’s edge may lie in combining “robots, AI, and production processes into a superior operating system.” Toyota also has about two decades of data to draw on, which Semafor framed as an edge over Chinese rivals. Those upstarts are strong on hardware, yet they’re behind Western companies on robot AI, according to the same report.
One investor put it bluntly: “Whoever controls the software layer will control global manufacturing.” That’s a big claim. But you can see the logic. If plenty of companies can build the body, the brain is what sets you apart.
Another investor told Semafor that “Toyota is developing its own large behavioural models.” In other words, Toyota wants AI that’s trained on how bodies move and work, rather than on text. Our guess? That’s where the real fight happens.
Okay, are Toyota humanoid robots for real?
Maybe. But a little caution won’t hurt. Some investors remember Toyota’s past bets on hydrogen cars and solid-state batteries, where it fell short of expectations. Others doubt robotics can become a major business for a company that makes around 10 million cars a year.
On the other side, Goldman Sachs thinks Toyota could take 3% to 8% of the humanoid market by 2035. That would mean somewhere between 190,000 and 540,000 units. Those are analyst projections, though, not Toyota promises.
Meanwhile, the competition isn’t waiting. Tesla, Xpeng and Hyundai are all heading the same way, and Hyundai plans to start building Atlas in the US in 2028. Chinese firms, as we covered in our look at Chinese humanoid robots, keep pushing hard too.
So what does it mean for you? If you work in manufacturing, the takumi idea is worth noticing. Toyota is treating its veterans’ skill as training data, which makes experience more valuable, not less. For engineers in Armenia and other emerging markets, we’d bet the openings sit in software, sensors and the unglamorous integration work that makes robots useful on a real line.
Here’s our honest take. Toyota humanoid robots won’t flood factories next year. But if Toyota gets this learning loop right, it could end up with some of the best-trained robots in the business. Would you rather train a robot teammate or compete with one?