Toyota Factory Robots: A ¥1 Trillion Bet on Retiring Craft Skills
Toyota factory robots are shifting from pilot lines to a fleet counted in the hundreds of thousands. The automaker estimates that modernising production across its own plants, group companies and major suppliers could require roughly 400,000 robots, with annual spending of about ¥1 trillion, or $6.4 billion, beginning in 2028. The deployment splits into 150,000 units for Toyota's own factories and 250,000 across group facilities, spread over close to 60 sites worldwide.
Eley, Toyota's wheeled humanoid, anchors the programme. It weighs around 50 kilograms, moves on wheels rather than legs, and learns tasks by watching human workers instead of executing hand-coded routines. The method rests on Large Behavior Models, which derive physical skills from minutes of human demonstration. Workers already take part in training, wearing finger-shaped jigs that record the real movements of an assembly task.
Eley's wheeled base is a deliberate engineering trade. Wheels carry a 50-kilogram machine cheaply across flat factory floors and cut the balance-control problem that makes legged humanoids expensive to build and maintain. The same choice limits the robot to routes a wheel can take, which suits assembly halls but not the full range of plant layouts. Toyota is optimising for units that can be deployed at volume rather than for general-purpose humanoid capability.
What the Robot Number Covers
The 400,000 figure is not a count of humanoids. Toyota's estimate spans conventional industrial machinery alongside newer learning-based systems, so the headline number describes a modernisation budget as much as a robotics fleet. About 150,000 machines go to Toyota's own plants. The remaining 250,000 land with group companies and major suppliers, which stretches the programme well beyond Toyota's direct control.
That division carries commercial weight. Capital cost for most of the fleet sits with suppliers and group companies, while the production standard stays Toyota's. Toyota frames the ¥1 trillion as an estimate of what modernisation could require rather than a locked-in commitment, which leaves room for the figure to move as deployment proceeds.
Toyota also intends to develop the machines in-house, a detail that shapes who gains from the spending. A requirement of this size would normally be a landmark order for outside humanoid vendors. Keeping development internal routes that value into Toyota's own engineering organisation and its supplier network instead.
Why Toyota Factory Robots Need Retiring Craftspeople
The plan lands against a demographic squeeze. The craftspeople whose judgement decides how a panel seats or a seam closes are retiring, and much of that knowledge was never written down. Behaviour models are the attempt to move the know-how into a machine while the people who hold it are still on the line.
Tacit skill resists capture. A veteran's decision to add pressure or hold a position for a fraction of a second longer is often made without being articulated, which is exactly why finger-mounted jigs and observation-based training exist. The jig records the motion because the worker cannot fully explain it.
Whether a few minutes of demonstration reproduce judgement built over decades is the open question the 2028 timeline will settle. The people supplying the training are also the benchmark for it: their expertise is both the asset being transferred and the standard against which the transfer gets judged.
What Behaviour Models Change
Conventional industrial robots are programmed task by task. Each new motion, part or variant needs an engineer, and that cost repeats at every plant where the task appears. Behaviour models reverse the sequence: a worker performs the motion a handful of times, the model generalises, and the same skill copies to other units without anyone rewriting code.
That mechanism is what makes the target plausible. If skills transfer in minutes, the economics of Toyota factory robots scale with capital rather than engineering headcount. Without the transfer, the deployment stays a capital project with a permanent engineering tail attached. Every demonstration captured on a live line also becomes a reusable training asset, which is why a fleet of this size can be described years before it exists.
The 2028 start date sets expectations for the intervening period. A deployment on this scale needs trained policies before it needs hardware, so the near-term work is data collection on running lines rather than fleet installation. That reframes today's pilots as the asset-building phase: what Toyota accumulates now is demonstration data, and the fleet size is a projection of what that data can eventually support.
The trade-off is verification. A statistical policy that performs well on recorded demonstrations can behave differently when a part arrives misaligned, a fixture wears, or a variant appears that never entered the training set. Hand-coded robots fail predictably. Learned ones fail in ways that are harder to enumerate in advance, and for a manufacturer whose reputation rests on defect rates, that is the engineering question sitting behind the fleet size.
Two Automakers, Two Automation Models
| Dimension | Toyota | Hyundai |
|---|---|---|
| Scale | ~400,000 robots, including conventional machinery | 25,000+ Atlas humanoids across Hyundai and Kia |
| Humanoid | Eley, ~50 kg, wheeled | Atlas, all-electric |
| Spend | Up to ¥1 trillion (~$6.4bn) a year from 2028 | Facility targeting 30,000 Atlas units a year |
| First sites | Toyota plants and group facilities, ~60 sites | Robot Metaplant Application Center, Georgia |
The comparison exposes different theories of adoption. Hyundai is building dedicated capacity to produce humanoids at a target of 30,000 Atlas units a year and plans more than 25,000 of them across its Hyundai and Kia manufacturing networks, starting from its Robot Metaplant Application Center in Georgia. Toyota is spreading a much larger count across a wider estate, and a large share of that estate belongs to suppliers.
Hyundai concentrates risk in one platform and one supply chain it controls. Toyota lowers the cost per site but depends on group companies and suppliers hitting the same standard. Neither route is obviously cheaper. Toyota's number is larger; Hyundai's is more vertically contained.
A published requirement of 400,000 units also works as a planning signal. Automation vendors, component suppliers and rival manufacturers now have a named benchmark to price and schedule against, which tends to pull forward investment across the sector regardless of how much of Toyota's plan is ultimately built.
Does Alongside Mean Replacement?
Toyota states that the machines will work alongside people rather than take their jobs, and it has repeated that framing as workers begin training Eley units on live tasks. The arithmetic invites scrutiny. If 400,000 machines join a manufacturing network without a matching rise in output targets, the labour effect is a change in the composition of the workforce rather than in its tooling alone.
That effect is hardest to read at the supplier tier, where 250,000 of the machines are headed and where redeployment options are typically narrower than inside Toyota's own plants. No headcount figures accompany the robot estimate, which leaves the substitution question open rather than answered.
The framing itself does operational work. Workers who treat the robots as colleagues are more likely to supply the demonstrations the models need, and training data is the one input Toyota cannot buy at scale.
Why This Matters
Toyota factory robots at this scale set a public benchmark for what physical AI costs and what it is expected to deliver, and the number is large enough to force every competitor to respond. If behaviour models genuinely absorb craft skills from short demonstrations, the constraint on factory automation stops being engineering labour and becomes capital, which favours the largest manufacturers. If the transfer only holds for repetitive tasks, 400,000 shrinks into a modernisation budget and the retirement of skilled workers stays an unhedged risk across the industry. The signals worth tracking are defect rates and headcount at the first plants to run the fleet.
Related Articles
- Toyota Humanoid Robots: 400,000 Units, ¥1tn a Year
- Walden Robotics Humanoid Robots Hit the Factory Floor: $300M Seed Round and a $1.1B Valuation
- China pushes 10,000 humanoid robots by 2026 as factories struggle to absorb them
✔Human Verified
Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.