Why Factories Come Before Homes
Factories are bounded, instrumented, and buying — homes are open, fragile, and personal. The first humanoid jobs live where the rules are clear.
When people picture humanoid robots, they usually picture a home. A robot folding laundry. A robot tidying the kitchen. A robot watching the kids.
That picture is reasonable. It is also far away. The first humanoid jobs are not in homes — they are in factories and warehouses. The reason is not about the robot. It is about the room.
If a robot can stand, why can it not just go and help in a kitchen?
Robots already live in factories
Industrial robots are not a new idea. The International Federation of Robotics reported 542,000 new industrial robot installations in 2024, with around 4.66 million industrial robots in operation worldwide.
Consumer robots exist too — robot vacuums, robot lawnmowers, robot pool cleaners. IFR counts close to 20 million consumer service robots sold in 2024. But almost all of them do one narrow job in a tidy, predictable space.
Robots scale where the room is bounded and the task is the same every day.
Why homes are harder than they look
A home looks calm. To a robot, it is chaos. Floors change material. Light changes through the day. Objects are soft, heavy, fragile, sticky, and never quite in the same place. Children, pets, and visitors move in unpredictable ways.
Manipulation research shows this clearly. In the HomeRobot benchmark, baseline systems reached around a 20% real-world success rate on open-vocabulary mobile manipulation tasks — useful progress, and a long way from a chore your family would trust.
- Variable surfaces
Rugs, tile, wet floors, low thresholds — a single misjudged step ends the trip.
- Unstructured objects
Clothes, food, cables, toys — every grasp is slightly new.
- People in the way
A home is shared space. The robot must yield, not just plan.
- No tolerance for damage
One broken mug is a story friends tell for a year.
Why factories and warehouses fit first
Factories solved most of the room problem before humanoids arrived. Floors are flat. Lighting is fixed. Parts are presented in the same orientation. Workflows are written down. Safety rules already exist. And there is a buyer with a budget who measures success in hours per shift, not feelings.
- 01Bounded space
A defined cell, aisle, or line — not the whole building.
- 02Repeated task
The same totes, the same fixtures, the same sequence.
- 03Trained people
Operators who can recover the robot when it stops.
- 04Clear value
A measurable hour, a measurable cost, a measurable output.
Homes have none of these. Factories have all four.
What the named examples actually show
The clearest public humanoid work is industrial. Agility Robotics has Digit moving totes at GXO. Figure has Figure 02 in BMW's Spartanburg plant. Apptronik has Apollo in trials with Mercedes-Benz. Hyundai has announced an Atlas rollout starting in 2028.
Notice what these have in common. Named site. Named task. Industrial floor. A customer who can afford a long pilot. None of them are home help.
Industrial pilots are not proof of a home product.
What people often misunderstand
- Mistake 01
“A humanoid shape means a humanoid life.”
Two legs and two arms do not give a robot judgement about a child or a pan of hot oil.
- Mistake 02
“Factory work is the easy version of home work.”
It is not easier — it is bounded. Bounded is what makes work possible at all today.
- Mistake 03
“If a robot can walk in a video, it can help at home.”
Walking through a hallway and surviving a real kitchen are different problems.
- Mistake 04
“The home is the bigger market, so it will come first.”
Size of market does not decide order. Cost of failure does.
What would need to change before homes are next
- A long stretch of industrial uptime that customers will actually pay for.
- Manipulation that handles soft, irregular, and unfamiliar objects.
- A safety story that holds up around children, elderly people, and pets.
- A price that makes sense for a household, not a factory line.
- A repair, support, and insurance system that ordinary buyers can use.
What single home task is valuable enough to justify a large, mobile, expensive machine?
- Factories already employ millions of robots; homes employ narrow appliances.
- Home manipulation is an unsolved research problem, not a near-term product.
- The first humanoid jobs are bounded industrial tasks with named customers.
- Bounded space, repeated task, trained people, and clear value are the four ingredients.
- Market size does not decide order of arrival — cost of failure does.
- Industrial robot
- A robot working in manufacturing, usually fixed to a cell, line, or arm base.
- Service robot
- A robot doing work for people outside manufacturing — domestic, professional, or both.
- Mobile manipulation
- Moving around a space and using a manipulator (arm, hand) to act on objects.
- HomeRobot benchmark
- A research test of open-vocabulary home manipulation tasks.
- Bounded environment
- A space with known layout, lighting, and rules — easier for robots to act in.
- Pilot
- A paid or unpaid trial of a robot system at a specific site with a specific task.
Sources and evidence notes
What this essay leans on
| Claim | Evidence | Strength | Note |
|---|---|---|---|
| Industrial robots are an established global market. | IFR World Robotics 2025 — 542,000 industrial installations in 2024, ~4.66M operational stock. | Strong | Industry standard count. |
| Consumer service robots are common but narrow. | IFR World Robotics 2025 — ~20M consumer service robots sold in 2024, mostly floor cleaners and lawnmowers. | Strong | Single-task home appliances dominate. |
| Open home manipulation is unsolved. | HomeRobot benchmark (Yenamandra et al., 2023) — ~20% baseline real-world success rate. | Strong | Research benchmark, not deployment. |
| Named humanoid pilots are industrial. | Agility/GXO; Figure/BMW Spartanburg; Apptronik/Mercedes-Benz; Hyundai Atlas plan (Reuters). | Strong | Industrial sites only; no comparable home deployments. |