The comparison everyone makes, and why it misleads
Most people put the price of a robot next to a yearly salary and conclude one of two things: that the robot is absurdly expensive, or that it pays back in a year. Both are wrong, for the same reason. The robot's price says nothing about how many hours it works, and the salary is a fraction of what a worker actually costs you.
The number that decides the case is simpler and harder: what does it cost to get one unit of work done, on this task, on this shift pattern, by a person and by a robot?
What an hour of labor really costs
Start with the wage for the job and the shift, including the night and weekend premiums. Then add what most people forget:
- Employer costs on top of the wage: taxes, insurance, benefits. Typically another twenty to forty percent, depending on the country.
- The agency's cut, if the shift is staffed through an agency. Commonly thirty to fifty percent on top.
- Overtime paid to cover the shifts nobody could be found for.
- Turnover: recruiting, training, and the weeks a new hire is slower than the person they replaced, spread over how long people actually stay.
- Absence, and the cost of covering it.
Add it up and divide by the hours people actually work, not the hours you pay for. Almost everyone who does this for the first time finds a number well above the wage they had in their head.
What an hour of robot really costs
- The robot itself: a monthly fee if you rent its time, or the purchase price and financing spread over its working life.
- Getting it working: fixtures, connecting it to your warehouse or labor system, safety measures. Spread over the deployment.
- Keeping it working: monitoring, maintenance, spare parts, the people watching it remotely.
- Electricity and charging. Small, but not nothing.
- Downtime: charging breaks and unplanned stops, which cut the hours it actually works.
Divide the yearly total by the hours it actually works. This is where the answer swings: nearly all of a robot's cost is fixed, so the same robot costs about three times as much per hour on one shift as it does on three.
Then compare on the work, not the hour
A robot does not match a person's pace on day one. So divide each side's cost per hour by how many units per hour it actually does on this task, using rates you have measured, not rates from a brochure. And allow for the ramp: the first weeks are slower.
A worked example, with round numbers
These figures are made up to show the mechanics. In a consultation, your own numbers go in their place.
| One shift | Three shifts | |
|---|---|---|
| Robot's yearly cost, all in | 100 | 100 |
| Hours it works per year | 1,800 | 5,400 |
| Robot cost per hour worked | 0.056 | 0.019 |
| Robot's pace as a share of a person's | 70% | 70% |
| Robot cost per unit of work, if a person's is 1.0 | about 1.6 | about 0.55 |
Read the last line. Same robot, same task: on one shift it costs sixty percent more than a person per unit of work; on three shifts it costs about half. That is why the shift pattern is one of the first things we ask, and why we build the case per task and per site, never per robot.
What two independent analyses put the per-hour figure at
JPMorgan, in research published in 2025, estimated the cost of running a humanoid robot at $10 to $12 per hour, calculated from current purchase prices and operating costs across the machine's working life. Roland Berger, in a 2025 automation report, projected the cost falling to around $2 per hour as production volumes grow and hardware costs follow. The difference is a time horizon, not a disagreement: JPMorgan models the robot in service today; Roland Berger models where the cost curve points.
Read alongside the table above, the $10 to $12 figure clarifies the single-shift case. On one shift, a robot at today's prices competes with agency rates in the higher-cost labor markets and falls short in the lower ones. On three shifts, it runs under most agency rates and closes on the $2 figure Roland Berger projects at scale. The number of shifts is still the deciding variable.
Where Chinese supply is pushing the price
Chinese makers have published lower prices in 2026 than any Western competitor. AgiBot lists its X2 Neo at $25,600 on its own store (October 2026); Unitree reported a 72% price reduction on its G1 research platform in late 2026. BYD targets 20,000 industrial humanoids in 2026 and states it will reach its cost target by sourcing more than 60% of components from its existing automotive supply chain; no public list price has been named. In China's rental market, a daily rate that ran at 10,000 yuan in early 2025 fell below 1,000 yuan by late 2026 as supply grew (Humanoid.guide, October 2026). Hardware costs are falling. For the calculation above, the deciding variables stay the same: shift count, integration cost and the local labor rate.
Where it usually works, and where it usually does not
With today's robots the case closes first where three things are true at once: the site runs more than one shift, the task is structured handling with a measured human rate, and the labor for it is expensive, scarce, or comes through an agency. Where one of the three is missing the case is usually not there yet. We will tell you so, because a pilot that cannot pay is not worth anyone's time.
