Every robotics proposal arrives with a headline number: the robot costs $X and replaces Y labor hours. The arithmetic is usually right and the conclusion usually wrong, because the hourly cost of a deployed robot is not the purchase price divided by hours. It is everything you spend to keep that robot productive, divided by the hours it is actually productive.
The market is now large enough that this distinction matters beyond a single pilot. Industrial robot installations reached a record market value of $16.7 billion, according to the International Federation of Robotics, and the global operational stock passed five million units in 2025, up 9%. Professional service robot shipments grew 24% to almost 250,000 units in the same year. Installations are no longer experimental; they are capital allocations that have to pay back on a schedule.
The number vendors quote and the number that pays
A vendor quote typically covers hardware, a warranty period, and a software subscription. It does not cover the integration work that makes the robot useful, the safety validation that makes it legal to run, the remote staff who intervene when it gets stuck, or the idle hours when there is no work for it. Those four items can add more than the hardware line over a four-year life, which is why the purchase price is the wrong denominator.
The right unit is cost per productive hour: total monthly cost of ownership divided by hours of useful work delivered. It is the only figure comparable to a labor rate, and it is unforgiving in a useful way, because it forces every hidden cost into one number that operations and finance can argue about together.
Hardware amortization: the visible cost
Start with capital cost and a realistic life. A four-year straight-line assumption is conservative and defensible for a robot that runs two shifts a day; five or six years may be reasonable for a simple fixed cell with light duty. Residual value is usually small and should be assumed near zero unless you have an actual secondary market for that model.
Two mistakes are common. Amortizing over the warranty period, which flatters early years and hides replacement risk. And treating the robot as the whole capital cost when the cell around it, including guarding, conveyors, fixtures, end-of-arm tooling, and network drops, frequently costs as much again. The model below rolls those into one capital number on purpose.
Maintenance and teleoperation: the costs that decide
Maintenance is best modeled as an annual percentage of capital cost, and it does not fall with age. A working assumption is 9% per year for a robot in two-shift operation, covering preventive service, spares, and the wear items that fail first: grippers, cables, suction cups, and vision hardware. For a fleet, add the cost of the technician who covers it, which for small fleets is a shared role rather than a dedicated one.
Teleoperation is the line most often missing from the business case, and in manipulation-heavy deployments it dominates. A remote operator intervention model is a staffing ratio: one person supervises some number of robots. At one operator per ten robots and a fully loaded cost of $9,500 per month, supervision adds $950 per robot per month. At one to five, it adds $1,900, more than maintenance. The IFR notes that most current humanoid applications remain specialized and often require human teleoperation, and flags training and maintenance costs as a barrier to adoption; treat the ratio as the central design parameter, not a detail.
Utilization: the multiplier nobody puts on the slide
Utilization is productive hours divided by available hours, and it is the single largest lever in the model. Two shifts, five days a week, is about 320 available hours a month once you subtract breaks and changeovers. A robot that is useful for 65% of that time delivers around 208 productive hours. A robot that sits idle while the line is down, or while waiting for an operator, delivers far fewer.
Most pilots report utilization optimistically because they measure scheduled time rather than delivered work. The honest denominator is productive hours, verified against a throughput log, not the hours the cell was powered on. If you cannot measure it, use a conservative assumption and say so, then replace it with data after eight weeks.
Deployment is a project, not a purchase
Deployment covers site survey, safety assessment and validation, integration with the warehouse or manufacturing system, fixture design, commissioning, operator training, and the productivity dip while people learn the new process. Safety work is not optional: robots operating near people have to be designed and validated against recognized safety standards, and the IFR points to ISO safety standards as a requirement for certification and clear liability. In the United States, OSHA publishes robotics guidance covering worker safety around automated equipment, and it belongs in the cell review rather than in a health and safety appendix.
A useful planning assumption is a deployment cost of roughly a third of robot capital cost for an established use case, and more for a first deployment in a new site. Amortize that over the same four years rather than expensing it in year one, so the hourly figure reflects the true cost of getting to a working cell.
The worked cost per hour, with labeled assumptions
Everything in the table is an assumption for illustration, not a market quote: capital cost of $134,400 including tooling, guarding, and installation; $48,000 of one-time deployment and integration; four-year straight-line amortization; maintenance at 9% of capital per year; $450 per month of fleet software and connectivity; $350 per month of power and facilities; remote supervision at one operator per ten robots at $9,500 per month; and 320 available hours per month at 65% utilization, or 208 productive hours.
| Line item | Monthly cost | Per productive hour | Basis (assumption) |
|---|---|---|---|
| Hardware amortization | $2,800 | $13.46 | $134,400 over 48 months |
| Deployment and integration | $1,000 | $4.81 | $48,000 over 48 months |
| Maintenance and spares | $1,008 | $4.85 | 9% of capital per year |
| Fleet software and connectivity | $450 | $2.16 | Platform subscription per unit |
| Power and facilities | $350 | $1.68 | Cell space, power, cooling |
| Remote supervision | $950 | $4.57 | One operator per 10 robots |
| Total | $6,558 | $31.53 | 208 productive hours per month |
The derived figure is about $31.53 per productive hour, or $78,696 a year for 2,496 productive hours. Compare that with a fully loaded human cost. At an assumed $28 per hour, the robot reaches parity at about 234 productive hours a month, which is 73% of the available 320 hours. That is the number to argue about in the business case, not the purchase price.
The second table shows why. Cost per hour more than doubles across plausible utilization, and the teleoperation ratio moves it across a similar range.
| Utilization (of 320 available hours) | Productive hours | Cost per hour |
|---|---|---|
| 40% | 128 | $51.23 |
| 50% | 160 | $40.99 |
| 65% | 208 | $31.53 |
| 75% | 240 | $27.32 |
| 85% | 272 | $24.11 |
At the same 65% utilization, changing only the supervision ratio gives $29.25 per hour at one operator to twenty robots, $31.53 at one to ten, $36.10 at one to five, and $49.80 at one to two. This is why humanoid pilots built on remote assistance are expensive in a way that hardware cost reductions do not fix: the staffing model has to change before the economics do.
Robots-as-a-service is the other route, and the IFR reports that subscription models continue to grow, already the dominant model for transport and logistics robots in the United States. RaaS moves capital and utilization risk to the vendor, which is real value for a buyer who cannot forecast demand. It also means the vendor has priced that risk in. A $5,500 monthly fee for a comparable unit at 208 productive hours works out to about $26.44 per hour, so the comparison you need is that rate against your own utilization, not against the sticker price of the hardware.
What to do next
- Build the model with six lines: capital, deployment, maintenance, software, facilities, and supervision. If supervision is missing, the model is not finished.
- Set utilization from a measured throughput log, or state it as an explicit assumption and commit to revisiting it after two months of operation.
- Decide the supervision ratio with the operations team before signing, since it sets the headcount plan and the hourly cost.
- Price the safety validation and integration as a project with its own budget and owner, not as a line in the hardware purchase order.
- Compare RaaS fees to your own modeled cost per productive hour, and note which party carries the utilization risk in each case.
Robotics is becoming ordinary capital equipment, and ordinary capital equipment is judged on cost per unit of output. The teams that win the next round of deployments will be the ones that can state that number, defend the assumptions behind it, and show the utilization data that makes it true.
