Industry and Economics
Robotics as a Service: When Renting Beats Buying
Paying per hour or per month shifts capital risk to the supplier and costs more over a long life. The break-even is usually around three years, and utilisation decides it.
Renting wins when the deployment is short, uncertain or seasonal, and when the customer lacks the engineering capacity to own the machine. Buying wins on anything running for more than about three years at steady utilisation, because the subscription includes a financing charge, a service margin and a risk premium. One humanoid programme now prices at roughly 25 US dollars per robot operating hour, which is about 100,000 a year at 4,000 hours.
The three structures
| Model | Customer pays | Supplier bears | Suits |
|---|---|---|---|
| Fixed monthly fee | Same amount regardless of use | Hardware, service, uptime | Steady predictable operations |
| Per operating hour | Only for hours the robot works | Hardware, service, utilisation risk | Variable demand, pilots |
| Per unit of output | Per pick, per pallet, per part | Everything including performance | Well-defined repetitive tasks |
The models differ in who carries utilisation risk. A fixed monthly fee leaves it with the customer, per-hour pricing splits it, and per-output pricing moves it entirely to the supplier, who then prices that risk into the rate.
Where the break-even sits
| Year | Purchase, cumulative | Subscription, cumulative | Position |
|---|---|---|---|
| 0 | 120,000 | 0 | Purchase pays up front |
| 1 | 124,200 | 54,000 | Subscription ahead |
| 2 | 128,400 | 108,000 | Subscription ahead |
| 3 | 132,600 | 162,000 | Purchase ahead |
| 5 | 141,000 | 270,000 | Purchase clearly ahead |
| 8 | 153,600 | 432,000 | Purchase far ahead |
Assumptions: 120,000 installed cost, 4,200 a year in service and energy, against a 4,500 monthly subscription including service. Crossover falls between year two and three, which is the figure most published comparisons converge on. Change the subscription rate and the crossover moves, and the shape does not.
The arithmetic in one place
| Item | Purchase | Subscription |
|---|---|---|
| Up-front capital | 120,000 | 0 |
| Annual service and spares | 3,600 | included |
| Annual energy | 600 | usually excluded |
| Annual subscription | 0 | 54,000 |
| Programming a new part | internal or 1,500 to 6,000 | contract dependent |
| Residual value after 8 years | 10 to 25 % of cost | 0 |
| Balance sheet treatment | capital asset | operating expense |
| Obsolescence risk | customer | supplier |
Two rows decide most real cases and neither is the price. The balance sheet row determines which budget the money comes from, and the obsolescence row determines who loses if the technology moves. On mature industrial arms obsolescence risk is small; on humanoids and learned systems it is the dominant consideration.
When renting is right anyway
- The deployment is short or uncertain. A pilot, a seasonal peak or a contract of unknown length. Paying a premium to avoid a stranded asset is rational.
- No engineering capacity. A subscription that includes programming, maintenance and uptime buys a capability the organisation does not have and would take years to build.
- The technology is young. Humanoids and advanced learned systems change quickly. Renting transfers obsolescence risk to the supplier, which is why that category priced this way early.
- Capital is constrained. An operating expense that needs no capital approval can be the difference between a project happening and not happening, even when purchase is cheaper.
- Uptime risk must sit with the supplier. If a stoppage is very expensive, a contract that pays only for working hours puts the incentive in the right place.
What the contract must define
- What counts as an operating hour, and whether idle, standby and fault time count.
- Uptime commitment and remedy, with a defined measurement method rather than a percentage alone.
- Who owns the programs and the data the robot produces, and what happens to both at the end.
- Exit terms. Notice period, removal, restoration of the site, and whether a purchase option exists.
- Change handling. What a new part or a new task costs, since that is where subscription agreements usually turn expensive.
- Who carries conformity. The cell still needs a risk assessment and a declaration, and the contract should say who issues it.
Frequently asked questions
When does renting a robot beat buying?
For deployments shorter than about three years, uncertain or seasonal demand, organisations without engineering capacity, and young technology where obsolescence risk is real. Beyond three years at steady utilisation, purchase is normally cheaper.
What does a robot subscription cost?
It varies by model. One published humanoid programme prices at roughly 25 US dollars per robot operating hour, which is about 100,000 a year at 4,000 hours. Industrial cell subscriptions are commonly quoted as a monthly fee including service.
Which pricing model is best?
Per unit of output aligns incentives best, since uptime becomes the supplier's problem, and it is the hardest to contract. Per hour splits utilisation risk. A fixed monthly fee is simplest and leaves utilisation risk with the customer.
Why do humanoid programmes price by the hour?
Because the technology changes quickly and reliability is still improving. Hourly pricing transfers obsolescence and uptime risk to the supplier, which is the correct place for it while both are unsettled.
What is the most common contract gap?
The definition of an operating hour, and the cost of change. Whether standby and fault time count, and what adding a new part number costs, are where subscription agreements most often become expensive after signature.
Sources
- The Robot Report, humanoid commercial coverageReporting on operating-hour pricing and fleet contracts
- World Robotics report seriesInternational Federation of Robotics, deployment and service robot data
- Regulation (EU) 2023/1230 on machineryEU-OSHA, conformity obligations that a rental contract must allocate