When Does a Humanoid Robot Pay for Itself in German Manufacturing?
A €165,000 single-task example: how productive hours, internal support and task performance determine humanoid robot costs and payback in Germany.
Humanoid robots are moving from impressive demonstrations into actual factories. Figure reported more than 1,250 hours of runtime and 90,000 parts loaded during its Figure 02 deployment at BMW's Spartanburg plant. BMW also reported initial testing of Hexagon's AEON at Leipzig in December 2025. These are examples of deployment progress, not proof of a universal business case. Figure deployment report, BMW's Leipzig pilot.
For manufacturers, the important question is simple: when does the investment actually pay for itself?
With a €165,000 initial investment, €3.50/hour of internal support and 70% human-equivalent productivity, our example yields simple payback of 3.3 years at 2,000 productive robot hours a year, 1.5 years at 4,000, and 0.9 years at 6,000. Those results depend on achieving the utilization and actually avoiding the labor costs assumed below.
We work with manufacturers on humanoid integration, and the answer depends on much more than the hardware price: productive hours, task-level productivity and the labor cost that can actually be substituted.
Start with German manufacturing labor costs
According to the German Federal Statistical Office, employers in German manufacturing paid an average of €49.50 per hour worked in 2025. This includes employer labor costs rather than just gross wages. Destatis, 2025 labor costs.
For this calculation, we round that to €50 per human hour. A production helper may cost less; a skilled shift worker may cost more. Use the cost of the actual task in your own business case.
Time freed up does not automatically become a cash saving. The payback calculation assumes that the equivalent labor cost is genuinely avoided, for example through reduced overtime, agency labor or an otherwise necessary hire. Redeploying an employee can create valuable capacity, but should be valued separately if payroll does not fall.
What does a humanoid robot actually cost?
Buying the robot is only part of the project. A robot standing in a crate does not generate ROI: it needs integration into a real production process.
For early planning, we use the following broad ranges. They are SchmidtFactories planning estimates, not vendor quotes or universal market prices.
| Cost component | Planning range |
|---|---|
| Humanoid hardware | €80,000–€150,000 |
| Integration and commissioning | €50,000–€100,000 |
| Safety, tooling, end effectors and infrastructure | €20,000–€50,000 |
| Initial project total | €150,000–€300,000 |
For one clearly defined task, we use €165,000 as a representative initial project budget, based on our internal planning experience. This sits toward the lower end of the wider range and includes hardware, integration, tooling, safety and commissioning. It is not a promise that every application fits this budget.
More complex tasks, additional workstations or substantial process changes require their own estimate. Any engineering needed to add tasks later must also be included when evaluating the benefit of higher utilization.
Recurring costs: external service and internal support
We currently allow €10,000–€15,000 per robot per year for maintenance, technical support, software, cloud infrastructure, updates and related services. Our model uses €12,500/year. This is a budget assumption to validate against the supplier's scope, including wear parts and battery replacement where required.
Internal plant support is budgeted separately at €3.50 per productive robot operating hour. This allocates the total annual internal support effort across productive hours; it does not mean someone stands beside the robot all day.
It covers occasional resets, exception handling, changeovers, minor troubleshooting, charging-related intervention, monitoring and coordination with maintenance. At 4,000 hours/year, that is €14,000 of internal support. At a €50 staff-hour cost, it buys 280 support hours a year, or about 4.2 minutes per productive robot hour.
Measure the actual support burden during validation. A deployment needing continuous supervision has a very different business case. Larger fleets may share support resources, but those savings must be demonstrated.
Electricity is a small part of this model
Figure publishes 2.3 kWh of battery capacity and five hours of runtime for Figure 03. Dividing those figures gives an indicative 0.46 kWh per operating hour; it is not a measured energy consumption figure for your task. Figure's battery development report.
At an assumed electricity price of €0.20/kWh, that is €0.092/hour before charging losses. We use a rounded €0.10/hour allowance. Actual duty cycle, charging losses and auxiliary equipment should be checked at the site.
In this example, productive utilization has a much greater economic effect than electricity.
One robot hour is not necessarily one human hour
If a person completes 100 acceptable cycles per hour and the robot completes 70, one productive robot hour substitutes 0.7 human hours. We call this human-equivalent productivity.
We calculate three task-dependent scenarios:
- 60%: lower productivity scenario.
- 70%: base case.
- 80%: higher productivity scenario.
These are modeling assumptions, not guaranteed performance levels. Measure comparable output at the required quality on the actual process.
Productive robot hours exclude charging, faults, changeovers and waiting without useful output. Productivity then compares output within those productive hours. Keep these definitions consistent so downtime is not counted twice. Scheduled shift hours alone are not productive hours: 4,000 or 6,000 hours require an operating plan that accommodates the losses.
Five years of economic life does not mean everything becomes worthless
We use five years as the economic life for the cost comparison, spreading the full initial investment across that period. This is a planning horizon, not a claim about a mandatory replacement date or tax depreciation.
Code, integrations, trained AI models, process knowledge and some tooling may remain useful beyond year five, including with replacement hardware. Reuse depends on compatibility, licenses, documentation and the need for adaptation and revalidation.
To keep the comparison straightforward, we give that potential future value no residual-value credit in the numbers below. It can improve the longer-term economics, but the payback figures do not rely on it. Replacing hardware therefore need not mean paying for the entire original engineering project again.
The assumptions and formulas
| Input | Model assumption |
|---|---|
| Initial investment | €165,000 |
| Annual maintenance, software and external support | €12,500 |
| Allocated internal support | €3.50/productive robot hour |
| Electricity allowance | €0.10/productive robot hour |
| Avoided human labor cost | €50/hour |
| Economic life for cost comparison | 5 years |
| Residual value credited | €0 |
Let H be annual productive robot hours and p human-equivalent productivity (0.70 in the base case).
Robot cost per productive hour = (€165,000 ÷ 5 + €12,500) ÷ H + €3.60.
Cost per equivalent human hour = robot cost per productive hour ÷ p.
Annual net operating savings = H × (€50 × p − €3.60) − €12,500.
Simple payback in years = €165,000 ÷ annual net operating savings.
There is no payback if annual net savings are zero or negative. The five-year capital allocation belongs in the hourly cost comparison; it is not deducted again from operating savings in the payback calculation.
The model uses constant annual utilization and costs, before financing, taxes, discounting and ramp-up losses. It assumes no major replacement expense beyond the stated budgets. Add ramp-up, exceptional repairs and other site-specific costs where relevant. Payback is a recovery period, not an annual investment return.
Three utilization scenarios
| Productive robot hours/year | Cost/robot hour | Cost/equivalent human hour at 70% | Annual net savings at 70% |
|---|---|---|---|
| 2,000 | €26.35 | €37.64 | €50,300 |
| 4,000 | €14.98 | €21.39 | €113,100 |
| 6,000 | €11.18 | €15.98 | €175,900 |
Hourly costs are rounded to cents. Equivalent-hour costs and payback are calculated from unrounded values.
2,000 productive hours: a narrower economic margin
At 2,000 hours/year, the robot delivers 10,000 productive hours over five years. Its cost is €26.35 per robot hour, or €37.64 per equivalent human hour at 70% productivity. Simple payback is 3.3 years.
This can be viable for a well-bounded task, but offers less room for ramp-up, lower throughput or extra support costs than higher-utilization deployments. Labor availability and ergonomics may add value beyond the quantified labor savings.
4,000 productive hours: a stronger business case
At 4,000 hours/year, robot cost falls to €14.98/hour, or €21.39 per equivalent human hour in the base case.
The robot substitutes 2,800 human hours, valued at €140,000/year. Subtract €12,500 for external maintenance and software, €14,000 for internal support and €400 for electricity: €113,100 in annual net operating savings.
The €165,000 investment pays back in 1.5 years. At 60% productivity, payback is 1.8 years; at 80%, 1.2 years.
6,000 productive hours: high utilization must be earned
At 6,000 hours/year, robot cost falls to €11.18/hour, or €15.98 per equivalent human hour at 70%. Simple payback is 0.9 years, approximately 11 months.
This is an ambitious utilization scenario, not a default expectation for an early deployment. The plant needs enough suitable work, reliable operation and a charging and maintenance plan. If achieving those hours requires more integration or hardware, increase the investment accordingly.
The payback table
Using €50/hour of genuinely avoided labor cost:
| Productive robot hours/year | 60% productivity | 70% productivity | 80% productivity |
|---|---|---|---|
| 2,000 | 4.1 years | 3.3 years | 2.7 years |
| 4,000 | 1.8 years | 1.5 years | 1.2 years |
| 6,000 | 1.1 years | 0.9 years | 0.8 years |
These are scenario calculations, not performance guarantees. At 70% productivity, about 1,449 productive robot hours/year are needed to recover the investment within five years under the same assumptions.
Utilization can matter more than purchase price
Two factories can buy the same robot and achieve very different economics if one operates it for 2,000 productive hours and the other for 6,000.
One task repeated across several shifts may be sufficient. Where demand for that task is limited, several compatible tasks can raise utilization: machine tending, material handling, assembly support and intralogistics, for example. See our industrial application overview.
That flexibility is a potential benefit of a human-scale platform in an existing factory. It is not free: include task changes, travel time, tooling, validation and additional engineering in the operating plan and budget. A multi-task deployment cannot automatically inherit the €165,000 single-task estimate.
Choose the right automation for the task
If a part arrives in exactly the same position every six seconds and needs the same movement millions of times, a conventional industrial robot will often be a better candidate.
Humanoids offer potential where manipulation, mobility, high product variation, several workstations and existing human-oriented infrastructure combine. Compare them with fixed robots, cobots, mobile manipulators and AMRs before committing.
The economic promise is flexibility across useful industrial work. The test is measured output, reliable operation and a supportable integration cost.
When does a humanoid pay for itself in Germany?
For this €165,000 single-task example, the 60–80% productivity scenarios give:
- 2,000 productive hours/year: approximately 2.7–4.1 years.
- 4,000 productive hours/year: approximately 1.2–1.8 years.
- 6,000 productive hours/year: approximately 0.8–1.1 years, if that utilization is achievable within the budget.
Five years is the comparison period, and reusable software and integration work can still carry value afterward. No such value is required to reach the results above.
The useful question is: which task, or combination of tasks, can keep this robot working productively—and what does it cost to deliver that reliably?
That is where we help manufacturers: identifying suitable applications, measuring task fit, planning integration and building the business case before scaling. Request a humanoid pilot fit check or explore our hardware-fit test for one factory task.