Automation cases are usually argued in payback months or five-year totals. Both hide the same thing: how many hours the machine is switched on. This model prices a cobot in dollars per productive hour and runs it at one, two and three shifts, against a fully loaded US manufacturing labour rate for the same shifts.
It is a different question from two we have already answered — payback months by application and the five-year total cost of an industrial robot. Neither varied the shift count. That variable turns out to move more than anything else in the model.
The subject and the inputs
A single UR5e-class cobot cell, five-year horizon.
| Line | Value | Source |
|---|---|---|
| Hardware (arm, controller, pendant) | **$39,083.16** | Automation Distribution list price, [our 28 July check](https://www.grabarobot.com/blog/cobot-payload-class-price-curve-2026/) |
| Integration and tooling | 40% of hardware = $15,633 | Integrator range of 30–50%, per our five-year TCO model |
| Deployed capital | **$54,716** | |
| Maintenance | 5% of hardware per year at one shift, scaled with running hours | Industry 3–8%/yr range |
| Energy | 0.5 kW average, $0.12/kWh | Assumption, stated |
| Availability | 90% at one and two shifts, 85% at three | Assumption: at three shifts, maintenance comes out of production time |
Running hours are 2,000, 4,000 and 6,000 per year.
The result
| 1 shift | 2 shifts | 3 shifts | |
|---|---|---|---|
| 5-year maintenance | $9,771 | $19,542 | $29,312 |
| 5-year energy | $600 | $1,200 | $1,800 |
| **5-year total cost** | **$65,087** | **$75,458** | **$85,829** |
| Productive hours over 5 years | 9,000 | 18,000 | 25,500 |
| **Cost per productive hour** | **$7.23** | **$4.19** | **$3.37** |
Going from one shift to three raises total spend 31.9% and cuts cost per hour 53.4%.
The whole advantage is capital spreading
Split the hourly figure into the capital part and the running part and the mechanism is unmistakable:
| 1 shift | 2 shifts | 3 shifts | |
|---|---|---|---|
| Capital per productive hour | $6.08 | $3.04 | $2.15 |
| Maintenance + energy per hour | $1.15 | $1.15 | **$1.22** |
Capital per hour falls 64.6%. Operating cost per hour is flat — and at three shifts it is *higher*, because the availability assumption takes hours away faster than the extra running adds them.
This matters because operating cost is the line buyers negotiate. Service contracts, energy tariffs and consumables are where procurement spends its attention, and they are worth about $0.07 per hour across the entire range of shift patterns. Hours are worth $3.86.
Against a person
US manufacturing production and non-supervisory workers averaged $30.27 per hour in June 2026 (BLS Table B-8). Wages are 69.9% of total employer compensation cost in private industry (BLS ECEC, March 2026), which puts the fully loaded first-shift rate at $43.31. The 2025 CompTool shift differential study of 227 organisations finds weekday second-shift premiums clustering at $1.00/hour and third shift at $1.50/hour in manufacturing.
| 1st shift | 2nd shift | 3rd shift | |
|---|---|---|---|
| Loaded labour, per hour | $43.31 | $44.74 | $45.45 |
| Cobot, per productive hour | $7.23 | $4.19 | $3.37 |
| Ratio | **6.0×** | 10.7× | **13.5×** |
| Cobot as % of labour | 16.7% | 9.4% | 7.4% |
Adding shifts moves the two costs in opposite directions. The robot gets 53% cheaper per hour; the person gets 4.9% more expensive. The gap widens from six-fold to thirteen-fold — and none of that widening comes from the robot getting better.
What actually moves the number
Holding everything else at base and flexing one input at a time, on the one-shift case:
| Input | Range tested | Cost per hour | Spread |
|---|---|---|---|
| Availability | 70% → 95% | $9.30 → $6.85 | **$2.45** |
| Hardware price | $39,083 → $43,544 (UR5e list range) | $7.23 → $8.05 | $0.82 |
| Integration share | 30% → 50% of hardware | $6.80 → $7.67 | $0.87 |
Uptime swings the answer about three times as hard as the price of the arm. A buyer who negotiates $4,000 off the robot and then accepts a cell that runs at 75% availability has made the number worse, not better.
Where this model breaks
Four things it does not capture, stated plainly:
It compares paid hours to productive hours. The labour figure is what an employer pays per hour on the clock; the robot figure is per hour of production. That is generous to labour, so the true gap is wider than the table — but it means these are not identical units and should not be quoted as if they were.
A cobot does not replace one person. It replaces a task. One operator commonly tends several cells, and the cell still needs loading, changeover and inspection. The correct comparison is against the fraction of a role displaced, which is site-specific and which we are not going to invent.
Dollars per hour is not dollars per part. If the human is faster at the operation, the hourly gap can invert on a per-piece basis. Throughput belongs in the case and is not in this model.
The maintenance line survives only until something structural fails. Our teardown of 96 Universal Robots spare-part listings put a single Size 2 wrist joint at $6,507 — 3.3 years of the one-shift maintenance budget in this model, in one part. The 5%-per-year line describes a robot that does not break, and at three shifts you are asking it not to break three times as hard.
The takeaway
The cost-per-hour question has an answer, and the answer is that it is barely about the robot. Between $3.37 and $7.23 per hour, the machine is cheap against $43 to $45 of loaded labour under every assumption we tested. What decides a project is whether there are hours to put on it — because an idle cobot is not cheaper per hour, it is only cheaper in total, and that is the version of the number that gets automation projects approved and then disappoints.
For the fleet-scale version of the same logic, see our warehouse AMR robots-per-worker breakeven, where density rather than shift count is the dominant variable. And for the hardware prices underneath all of it, the collaborative robot category page.
Sources
- BLS, Table B-8: average hourly earnings of production and non-supervisory employees, June 2026 — manufacturing $30.27
- BLS, Employer Costs for Employee Compensation, March 2026 — private industry $46.60/hour total, wages 69.9%, benefits 30.1%
- CompTool 2025 Shift Differential Best Practices Study — 227 organisations, manufacturing differentials $1.00–$2.50/hour
- Automation Distribution — Universal Robots UR5e — hardware list price
- GrabaRobot — five-year industrial robot TCO model — integration and maintenance ranges
- GrabaRobot — Universal Robots spare-parts cost share — $6,507 wrist joint
