Does a Humanoid Cost $2 or $35 an Hour?
Vendors publish shipments and hours logged, not payback. Here’s a simple, open cost-per-useful-hour model — every input sourced or labeled as an assumption, and open to correction.
You’ll see humanoid robots priced at $2 an hour, $10 to $12, or more than $35. Those figures are measuring different things. I built a simple, open model to show why, and the answer comes down to two numbers vendors rarely publish: how many hours the robot really works, and how much it gets done in each one.
To give you broad brush strokes: a robot’s hourly cost is its yearly cost divided by its productive hours. The first number is easy to find. The second is the one the pitch decks usually leave out.
I run consumer brands that make products in Chinese factories. In my world, nobody quotes the price of a machine without the units it produces. Robots should be held to the same standard.
Why the published numbers disagree
Fortune laid out the spread in October. Roland Berger put robot cost at $2 an hour, JPMorgan at $10 to $12, and reports on early industrial deployments at more than $35 (Fortune).
Per Fortune, Roland Berger’s $2 covers energy, software, network and scheduled maintenance — it excludes buying or leasing the machine. So it’s a running cost at volume, not the cost of a deployment (Fortune).
The same piece walks through a $90,000 industrial robot over five years. At 2,000 working hours a year, the hardware adds about $9 to each hour, and maintenance takes it toward $12. Run it only 1,000 hours a year and the hardware cost per hour doubles (Fortune). The same robot, at the same price, costs twice as much per hour if it only runs half the hours.
Agility’s published math
Agility Robotics filed an S-4 ahead of its merger with Churchill Capital Corp XI. As summarized by Fortune, its Digit robots had logged more than 65,000 hours across nine customer sites, on $1.78 million of 2025 net sales against a $140.2 million operating loss (Fortune).
The filing also has Agility’s own cost model for its next robot: roughly $400,000 over a five-year life. One shift of 2,000 hours a year works out to about $40 an hour. Two shifts, close to $20. Against Agility’s fully loaded worker benchmark of $30 an hour, the robot only wins above about 2,700 hours a year (Fortune).
That’s honest math, and I give them credit for publishing it.
How much work per hour
Hours aren’t the whole story. Fortune cites a JPMorgan estimate that it currently takes about two humanoids to match the output of one human worker — which doubles the effective hourly cost (Fortune).
That’s the other half of the equation, and it’s the number vendors rarely publish next to cost. A robot can log a lot of hours and still cost more than the labor it replaces. So I built a simple model that puts cost and output on the same page.
An open cost-per-useful-hour model
My formula:
Cost per useful hour = (yearly robot cost ÷ scheduled hours + supervision cost per robot-hour) ÷ productivity vs. a human
Every input is either sourced or labeled as my assumption. Change any of them and the answer moves — that’s the point.
| A. Pilot, 1 shift | B. Pilot, 2 shifts | C. Cheaper robot, 2 shifts | D. 2030-style, ~3 shifts | |
|---|---|---|---|---|
| Robot cost per year | $80,000 (Agility’s ~$400k over 5 yrs, via Fortune) | $80,000 (same) | $32,940 = $114,700 ÷ 5 (IDTechEx 2024 avg. price) + $10,000 upkeep (ASSUMPTION) | $12,400 = $37,000 ÷ 5 (IDTechEx 2030 forecast price) + $5,000 upkeep (ASSUMPTION) |
| Scheduled hours per year | 2,000 (Fortune) | 4,000 (Fortune) | 4,000 (ASSUMPTION) | 6,000 (ASSUMPTION) |
| Robot cost per hour | $40.00 | $20.00 | $8.23 | $2.07 |
| Supervision per robot-hour | $6.00 (ASSUMPTION: one $30/hr person per 5 robots) | $6.00 (same) | $3.00 (ASSUMPTION: 1 per 10) | $1.50 (ASSUMPTION: 1 per 20) |
| Productivity vs. a human | 0.5 (JPMorgan, via Fortune) | 0.5 (same) | 0.7 (ASSUMPTION) | 0.8 (ASSUMPTION) |
| Cost per useful hour | $92 | $52 | $16 | $4.46 |
What it’s compared against:
- The BLS median wage for hand laborers and material movers was $19.35 an hour in May 2025 (BLS OEWS).
- In private industry, wages are 70.0% of total employer compensation cost (BLS ECEC, June 2026). Grossing up $19.35 by that ratio gives roughly $27.60 loaded. (That’s my rough approximation — the 70% is an all-industry average, not specific to warehouse jobs.)
- Agility’s own benchmark is $30 fully loaded (Fortune).
Notes on the model, candidly:
- I’m treating Agility’s ~$400,000 as an all-in five-year cost and not adding separate upkeep in A and B. If it excludes items, A and B get worse.
- The supervision cost in A and B is my addition, not Agility’s. Without it, A and B are $80 and $40 per useful hour.
- Two-shift scenarios assume the robot is actually busy for the whole shift. Idle time goes straight into the denominator.
Reading the four scenarios
Today’s pilot math (A and B) loses to a human by a wide margin once you adjust for output. The robot is the expensive worker.
The cheaper-robot case (C) beats a loaded warehouse wage on paper — but only with two full shifts, light supervision and solid productivity. Miss on any one of those and it slides back toward B.
The 2030-style case (D) is where the “$2 to $5 an hour” headlines live. IDTechEx projects high-utilization industrial scenarios could get operating costs below $5 an hour by around 2030, with average selling prices falling from about $114,700 in 2024 to around $37,000 by 2030 (IDTechEx). It also estimates 2026 payback of around six months in high-utilization scenarios versus about 15 months at medium utilization — and warns that a short payback isn’t guaranteed profitability (IDTechEx).
My read: the sticker price matters. The schedule matters more. Productivity per hour matters most, and it’s the number nobody discloses consistently.
The output data that does exist is hard to compare across vendors. Hours logged, totes moved, and cycles completed are all useful internally. They aren’t a payback figure, and they don’t let an operator compare two robots on the same task.
What I’d want in writing
- If you’re buying, ask for three numbers in writing: autonomous hours per robot per week, interventions per hour, and output versus your current crew on your task. If a vendor won’t give them, price the deal as scenario A. I’d look at this differently if we had published, comparable output-per-hour data from real customer sites. If productivity is already closer to 1.0 than 0.5, scenario B starts to look fine.
- If the vendor sells “work” instead of robots — per tote, per pallet, per hour — that shifts the utilization risk to them. Fortune argues for exactly this labor-as-a-service model (Fortune). I like it. Just compare the price to your fully loaded cost, not the wage.
- If you’re investing, the schedule is the product. A company that can keep robots busy across shifts beats one with a better robot sitting idle.
If you’re already running these machines: what productivity ratio are you actually seeing on the floor?
Disclosure: I hold no positions in any company mentioned. No company reviewed this piece. My own consumer brands are manufactured in China. This is a simplified model, not investment advice.
Sources & notes
- Fortune, “Does a robot cost $2 or $35 an hour? A multitrillion-dollar market depends on how the answer,” Navin Chaddha and Mark Minevich, Oct 16, 2026 (commentary; the authors disclose that Mayfield is an investor in Rhoda AI). Source for the Roland Berger, JPMorgan and early-deployment figures, the $90,000 example, and the Agility S-4 figures: https://fortune.com/2026/10/16/when-will-physical-ai-robots-pay-off/
- IDTechEx, “Humanoid Robots Show ROI, but Success Depends on Effective Output,” Shihao Fu, May 12, 2026: https://www.idtechex.com/en/research-article/humanoid-robots-show-roi-but-success-depends-on-effective-output/34724
- U.S. Bureau of Labor Statistics, OEWS May 2025, SOC 53-7062, national median hourly wage $19.35: https://data.bls.gov/oesprofile/?areas=INDUSTRY%2CSTATE%2CMSA&major_group=530000&measure=01&occupation=537062&year=2025
- U.S. Bureau of Labor Statistics, Employer Costs for Employee Compensation, June 2026 (private industry wages = 70.0% of compensation): https://www.bls.gov/news.release/ecec.nr0.htm
- Model assumptions (upkeep, hours in C and D, supervision ratios, productivity in C and D) are the author’s own and labeled in the table.