SHARE
SHARE
We’ve all seen the videos.
A humanoid does a backflip. Another folds a shirt. A robotic arm pours a drink with the confidence of a slightly overqualified bartender. Somewhere on X, someone announces that physical labor is basically solved.
Cool. Now show me the gross margin.
Because while the world waits for humanoids to become cheap, reliable and ubiquitous, the robotics companies actually generating revenue today tend to look far less cinematic. They pick crops. Move pallets. Clean floors. Inspect pipes. Sort packages. Mow grass.
They are specialized, slightly boring and – if the economics work – potentially very valuable.
For venture investors, that creates an awkward problem. Most of the metrics we instinctively understand were designed for software companies.
Robots, unfortunately, have batteries, bearings, technicians, shipping crates, warehouses, and an irritating habit of physically breaking. So the question isn’t whether robotics can be evaluated like SaaS – it can’t.
The better question is: which robotics metrics tell us whether a hardware-heavy company can eventually behave economically like software?
Here are the ones I would watch.
1. Availability: Is the Robot Actually Alive?
In SaaS, churn kills companies. In robotics, downtime often causes the churn first.
A customer doesn’t particularly care that your autonomy stack is state-of-the-art if the machine spends Tuesday afternoon waiting for a technician. Think of a robot as an employee who occasionally refuses to turn on. Do that often enough and HR gets involved.
The KPI: Fleet Availability
How much of the contracted operating time is the robot technically capable of doing its job?
A fleet with 100 deployed robots sounds impressive. A fleet with 100 robots where 28 are waiting for repairs, spare parts or divine intervention is less impressive.
For mature commercial deployments, investors should want availability moving toward industrial-equipment territory – ideally well above 90% and, depending on the application, pushing toward 95%+.
The exact number matters less than the trajectory. If availability isn’t improving as the fleet scales, you probably don’t have a software scaling problem. You have a machine problem.
And machines are expensive problems.
2. Productive Duty Cycle: Does It Work More Than It Sleeps?
Availability alone can be misleading. A robot may be perfectly functional while spending half the day charging.
This matters particularly for drones, warehouse robots, agricultural machines and other battery-dependent systems. Imagine hiring somebody who works for 30 minutes and then takes a 60-minute nap. Technically available. Economically questionable.
The KPI: Productive Duty Cycle
What percentage of the theoretically available operating window is spent doing revenue-generating work?
If one robot can only perform four productive hours inside an eight-hour shift, the customer may need two robots to replace one human. Congratulations! You have automated the job and doubled the headcount.
Battery swapping, opportunity charging, fleet orchestration and energy-efficient drivetrains aren’t engineering footnotes. They are unit economics.
3. Revenue per Robot: How Much Money Can One Piece of Metal Make?
Software investors obsess over ARPU. Robotics investors should obsess over ARPR – Average Revenue per Robot.
One deployed robot is effectively a tiny revenue-producing asset. So ask: how much annual recurring revenue can each machine generate?
And, more importantly: can that number increase without replacing the machine?
This is where robotics starts getting interesting. If the company can increase revenue per deployed unit through software modules, expanded workflows, additional shifts, autonomy upgrades or premium services, the installed hardware base starts behaving like a software distribution channel.
That is a beautiful thing.
Building another robot costs money. Shipping a software upgrade costs almost nothing.
The dream isn’t merely robots-as-a-service. It’s hardware installed once, monetized repeatedly.
4. Contribution Margin per Robot: Does the Robot Actually Make Money?
This is where attractive demos go to die.
A robot generating $3,000 per month in revenue tells you almost nothing. You need to know what happens after paying for everything required to keep that specific machine earning money:
- energy;
- maintenance;
- replacement parts;
- field technicians;
- connectivity;
- remote operators;
- fleet support.
What remains is the robot’s economic contribution.
Call it robot contribution margin, robot gross profit or deployed-unit margin – the terminology matters less than the discipline. Every deployed unit should increasingly resemble its own profitable micro-business.
If revenue grows but service costs grow almost proportionally, you haven’t built scalable automation. You’ve built a field-services company with cool hardware.
5. Hardware Payback: How Long Until the Robot Pays for Itself?
This may be one of the most underappreciated robotics metrics.
Every new unit requires capital before it generates revenue. So calculate:
Hardware Payback Period = Fully Deployed Hardware Cost / Monthly Contribution Profit
If a robot costs $15,000 to manufacture and deploy and generates $1,500 of monthly contribution profit, the hardware pays back in roughly ten months. Lovely.
If payback takes three years, scaling becomes painful. Every new customer creates another funding requirement. Growth consumes cash instead of releasing it.
Software companies scale servers. Robotics companies scale inventory. Your balance sheet notices the difference.
For many VC-backed models, getting deployed-unit payback somewhere around or below 12–18 months is extremely attractive. Longer payback isn’t automatically fatal, but somebody has to finance it.
And if the answer is permanently “the next equity round,” investors should pay attention.
6. Human Intervention Rate: How Autonomous Is “Autonomous”?
Welcome to robotics’ favorite magic trick.
Founder: “Our fleet is fully autonomous.”
Investor: “Fantastic. How many people are remotely helping it?”
Silence.
One of the most important metrics in robotics is therefore not a robot metric at all. It’s a human metric. Track intervention minutes per robot-hour and robots per remote operator.
If every robot needs a dedicated human supervisor, you haven’t eliminated labor. You’ve relocated it to a cheaper office and added an expensive robot. That may still be a business, it just isn’t the business your pitch deck says it is.
The real prize is a steadily improving robot-to-operator ratio. 1:1 is teleoperation. 5:1 is interesting. 20:1 starts looking like genuine leverage. 100:1 gets very interesting indeed.
The absolute benchmark depends heavily on the application. The direction doesn’t.
Every quarter, the fleet should need less human attention per productive robot-hour. That is one of the clearest signals that autonomy is turning into economic leverage.
7. Deployment Time: How Much Consulting Comes Free With Every Robot?
Here is another metric founders don’t always put on slide 12: time from signed contract to productive deployment.
A robot that takes six engineers, three site visits and eight weeks of custom integration to install is not a product, it is a consulting engagement wearing wheels.
Early deployments are obviously messy. But investors should look for declining deployment effort as the company learns.
How many engineer-hours does deployment require? How much site-specific mapping? How much customer infrastructure? How much custom code? How quickly can Robot #500 be deployed compared with Robot #5?
True productization shows up when deployment becomes boring.
Boring is good. Boring scales.
8. Customer ROI: Does the Robot Beat the Human?
And this may be the most important metric of all.
Forget the robot for a moment. Look at the customer.
What is their economic reason to adopt it?
Robotics companies often begin with the assumption that replacing labor is inherently valuable. It isn’t.
Humans are remarkably capable machines. They climb stairs, open doors, improvise, recharge themselves overnight and handle edge cases. And, inconveniently for robotics founders, they arrive with zero manufacturing cost on the customer’s balance sheet.
So the real question is: does the robot deliver the same output materially cheaper, safer, faster or more reliably than the human alternative?
If a delivery robot costs $10,000, requires remote intervention, gets defeated by badly parked scooters and takes twice as long as somebody on an e-bike, then automation hasn’t improved the system. It has added capex.
The strongest robotics companies don’t automate tasks simply because they can. They automate tasks where machines have a structural economic advantage:
- dirty jobs;
- dangerous jobs;
- extremely repetitive jobs;
- labor-constrained jobs;
- 24/7 jobs, and
- tasks where humans are expensive, unavailable or simply don’t want to do the work.
That difference matters enormously.
The VC Robotics Scorecard
Metric | Mechanical Science Project | Venture-Scale Machine |
Fleet availability | Frequently offline | Reliably >90%, trending toward industrial-grade uptime |
Productive duty cycle | Spends half its life charging | Majority of available time producing value |
Revenue per robot | Fixed by hardware | Expands via utilization, software and services |
Robot contribution margin | Eaten by service & maintenance | Improves meaningfully with fleet scale |
Hardware payback | 24–36+ months | Ideally ~12–18 months or better |
Human intervention | Robot ≈ remote-controlled employee | Human supervision diluted across a large fleet |
Deployment burden | Engineers camp at every customer site | Increasingly repeatable and plug-and-play |
Customer ROI | “Robots are cool” | “Not using the robot would be economically irrational” |
These aren’t commandments. Different applications will have radically different numbers.
A surgical robot shouldn’t be evaluated like a floor-cleaning robot. A warehouse AMR isn’t a drone. And an agricultural robot operating seasonally won’t resemble a factory machine running three shifts.
But the economic questions are remarkably consistent.
The Slightly Uncomfortable Conclusion
The next great robotics company may not build the smartest robot. It may build the robot that breaks less, needs fewer humans, deploys faster, works longer, pays itself back sooner, and saves the customer an embarrassingly obvious amount of money.
That doesn’t look nearly as good in a viral demo. It looks much better in a spreadsheet.
We are entering a fascinating bridge period in robotics. Humanoids may eventually become the general-purpose labor platform everyone expects. But before that happens, enormous companies can be built by specialized machines quietly dominating narrow categories, one ugly workflow at a time.
So when the next robotics pitch lands in the partner meeting, don’t fall in love with the metal. Ask how often it breaks. Ask who rescues it. Ask how long it charges. Ask what one deployed unit actually earns. Ask how quickly the hardware pays back.
And then ask the question that kills surprisingly many robotics startups: Why is this actually better than hiring a human?
Because a robot that performs a backflip is impressive. A robot that autonomously earns money is investable.
