Humanoid robots: stop funding the body before the job
Humanoid robots need one paid job before they deserve general-purpose labor claims. Use this founder filter before you build or fund one.
Most humanoid robot hype is investor theatre with legs.
That sounds harsh until you ask the question that kills weak robotics ideas fast: what job does the human-shaped body do better, cheaper and safer than a cart, arm, drone, camera rig or fixed machine?
If the answer is vague, the robot is probably a costume for a funding story.
TL;DR: Humanoid robots are human-shaped machines built for environments designed around people. They may matter in factories, warehouses, hospitals, care homes and homes when the human form lets them use existing doors, stairs, shelves, tools and workflows. The trap is pretending "general-purpose labor" exists before one narrow paid job works. Founders should choose one site, one task, one buyer metric, one safety plan and one repair model before building a full humanoid company.
I am Violetta Bonenkamp, founder of Mean CEO, CADChain, and F/MS Startup Game. I like hard tech. I do not like hard tech dressed as a magic labor story for people who have never had to repair anything with a deadline and an angry buyer watching.
Humanoid robots might become a serious category.
They are also one of the easiest ways to burn money while looking impressive.
What Humanoid Robots Actually Are
Humanoid robots are robots shaped roughly like people.
Usually that means a torso, head, two arms, hands and legs, although some industrial versions use wheels instead of feet. The promise is simple: if the world was built for human bodies, a robot with a human-like form may work in that world without rebuilding every room, shelf, door, handle, fixture and tool.
The International Federation of Robotics paper on humanoid robots frames the vision as general-purpose robots based on human motion mechanics, while warning that mass adoption remains uncertain. Its humanoid robot infographic also lists the tradeoffs founders should stop hiding: limited battery life, safety risks from falls, small production volumes, public discomfort and weaker speed, precision and repeatability than many traditional industrial robots.
That last point matters.
Humanoid robots are not better because they look like us.
They are better only when the human shape avoids a more expensive site redesign or lets the machine use tools, spaces and movement patterns already built for people.
This is why the earlier piece on robotics startups moving beyond warehouses matters for this article. The founder question stays the same: what narrow job will a buyer fund before the story expands?
The General-Purpose Labor Trap
"General-purpose labor" is a seductive phrase.
It makes investors imagine one robot doing warehouse picking, elder care, cooking, factory support, hotel cleaning, shelf scanning and home chores. It makes founders imagine a huge market before one paid task survives a Tuesday shift.
The phrase hides three problems.
First, labor is not one thing.
Lifting a box, loading sheet metal, helping an older person stand, folding laundry, scanning shelves and carrying sterile supplies are different jobs with different risks, buyers, workflows, spaces and blame paths.
Second, the human body is expensive to copy.
Balance, hands, wrists, perception, battery, motors, force control, heat, falls, cabling, repair access, cleaning, safety and operator recovery all add cost.
Third, buyers pay for results, not resemblance.
A factory buyer does not wake up wanting "a humanoid." The buyer wants fewer unsafe lifts, more stable material flow, less awkward manual work, cleaner records or a way to keep production moving when hiring is hard.
This is where physical AI for manufacturing and field operations becomes the useful lens. AI in a body is interesting only when the body performs paid work in a real site.
Where Humanoids Are Getting Real First
The least ridiculous humanoid use cases are not in homes.
They are in factories, warehouses and logistics sites where there is already paid physical work, controlled access, repeatable tasks and a buyer who can measure whether the robot helped.
Agility Robotics positions Digit for manufacturing, distribution and logistics, which is a smart wedge because those sites can test repetitive material movement. Figure’s BMW deployment recap says Figure 02 ran weekday shifts, loaded sheet-metal parts and generated plant data during an 11-month deployment. BMW’s Leipzig humanoid robot pilot also points to production and intralogistics as the near-term arena for physical AI.
That does not mean every humanoid startup should copy those companies.
It means founders should learn from the pattern:
- Pick a controlled site.
- Pick a repetitive job.
- Measure human resets.
- Measure task time.
- Measure error rate.
- Learn which part breaks first.
- Price support from the start.
Notice what is missing.
Nobody serious starts with "the robot will do everything."
The Body Only Wins In Human Spaces
A humanoid body makes sense when the environment is hard or expensive to change.
Factories already have aisles, fixtures, racks, tools and stations designed around workers. Hospitals have doors, lifts, carts, supply rooms and mixed human traffic. Homes have stairs, kitchens, bathrooms, shelves, clutter and furniture. Retail stores have aisles, shelves, stock rooms and customers.
That is the pro-humanoid argument.
Now the founder has to earn it.
If a wheeled autonomous mobile robot can move the tote, use the wheeled robot. If a fixed arm can load the part, use the arm. If a camera rig can inspect the site, use the camera rig. If a drone can scan the roof, use the drone. If a simple cart saves staff time, the humanoid has already lost.
Humanoid robots should compete only where:
- The space is built for human reach, height and movement.
- The job changes enough that a fixed machine becomes awkward.
- The buyer cannot redesign the site cheaply.
- The task is repetitive enough to train and support.
- The safety case can be explained in plain language.
- The service model survives real use.
If those are missing, the humanoid body is vanity hardware.
Humanoid Robot Founder Filter
Use this table before you build, invest, hire or pitch.
Uses fixtures, racks and spaces made for people
Load one part type into one station
Fewer resets and steady task time
A fixed arm can do it cheaper
Works in aisles and human stations
Move one tote path repeatedly
Safe passes near workers and fewer manual lifts
A cart or AMR solves the job
Reaches shelves and moves around shoppers
Scan one shelf category after hours
Cleaner stock data and fewer missed items
The robot becomes a novelty host
Uses doors, lifts and supply rooms
Move supplies between two points
Staff time returned without new risk
Infection or privacy rules are ignored
Works in spaces built for residents and carers
Carry, fetch or remind around one routine
Staff relief and resident acceptance
The pitch drifts into medical claims
Handles spaces that change daily
Do one chore in one room type
Fewer failures than the existing appliance
The buyer needs a manual
Handles stairs, tools and uneven areas
Carry small tools or scan one defect type
Fewer repeated trips and cleaner records
Rugged special robots are better
Uses human access paths and tools
Repeat one inspection route
Same report format every run
A camera on wheels works fine
Helps robot teams train and test skills
Teleoperation or simulation workflow
Faster learning for one task
The buyer still needs custom hardware
Sells hands, actuators or sensors to builders
One part that fails less often
Repeatable manufacturing and repair data
One big buyer controls your fate
This table is deliberately rude.
It should be. Humanoid robots deserve rude questions before they deserve founder savings.
The Technical Problems Founders Underprice
Humanoid robots fail in ways software founders are not used to paying for.
They fall. They wear out. They drift out of calibration. They overheat. They break fingers, wrists, forearms, cables, joints and sensors. They need batteries, chargers, cleaning, service visits, spare parts and operator training.
They also create human trust problems.
People may be fascinated by a robot in a video and uncomfortable with the same robot in a corridor. The IFR infographic points to public fear and the "uncanny valley" as a real barrier. That is not soft psychology. It can decide whether a care home, school, hospital, shop or factory lets your robot stay.
The EUR-Lex text of Regulation 2023/1230 on machinery is another reminder for European founders: machinery placed on the European market must meet health and safety requirements, and the regulation better covers technologies such as autonomous mobile machinery and AI-related safety functions.
Here is the uncomfortable founder math:
- A human-shaped robot needs more moving parts.
- More moving parts create more repair risk.
- Repair risk raises service cost.
- Service cost changes pricing.
- Pricing changes which buyers can afford the product.
- Buyer affordability decides whether the company is real.
This is why deep tech founders should keep CADChain’s robotics CAD protection guide in mind. A humanoid robot is not one file. It is a stack of mechanical design, suppliers, software, test data and IP that can leak or get copied while the company is still learning how to sell.
Business Models That Make More Sense Than Magic Labor
Bootstrapped founders should avoid selling a whole humanoid dream too early.
Better entry points include:
- A paid factory pilot for one repeated material task.
- A repair or maintenance service for humanoid fleets.
- Teleoperation services for one robot category.
- Simulation workflows for one task family.
- Hands, grippers, actuators or sensor modules.
- Safety testing support for legged robots.
- Data collection for one physical skill.
- Operator training for warehouses and factories.
- CAD file protection for robotics supplier networks.
- Content and buyer education for a narrow robotics niche.
There is nothing embarrassing about selling the picks and shovels of humanoid robotics.
In fact, it may be the smarter bootstrapped move.
The companies building full humanoid bodies need capital, hardware talent, factories, safety teams, supplier access and years of field learning. A smaller founder can sell the narrow layer those companies need: data, tools, simulation, components, repair, documentation, testing or IP control.
If you still want to build the full robot, use the F/MS Startup Game mindset: test demand by making the buyer act. A founder who cannot sell a narrow pilot will not be rescued by adding knees.
The 60-Day Founder Test
Here is a practical test before you commit to humanoid hardware.
Day 1 to 5: name the paid job.
Write it without drama. "Load sheet-metal parts into one fixture." "Move totes from one station to another." "Carry sterile supplies from storage to one ward." "Scan shelves after closing." If the job sounds like science fiction, shrink it.
Day 6 to 12: prove the human-shape reason.
Write why a humanoid body is needed. Stairs. Door handles. Human-height shelves. Existing tools. Mixed traffic. Reach. Two-arm handling. If a simpler robot can do it, stop.
Day 13 to 20: watch the work.
Track task time, handoffs, awkward movements, injury risk, error points, reset moments, worker workarounds, lighting, surfaces, storage, noise, cleaning rules and battery windows.
Day 21 to 30: sell a paid test.
Ask the buyer to pay for one small trial with one success metric. Not applause. Not press. Payment.
Day 31 to 45: define the failure script.
What happens if the robot falls, blocks an aisle, drops a part, misses a shelf, loses network, scares a customer, overheats, breaks a hand or needs cleaning? A serious buyer will ask. A serious founder answers.
Day 46 to 60: price the service.
Include setup, training, support, spare parts, repair time, data review, insurance assumptions, removal terms and staff time. The robot price alone is fantasy accounting.
If the buyer still wants the test after seeing the full service model, you may have a company.
What European Founders Should Do Differently
Europe does not need to copy the loudest humanoid robot race.
Europe can build around safety, industrial buyers, component skill, robotics research, manufacturing know-how, public funding and real buyer proof. That is less glamorous than announcing a household robot. It is also more likely to survive contact with procurement.
The better European wedge may be:
- Factory and warehouse pilots.
- Safety and testing tools.
- Components for hands, sensors and joints.
- Simulation and data workflows.
- Robot repair and maintenance.
- Machine vision for physical tasks.
- Industrial training data.
- CAD and supplier file protection.
- Paid consulting that becomes a product.
This connects to European deep tech because hard technology fits Europe when founders pair technical depth with commercial discipline. It also connects to AI inspection and factory maintenance because many buyers will pay sooner for narrow factory tools than for a full humanoid workforce.
Use grants if they buy time.
Use pilots if they buy truth.
Use both if you can, but do not let public funding turn the company into a paperwork machine.
The Bottom Line
Humanoid robots may become useful.
Some already are useful in narrow industrial settings.
But "general-purpose labor" is still a claim founders must earn one paid task at a time. The shape is not the product. The job is the product. The service model is the product. The safety case is the product. The repair plan is the product.
If a simpler machine does the work, build the simpler machine.
If the human shape truly earns its keep, prove it with a buyer who pays before the hype does.
FAQ
What are humanoid robots?
Humanoid robots are robots shaped roughly like human beings. Most have a torso, head, arms, hands and legs, although some use wheels for movement while keeping a human-like upper body. The reason for that shape is practical: many workplaces, homes, hospitals and shops are designed around human bodies. A humanoid robot may use existing shelves, doors, tools, stairs, handles and workstations without forcing the buyer to rebuild the whole site.
Why are humanoid robots attracting so much attention?
Humanoid robots attract attention because they combine AI, robotics, labor shortages and science-fiction imagery in one dramatic package. Factories and warehouses want help with repetitive physical work. Care systems face aging populations and staff pressure. Investors see a possible new hardware category. The danger is that attention can arrive before buyer proof. A founder should separate market excitement from paid adoption.
Are humanoid robots ready for general-purpose labor?
Humanoid robots are not ready for broad general-purpose labor in the way many headlines imply. Some are being tested or used in narrow factory, logistics and material-handling tasks. That is very different from one robot doing every job in a workplace or home. General-purpose labor will require better safety, battery life, repair models, dexterity, perception, task learning, buyer trust and service economics.
Where will humanoid robots work first?
Humanoid robots are most likely to work first in controlled industrial sites: factories, warehouses, distribution centers and logistics spaces. These sites already have paid physical work, measurable tasks and staff who can be trained. Homes and care settings may come later, but they are harder because the environment changes constantly and the trust bar is higher.
Should bootstrapped founders build humanoid robots?
Most bootstrapped founders should not start by building a full humanoid robot. The body is expensive, slow and hard to maintain. A smaller founder may have a better chance selling a narrow layer: components, simulation tools, data workflows, teleoperation support, maintenance services, safety testing, CAD file protection or a specialized robot for one task. Build the full body only when the human form clearly beats simpler machines.
What is the safest first use case for humanoid robots?
The safest first use case is usually a repetitive task in a controlled site where failure can be managed without harming people. Material handling in a factory or warehouse is a good example when the robot works away from crowded public spaces, moves known objects and follows clear routes. Patient care, public spaces and homes raise the trust and safety bar much higher.
How do humanoid robots differ from industrial robots?
Traditional industrial robots are usually fixed arms or specialized machines built for speed, precision and repeatability in designed work cells. Humanoid robots are mobile, human-shaped and meant to work in spaces built for people. That flexibility comes with tradeoffs: more moving parts, harder balance, more safety questions, more service work and lower speed or precision than a machine built for one fixed task.
What are the biggest technical limits of humanoid robots?
The biggest limits include battery life, balance, hand dexterity, safe movement near people, perception in messy spaces, heat, repair access, part wear, task training and recovery after failure. A robot that looks good in a video may still struggle with repeated work across long shifts. Founders should track resets, errors, repairs and operator interventions from the first pilot.
How should a founder validate a humanoid robot idea?
A founder should validate a humanoid robot idea by choosing one buyer, one site and one physical job. Then watch the current human workflow, prove why the human-shaped body is needed, price a paid pilot and write the failure recovery plan. If the buyer will not pay for a narrow test, the founder should not add more features. Payment is better evidence than applause.
What should European founders watch before entering this market?
European founders should watch safety rules, machinery requirements, public funding incentives, buyer procurement, component supply, IP protection and service cost. Europe can do well in industrial robotics, components, testing, simulation, safety, CAD protection and field services. The mistake would be copying the loudest humanoid robot hype while ignoring the boring buyer work that makes hard technology survive.