It seems that humanid robots are overtaking the world. And you might think this statement sounds crazy. Fair enough. Stay with me and in this post I will show you some strong, structural reasons why most humanoid projects are likely to fail and disappoint their investors. If you're a humanoid robot startup investor, right now, I'm saving you from incurring losses. Thank me later.
Humans are shaped the way we are because of biological constraints, not because it's optimal for machines. Two legs are unstable and require energy-inefficient complex control. Hands with 20+ DOF are hard to actuate, fragile and expensive and upright balance means constant computation and failure modes. This last one has serious implications and I'll get to this shortly below.
Wheels, tracks, arms on rails, gantries, drones, and fixed robots on the other hand are cheaper, are faster, break less and are easier to certify.
Industry already learned this lesson decades ago.
Result: You pay 5x - 10x more to be "human-shaped" for no functional gain, just for hype!
This is the favorite Silicon Valley justification - and it's weak, oh so weak. Yes, door handles, stairs, and tools are human-scaled. But robots don't need to copy humans to interact with human spaces.
Examples:
It's far cheaper to change the environment than to build a robot that perfectly mimics a person. Dear humanoid robot startup investors. Do you hear me? Things are getting interesting.
Even if humanoid robots worked, they don't make sense financially.
Rough reality:
Compare that to a human worker, who is flexible, self-repairing, cheap or a task-specific robot that is fast, dumb, bulletproof!
Humanoids sit in the worst middle ground!
This is where hype really runs wild. And I get it. Oh those LLMs. They are to blame. They talk well, plan abstractly and hallucinate confidently - ouch!
Robots need precise perception, millisecond control loops, real-world physics handling and zero-mistake safety behavior!
A chatbot failing is funny. Trust me, when I want to relax a little bit from work, I try to drive ChatGPT in all kinds of stupidness and it's really funny. Once I made it write code and it leaked whole private GitHub repository! Funny, right?
A humanoid robot failing is property damage or injury! Imagine a humanoid robot dealing with a baby. I'd never trust my kid to one.
Bridging “reasoning” to reliable embodied action is an unsolved problem. And it might never get resolved. In computing we have certain things that are not computable. You might be pouring your hard-earned money into an impossible thing.
Biological bodies self-heal! Yes, damn it, we're a product of six billion year evolution. On the other hand humanoid robots have hundreds of failure points, need frequent calibration and degrade fast in dust, heat and moisture. They break fingers, joints and sensors constantly.
At scale, this becomes fleets stuck in repair, massive spare parts logistics and skilled technicians per robot cluster that might run wild.
This alone kills business models.
Do you honestly think that Silicon Valley kiddo is the first who's making the revolution? We've seen this movie before:
And did you have a chance to see that Russian humanoid robot that fell during the demo? That was really funny. Here's the video
Every cycle works like this: Amazing demo -> Viral videos -> Investor hype -> Pilot programs -> Silent shutdown!
The pattern is extremely consistent and you're guaranteed to get caught in this.
Humanoid robots exist mainly for PR and fundraising because they are visually impressive and sometimes funny like this poor Russian metal cra... pardon chap. They are easy to anthropomorphize, perfect for TED talks and investor decks and irresistible to media.
But virality ≠ viability.
Most humanoid startups are capital sinks, AI hype multipliers, talent magnets brand vehicles but in no way sustainable businesses!
If humanoid robots were actually useful, factories, warehouses, and hospitals would already be full of them.
Instead, we see wheels instead of legs, suction instead of fingers, rails instead of balance, software instead of bodies.
Because physics and economics always win over aesthetics!
Humanoid robot projects and startups won't "fail" overnight, but they will fade into niche irrelevance, replaced by simpler, cheaper, non-humanoid systems that actually work. That's how complexity works my friends.