A YC-backed humanoid robotics startup just died. K-Scale Labs, based in San Francisco, tried to build affordable humanoid robots. They never closed their Series A. By late 2025, it was over. Their COO recently published a brutally honest post-mortem — and every single lesson he learned the hard way is something actuator engineers have been screaming about for years.
I've spent 25 years building actuators. Before founding FIRGELLI, I worked at Rolls-Royce, BMW, and Ford. I've watched the robotics industry go through hype cycles, startup graveyards, and billion-dollar promises that never shipped a single unit. K-Scale's story isn't unique. It's a pattern. And it keeps repeating because the industry refuses to listen to the people who actually build the hardware.
Here's what they got wrong — and what the entire humanoid robotics industry needs to hear.
"The AI Will Handle It" Is the Most Dangerous Sentence in Robotics
K-Scale's COO describes something he calls "Large Model Chauvinism" — the belief that AI models are getting so good that hardware can afford to be dumb. At his startup, they genuinely debated whether to put end stops on the robot's joints. End stops. The most basic mechanical safety feature that exists — a physical limit that prevents a joint from destroying itself.
The argument against them? The AI policy should learn the joint limits. End stops are extra cost, extra weight.
Let me be blunt: this is what happens when software people build robots without listening to hardware engineers.
At FIRGELLI, every single actuator we manufacture has end stops built in. This isn't optional. This isn't a "nice to have." This is fundamental mechanical engineering. End stops exist because software fails. AI models hallucinate. Policies hit edge cases nobody anticipated. When a language model gets something wrong, you get a bad answer on a screen. When an actuator blows past its joint limit at full torque because the AI had one bad inference step, you get a broken machine — or a broken person standing next to it.
The model might be right 99.99% of the time. The end stop is for the 0.01%. And in the physical world, that 0.01% is the only number that matters.
Even Tesla, with all its autonomy ambitions, still puts brakes on the car.
Your Robot Is Not a Hoverboard
One of the most dangerous analogies in robotics fundraising is the cost curve comparison. "Humanoid robots will follow the same path as hoverboards — expensive novelty to Shenzhen mass production to commodity hardware." K-Scale bought into this. So do dozens of other startups pitching VCs right now.
Here's the problem: a hoverboard motor just needs to spin. That's it.
A humanoid robot's actuators need to be extraordinarily precise, explosively powerful, resistant to wear, and consistent unit to unit. One actuator slightly out of spec and the robot walks wrong, or falls. I've spent 25 years manufacturing actuators — the tolerance requirements for humanoid robotics are nothing like consumer electronics. Nothing.
Tesla's Optimus robot went from 11 to 22 degrees of freedom per hand in a single generation. Each degree of freedom requires at least one precision actuator. Each actuator requires rare earth NdFeB magnets that China controls over 90% of. This is not a hoverboard supply chain. This is aerospace-grade precision manufacturing at a scale that doesn't exist yet.
Analogies are compression algorithms — they make complex things simple by throwing away information. In a pitch deck, that's fine. In an engineering decision, the thrown-away information is usually the part that kills you. K-Scale learned this. Others will too.
Supply Chain Is a Capability, Not a Task
K-Scale's COO describes arriving to find no manufacturer relationships, no payment terms, no QC process, no logistics pipeline. He calls this the most common way hardware startups get into trouble. He's right.
I've negotiated with contract manufacturers across three continents for over two decades. Your relationship with your CM determines whether actuators come in within tolerance or 2mm off. Whether unit cost lands at $800 or $2,400. Whether your production timeline is six weeks or six months.
This isn't something you solve by hiring someone who speaks Chinese and pointing them at a factory. Manufacturing is a capability you build over years. At FIRGELLI, it took us the better part of a decade to build the supplier relationships and quality control processes that let us ship actuators reliably at scale. A startup trying to do this in months while simultaneously developing AI, designing hardware, and fundraising? That's not ambitious. That's delusional.
There Is No Such Thing as "Commodity" Robotics Hardware
This might be the most important lesson from K-Scale's failure, and it's one I've been saying for 25 years: there is no commodity hardware in robotics.
The narrative that robot hardware will become off-the-shelf components assembled in Shenzhen — with the real value sitting in the AI layer — is fantasy. There is no standard bill of materials for a humanoid. No off-the-shelf actuators that just work for bipedal locomotion. Every team building a walking robot right now is designing custom hardware.
When a company buys into the "hardware is commodity" story, the people building the physical product end up with less voice and less recognition than what they actually contribute. Power shifts to whichever function gets the "defensible" label — usually software — regardless of who's doing the hardest work.
K-Scale's engineers built a robot that walked. Their COO calls it the hardest engineering the company did. He's right. And yet hardware was treated as a checkbox.
The 3 D's: The Only Test That Matters
Twenty-five years ago, I coined a phrase that I've used as a filter for every automation product since: Dirty, Dull, or Dangerous. If a robot doesn't do at least one of these three things, it has no right to exist commercially.
K-Scale was trying to build affordable humanoid robots. Affordable for whom? To do what? If the answer is "walk around your house and look cool," you're building an entertainment product for early adopters — not a commercially viable robot.
The humanoid robotics market will follow a predictable path:
Phase 1 is entertainment. Early adopters buy humanoid robots for the cool factor. They're essentially walking AI assistants — a Siri with legs. They can follow you around, stand guard at the door, self-charge. But they can't do any of the 3 D's. The hands are simplified because full dexterity (an actuator per finger) is an unnecessary cost driver at this stage. This market is small.
Phase 2 is utility. This is when robots can reliably do at least one thing that's Dirty, Dull, or Dangerous. Better sensors, better AI training, better actuator control. This is when mainstream buyers show up — not because the robot is cool, but because it's useful. This is when the real market opens.
K-Scale tried to jump straight to affordable utility robots without solving the hardware fundamentals first. They're not the only ones. Half the humanoid startups funded in the last three years are making the same bet.
The Actuator Is Where the Robot Lives or Dies
Every lesson from K-Scale's failure points back to the same thing: the actuator. Not the AI model. Not the software stack. Not the pitch deck.
The actuator is where force meets the physical world. It's where precision matters. Where tolerances kill you. Where supply chains constrain you. Where safety isn't negotiable.
After 25 years of building them, I can tell you this with certainty: you cannot AI your way out of a bad actuator.
The companies that will win the humanoid robotics race are the ones that take hardware as seriously as software. That build supply chain capabilities over years, not months. That put end stops on every joint — not because the AI can't learn the limits, but because physics doesn't care what your model thinks it knows.
K-Scale's failure is a cautionary tale. But it's also an opportunity — for the companies that are willing to do the hard, unglamorous work of getting the hardware right.
The robot starts with the actuator. It always has.