ETH Zurich students build ‘first Swiss humanoid’, eye fundraising

Hardware is high-stakes. In the crypto world, we often talk about 'building' in the abstract, but for the team at ETH Zurich, building looks like 160 kilograms of metal, sensors, and actuators trying not to fall over. They call it Xalun. It is being marketed as the first Swiss humanoid, but the geography is the least interesting thing about it. What matters is the shift in how we approach robotics from a developer perspective.

The Pivot from Hydraulics to Electric Precision

For years, the gold standard for robots that move like humans was hydraulic power. It provided the raw force needed to move heavy limbs. But hydraulics are messy, maintenance-heavy, and difficult to calibrate with high-frequency software updates. The Xalun team decided to skip that legacy entirely. They built a system based on electric actuators. This is a builder-first decision. It trades raw, brute force for granular control.

When you are building a machine that needs to navigate a human environment, you don't actually need it to be a tank. You need it to be responsive. Electric systems allow for a tighter feedback loop between the AI's neural network and the actual physical movement. This is where the intersection of AI and robotics gets real. If the software can't talk to the hardware without a massive lag or mechanical friction, the 'intelligence' part of the robot is useless.

Learning to Walk in Simulation

One of the most impressive parts of the Xalun project isn't the metal itself, but the training ground. The team utilized reinforcement learning in simulated environments. In the old days of robotics, you would write thousands of lines of explicit code telling a robot exactly how high to lift its foot. If the floor was a different texture, the robot failed.

The ETH students are leaning into the modern stack: let the robot fail ten million times in a virtual world before it ever takes a step on a laboratory floor. This is exactly how we see LLMs being trained, except the output isn't a string of text; it's a stable center of gravity. For founders, the takeaway here is clear: simulation is no longer a luxury. It is the only way to scale hardware development without breaking your budget on physical prototypes.

Why Switzerland?

Switzerland has a reputation for precision—watches, banking, pharmaceuticals. But it is also becoming a concentrated hub for robotics. Being a student at ETH Zurich provides a unique moat. You have access to deep technical talent and a regulatory environment that isn't quite as hostile to experimental hardware as other regions. The team behind Xalun used their university foundation to bridge the gap between academic theory and a viable startup model.

They are currently looking at fundraising, which is the true test for any student project. Moving from a lab to a recurring revenue model is where most humanoids die. There is a massive 'valley of death' between a robot that can walk in a demo and a robot that can move boxes in a warehouse for 12 hours a day without an engineer standing by with a remote control.

The Problem with the 'General Purpose' Dream

The industry is obsessed with 'general purpose' robots. Everyone wants the Rosie the Robot from the Jetsons. But as a founder, I’m skeptical of the 'do everything' pitch. The Xalun team is aiming for versatility, but the real wins in robotics usually come from hyper-specific applications. Whether it's logistics, inspection, or hazardous environment navigation, the hardware needs a job.

Building a humanoid just because it's the shape of a human is a massive engineering tax. We have chairs, stairs, and handles designed for our specific anatomy, so the humanoid form factor makes sense for navigating our world. However, the complexity of balancing two legs is exponentially harder than four legs or wheels. The Xalun team is taking the hard road. If they can solve the balance and battery life issues using an all-electric kit, they might have a platform that others can build on top of.

Building for the Ecosystem

What I find most compelling about this group of 20-somethings is their approach to the tech stack. They aren't trying to invent a new type of screw. They are using existing high-end components and focusing their 'alpha' on the software integration. This is the same pattern we see in successful crypto projects: don't rebuild the base layer if you don't have to. Focus on the middleware and the user experience.

For Xalun, the user experience is the robot's ability to operate autonomously. If an operator has to babysit the machine, it's just an expensive toy. The goal is a machine that can take a high-level command—'go to the loading dock'—and figure out the millions of micro-adjustments needed to get there. That requires a massive amount of onboard compute power and a very efficient battery management system.

The challenge isn't making a robot walk; it's making a robot walk when it doesn't know what the floor looks like.

The Founder's Takeaway

If you are looking at the robotics space, don't get distracted by the shiny exterior of the Xalun robot. Look at the efficiency of their development cycle. They moved from concept to a functioning humanoid in a fraction of the time it took legacy players like Boston Dynamics to reach similar milestones. This is a testament to the democratization of AI tools and simulation software.

The era of the 'garage hardware' startup is coming back, but this time, the garage is full of GPUs. For builders, the lesson is simple: stop waiting for 'perfect' hardware. Use what is available, focus on the neural networks that control the movement, and move fast. The Swiss might be known for being slow and steady, but this team is sprinting.

Final Thoughts for the Skeptics

We see a lot of vaporware in AI. A lot of videos that are edited to make robots look more capable than they are. The Xalun project feels different because of its academic transparency. They aren't hiding behind a marketing firm. They are showing the struggle of weight distribution and the limitations of current battery technology. That honesty is what will eventually attract the right kind of capital—not the 'hype' money, but the 'build' money.

Keep an eye on the electric actuator space. If Xalun proves that you don't need hydraulics to move 160kg efficiently, the cost of entry for humanoid robotics is going to plummet. When the cost drops, the number of developers increases. That is when the real revolution starts.

Read the original at Sifted →