The Robotics Companies No One's Talking About (Yet)

Humanoid robots are everywhere in 2025. They’re making news, going viral in demo videos, and showing up in places like hospitals in China. A robot marathon recently took place, and a recent robot boxing event made headlines. The front-end of robotics is capturing global attention.

But the more important story is what’s happening behind the scenes.

Think back to the AI boom. Everyone focused on ChatGPT. But the biggest winner was NVIDIA. The value accrued to the infrastructure. The same pattern might be playing out in humanoid robotics.

This blog looks at both sides: the startups building humanoid robots that grab attention, and the lesser-known companies creating the systems that power them, like chips, sensors, control software, and physical intelligence.

If you want to understand where the machine economy is really heading, and where the alpha may be, look beyond the robots.

‍

The Companies Rebuilding Intelligence in Physical Form

Figure AI

We’ll start with Figure AI because chances are you’ve probably heard of them. With major backing ($1.5B) to be exact, and an ambitious rollout plan, Figure AI is building a vertically integrated stack around general-purpose humanoids. 

Figure AI have already shown their robot (F.02) responding to voice commands and doing basic tasks like picking up household items. That’s thanks to Helix, their own vision-language-action model. It’s one of the more complete humanoid stacks out there, following a strategy very similar to Tesla Even with a high valuation, if they’re first to scale, they could set the standard for what’s to come.

Apptronik

Apptronik, based in Texas, is one of the top contenders in the humanoid space. Their Apollo robot is already performing trial work with Mercedes, and they have also a partnership with Jabil, a major manufacturing company, that could help them scale

The most critical partnership is the one they have  with Google DeepMind to accelerate the development of humanoid robots minds. Currently, Apollo’s being trained to handle inspection, sorting, and other practical jobs in factories and warehouses.

Read our latest blog fo a further look into Apptronik's humanoid robots.

Sanctuary AI

We’ve looked at companies developing the complete humanoid body, but what about those focused on more specific, niche elements? Sanctuary AI is not only building humanoids, but their work is heavily focused on developing the perfect robotic hand. Their advanced robotic hands are designed to mimic human hand capabilities. These hands, integrated into their Phoenix humanoid robot, feature 21 degrees of freedom (DOF) and utilize miniaturized hydraulic valve actuators. What is most impressive is the large patent portfolio that they have accumulated over the years.

That includes grip, touch, and motor control, key components for real-world interaction. Geordie Rose, Co-Founder of Sanctuary strongly believes that “general purpose robots must be grounded in the world the way humans are.” 

1X Technologies

1X Technologies is a little different. Their approach is a mix of autonomy and remote control. If you imagine a future where robots are semi-autonomous but can still be guided by people, 1X is building for that world. 

They’re rolling out trials for their NEO Gamma humanoid to work inside homes. Although instead of aiming for 100 percent independence from these machines, they’re making sure the robot, when needed, can always rely on a human operator when they get into tricky situations. The remote control technology will be critical for mass deployments and it’s no surprise that they are in partnership with NVIDIA. 

Neura Robotics

Based in Germany, Neura Robotics is already selling non-humanoid robots across multiple use cases. They’ve got customers, real deployments, and a growing base of enterprise partners. 

Now they’re taking that experience and applying it to humanoids. Their new robot, 4NE-1, can see, hear, and even sense touch. Neura calls it “the first cognitive robot platform for real-world environments,” and while that’s a strong statement, they will be put to the test with their coming deployment of the MiPA, an “Intelligent Personal Assistant for Work and Life”. It’s a robot on wheels with arms and hands. 

Physical Intelligence 

This team is building π0 (pi-zero), which is basically a foundation model for robots. Think of it as what GPT is for language, but for physical interaction. Physical Intelligence believes robots shouldn’t just follow instructions, they should be able to reason about them in context. It’s still early, but this is a serious attempt  to bring large-model intelligence into the real world.

‍

Skild AI

Skild AI is on a similar path, training what they call the Skild Brain. Their focus is on continual learning. They want to develop robots that get better over time by learning directly from their actions. The Skild Brain is built to work across different types of robots and tasks, combining locomotion, manipulation, and navigation into one transferable intelligence layer.

The Brains Behind the Humanoids

Even the smartest robots need chips that can keep up. These next companies are working at the silicon layer, and they may end up powering every robot that comes after.

Groq

Groq makes ultra-fast AI chips that are optimized for inference. That might sound like a backend tool, but the faster and more efficient robots get, the more important Groq becomes. Groq chips are already being used in AI systems, and if they end up inside physical robots, their upside is huge. 

What makes Groq stand out is how energy-efficient their chips are. This is going to matter when you’re trying to run a robot that needs to make decisions on demand without burning through a battery in ten minutes.

Etched

Etched is taking a slightly different route. They’re building chips specifically for transformer models, the kind that power LLMs and, more recently, in the AI systems behind robots.

By focusing on this kind of processing, they’re developing hardware that can support the kind of ‘on the spot’ decision-making these systems demand. Those chips are likely going to be used for everything related to Physical AI : self-driving cars and humanoid robots. 

‍

Precision In Robotics

To define perfect precision in a robot whether it’s movement or grip might just be as hard as building the entire thing. That’s why these companies are focusing entirely on that part alone.

Proception AI

This company is all about perception, especially robotic hands. They’re building robotic hands with touch feedback while trying to close the gap between human grip and robotic precision. If you’ve ever watched a robot fumble with an object, you know how far we still have to go, and Proception AI is working on this particular problem.

‍

SharpaWave 

SharpaWave is a bit more under the radar. We don’t know much yet, but early signals suggest they’re also working on hand control or fine-motor perception. Could be in its early stages, but worth keeping an eye on.

So that’s the list. Not just the names you already know, but the ones building the less visible layers bringing humanoid robotics to the next phase of real-world use.

‍

Bullish on robotics? So are we.

XMAQUINA is a decentralized ecosystem giving members direct access to the rise of humanoid robotics and Physical AI—technologies poised to reshape the global economy.

Join thousands of futurists building XMAQUINA DAO and follow us on X for the latest updates.

‍