Humanoid Robots at CES 2026: From Demo to Reality

For years, humanoid robots at CES were primarily exercises in technological showmanship. Fascinating, highly choreographed machines, often closer to performance than to real-world applications.

CES 2026, however, marked a clear turning point. This year, humanoids did more than simply “put on a show”: they started to work in real-world contexts, with clear industrial roadmaps and, in many cases, products already shipping or scheduled for deployment within the year.

Looking at the different companies and examples presented, one first impression is hard to ignore: we are no longer looking at distant promises, but at a discipline that is beginning to take a stable form in the real world.

As designers, what stands out is not just the technological leap, but the shift in design intent: less spectacle, more system; less wow factor, more usefulness.

A geography that speaks volumes

The distribution of companies at CES tells us much more than it might seem. The centre of gravity of humanoid robotics has clearly shifted towards Asia and, in particular, China, which accounts for around 60% of the companies and startups seen at CES 2026. These are not simply prototypes, but products designed to be mass-produced, sold, rented and upgraded.

South Korea and the United States follow with a similar presence, at around 15% each, often represented by highly structured players with strong integration into major artificial intelligence platforms.

Europe and Japan together account for the remaining 10%, with a more selective and targeted approach, placing greater emphasis on safety, engineering quality and design for shared environments.

This map reflects different industrial models and different visions of the relationship between technology, the market and society.

No longer spectacle, but intent

The real paradigm shift at CES 2026 is not just about hardware. It is about intent.

Atlas by Boston Dynamics is perhaps the most emblematic example: no longer a research platform or an “heroic” machine, but a robot designed to actually work. With continuous autonomy, industrial lifting capabilities and real-world deployments planned at Hyundai facilities, Atlas is not trying to look human. It is trying to be useful, reliable and easy to integrate.

At the opposite end of the spectrum, but with the same clarity of purpose, companies such as Unitree presented a complete range of humanoids already designed for scalability and the mass market. The fast, almost aggressive demonstrations of the G1 are not simply marketing: they reflect an obsessive focus on balance, motor control and cost, supported by a Robot-as-a-Service model that openly targets global adoption.

Italy also deserves a special mention, with GENE.01 by Generative Bionics, presented at CES alongside semiconductor giant AMD. It is an important signal for the European and Italian ecosystem, not so much because of the robot’s visual impact, but because of its modular architecture and native integration with advanced computing platforms. GENE.01 is positioned as a general-purpose humanoid, with a strong focus on human-machine collaboration and application-specific customisation rather than athletic demonstrations.

Outside CES itself, but impossible to leave out of the broader conversation, there is Tesla Optimus. Although Tesla did not present any specific new developments at the show, it remains an important reference point, not so much for what it is demonstrating today, but for its systemic ambition to bring humanoid robotics into an existing industrial, software and manufacturing ecosystem. Optimus remains a promise to watch, indirectly influencing the entire sector.

The humanoid enters human spaces

One of the strongest signals from CES 2026 was seeing humanoids leave the factory floor and enter environments where the body, gesture and human interaction become central.

In healthcare, LEM Surgical’s Dynamis demonstrates how a humanoid form factor can translate into surgical precision, ergonomics and adaptability. Not a prototype, but a system already authorised and in clinical use. Here, the humanoid does not imitate humans; it enhances their capabilities.

In care and assistance, Fourier’s GR-3 presents a radically different image: soft surfaces, controlled movements and a body language designed to feel reassuring. It is a robot designed to stay alongside people, rather than in front of or above them, operating along the fine line between function and relationship.

In public spaces and hospitality, robots such as AgiBot A2 are beginning to appear in museums, showrooms and reception areas. They are not always perfect, and are often still supervised, but they are already present, visible and integrated into everyday flows.

The home as the next testing ground

The domestic environment remains perhaps the most complex of all. CES 2026 showed, with a certain degree of honesty, just how difficult it is to bring a humanoid into our homes.

LG CLOiD is one of the more mature attempts. Rather than promising to “do everything”, it demonstrates its ability to perform a number of key tasks in real-world scenarios, such as folding laundry or loading a dishwasher. More than a robot, it appears to be a physical extension of the smart home ecosystem.

SwitchBot Onero H1, by contrast, abandons the bipedal form altogether in favour of stability and precision. It is a very clear design choice: less anthropomorphism, more reliability. A domestic assistant conceived as an advanced appliance rather than a theatrical presence.

Alongside these examples, a constellation of hybrid and specialised solutions suggests an important truth: the home will not have a single humanoid robot, but many different devices, each designed around a specific task.

The real enabler: Artificial Intelligence and sensing

What truly sets CES 2026 apart from previous editions is not so much the number of humanoid robots on display, but the quality of the systems that make them operational. The value no longer lies exclusively in mechanics, but in the deep integration of advanced sensing and generative Artificial Intelligence, transforming humanoids from execution platforms into systems capable of perceiving, interpreting and acting in the real world.

We are no longer looking at rigidly pre-programmed sequences. Robots such as Boston Dynamics’ Atlas demonstrate how AI may finally address one of the historical limitations of humanoid robotics: adapting to unpredictable environments. Through its collaboration with Google DeepMind and models such as Gemini Robotics, Atlas is designed not merely to walk or lift objects, but to understand the physical environment and learn from experience.

Sensing plays an equally crucial role. 360-degree vision, tactile sensors in the hands and perception systems distributed throughout the body allow robots to build a continuous representation of their surroundings, recognise people and objects, and adapt complex movements safely. These are no longer simply “eyes and ears”, but an integrated perceptual infrastructure in which real-world data and neural models work together to enable fast, reliable and context-aware decisions.

A discipline that concerns us all

CES 2026 paints a clear picture: humanoid robotics is no longer a technological curiosity. It is becoming a discipline that is redefining the way we design interactions, services and responsibilities.

For those working in design, the question is no longer whether these robots will work, but how we want them to exist in the world. What kind of presence will they have? What kind of body language will they use? What kind of relationship will they build with people?

The future is no longer behind glass. It is in the middle of the crowd. A little imperfect, sometimes unstable, but undeniably real.