Future of Humanoid Robotics

From Struggling to Walk to Running Faster Than Humans: What Comes Next for Humanoid Robots?

Around 2015, different prototypes by humanoid robotics companies often struggled to walk properly. They were bulky, chunky, and nowhere close to what we could call a finished product.

Fast forward a decade, and we have manufacturers like Unitree producing commercial-quality robots capable of performing basic movements and much more, from backflips and karate to freestyle dancing. China has even hosted the World Humanoid Robot Games, where robots competed in events ranging from athletics and football to practical tasks.

With Google and other major technology companies working on optimising gestures, motor controls, and other essential movement techniques for humanoid robots, we are moving closer to a future where humanoid robots can integrate into our existing social ecosystem. As this technology continues to progress, we can expect to see a wide range of applications for humanoid robots.

In this blog, we are going to discuss the future of humanoid robots.

From falling over to running faster than humans

In September 2026, China’s TianGong Ultra humanoid robot reportedly completed 100 metres in 8.64 seconds, faster than Usain Bolt’s 9.58-second human world record. At the World Humanoid Robot Games in Beijing, humanoids competed in running, football, table tennis and other events.

But there is an interesting contradiction. A robot can perform an impressive athletic movement and still struggle with something as ordinary as folding a shirt or using a tool.

Why?

Because the real world is unpredictable. Training for running a 100 meter race requires precise optimization, once it’s done, it’s done, but the real world doesn’t function like that.  Running along a controlled track is very different from picking up an unfamiliar object, deciding how firmly to grip it, adjusting to its weight, moving around another person and recovering when something doesn’t go as expected.

Humans do all of this almost automatically. For robots, it is an enormous challenge.

AI: The catalyst

Modern humanoids are increasingly being paired with AI systems that allow them to see, understand, plan and act.

For example, Google DeepMind’s Gemini Robotics systems, are designed to connect language and vision with physical action. Instead of programming every movement individually, a robot can be given an instruction and work out how to carry it out.

Imagine saying: “Put that object on the shelf.”

For a human, that sounds trivial.

For a robot, it involves identifying the object, understanding the instruction, navigating to it, grasping it, maintaining balance, walking to the shelf and placing it correctly.

This combination of robotics and AI is often described as integrated intelligence: intelligence that doesn’t just understand the world, but can physically interact with it.

And that could be the most important development in humanoid robotics.

In July 2026, Google DeepMind introduced Gemini Robotics 2, describing it as an intelligence layer designed to give robots whole-body intelligence, advanced dexterity and the ability to collaborate with other robots. Earlier robotic AI systems might control an arm performing a relatively contained task. Gemini Robotics 2 is designed to control an entire humanoid, from its feet to its fingertips.

Instead of following a fixed sequence, the system can interpret the instruction and act accordingly. Beyond movement, the system helps the robot to understand its environment, plan a sequence of actions, coordinate its entire body and adjust if something goes wrong. Google has also demonstrated Gemini Robotics 2 performing delicate manipulation using five-fingered robotic hands, including tasks such as tying knots and sealing ziplock bags.

The technology is not perfect. Google’s own demonstrations show that multi-finger dexterity remains challenging, with some tasks achieving considerably lower success rates than simpler pick-and-place operations. But that limitation is almost as interesting as the progress. It shows us where the frontier of robotics actually is.

Performance in Real-World Trials

The transition from impressive demonstrations to real-world work has already begun.BMW has deployed humanoid robots from Figure at its Spartanburg manufacturing facility. Figure says its robots have supported the production of more than 30,000 BMW vehicles, moving more than 90,000 components and accumulating around 1,250 operating hours during one deployment.

Boston Dynamics is moving its Atlas humanoid towards industrial deployment, while Hyundai has announced plans to build the infrastructure to manufacture tens of thousands of robots annually. Meanwhile, China’s Unitree has already brought humanoid robots such as the G1 to the commercial market.

And in September 2026, XPeng, the Chinese electric vehicle manufacturer, announced that its humanoid robot production line had begun operating, with the company targeting mass production of its IRON robot.

These developments suggest that humanoid robots are moving from concepts to real-world deployment.

The Challenges Ahead

It’s not all sunshine and rainbows in the humanoid robots industry. Reuters’ recent investigation into China’s robotics industry found that many humanoids remain slow, error-prone and dependent on highly choreographed routines. Some robots that look extremely capable in demonstrations still struggle with the unpredictability of real industrial environments.

Even TianGong Ultra’s record-breaking run illustrates this gap. After the race, all the robots that competed in the race failed to come to a halt. The organisers kept crash pads to prevent the robots from running into the audience.  The team is now focused on turning high performance into reliability and usability, with challenges including braking, adapting to different environments and making the robot’s AI capable of handling more complex tasks.

And economics will matter just as much as engineering.

A company may be able to build a robot that can walk, run and manipulate objects. But questions relating to cost, operational safety, consistency and reliability are still unanswered. Those questions will determine whether humanoid robots become a genuine industrial revolution or remain just another impressive piece of technology.

What could humanoid robots do next?

The first major opportunities are likely to appear in places where work is repetitive, physically demanding or dangerous. Factories and warehouses are obvious starting points. Robots could eventually move beyond material handling into assembly, inspection, machine operation and maintenance.

They could also find roles in construction, logistics, emergency response and other environments where human workers face physical risks. Further into the future, humanoids could enter homes, helping with cleaning, household tasks or assisting people who need physical support.

The next 10 years

Eleven years ago, the remarkable achievement was getting a humanoid robot to walk without falling. Today, robots are running faster than humans, competing in sports, entering factories and developing increasingly sophisticated AI brains.

The next breakthrough may not be another world record. It may be something much less spectacular.

A robot arrives at a mine or a factory in the morning. It understands what needs to be done. It encounters an unexpected problem, figures out how to deal with it, completes the task and moves on to the next one. No carefully choreographed demonstration. No human controlling every movement. Just a machine that can understand, adapt and work.

If humanoid robotics reaches that point, the biggest question will no longer be whether robots can imitate humans. It will be how much the world changes when machines can finally work alongside us.