Chinese Robots Can Do Backflips, Now It's a Race for the AI "Brain

Have you ever watched a video of a humanoid robot running, jumping, doing backflips, or even performing martial arts and wondered: "That's impressive, but can it actually do anything useful?" That question isn't just on your mind. It's the exact question that an entire multi-billion-dollar industry is trying to answer.

For years, Chinese robotics companies have amazed the world with impressive hardware demonstrations. Their robots can run fast, jump high, do backflips, and execute movements requiring extreme balance and complex coordination. But now, as hardware becomes more commoditized and prices drop, a new race is beginning—a race not focused on limbs, but on the brain.

And that race is being fueled by an unprecedented wave of capital. According to data from research firm ITjuzi, Chinese embodied AI companies—the software sector allowing robots to understand their environment, learn new tasks, and adapt without individual programming for each chore—raised 93.5 billion yuan (approximately $13 billion USD) across 322 deals in the first half of 2026. This figure is five times higher than the same period last year.

Join me, SilverZ, as we dissect this race from its roots. Because behind the investment figures lies a simple yet brutal truth: hardware is no longer a competitive advantage. The brain is everything.

When "Muscle" Is No Longer Enough to Win

Imagine owning a robot that can run marathons, leap over obstacles, and perform beautiful acrobatic maneuvers. That sounds amazing. But if you put it in a factory and ask it to assemble a component, or inside a warehouse and ask it to sort goods, that robot will likely stand still—or worse, break things.

That is the exact problem facing the humanoid robotics industry. For years, companies have focused on perfecting hardware: stronger motors, more flexible joints, and more precise sensors. And they have succeeded. Today's Chinese robots can perform movements that a decade ago only existed in science fiction movies.

But once the hardware is good enough, a harsh reality emerges: a perfect body without intelligence is just an expensive toy. And robot manufacturers have realized this.

Dermot McGrath, founder of Shanghai-based tech advisory firm ZenGen Labs, offered a sharp observation: "As competition drives down hardware costs, robot makers have concluded that profit margins won't come from hardware, but from the brain."

That is why investment capital is shifting heavily from hardware manufacturers to robot software developers. And that is why embodied AI is becoming one of the hottest fields in global technology.

The Robot Brain: The Missing Piece

So what exactly is a robot "brain," and why is it so difficult to develop?

Think about how a child learns to walk. The child doesn't need anyone to program every single step for them. They observe, experiment, stumble, and gradually adjust their body to walk without thinking. It's a complex learning process that combines sensory, motor, and cognitive functions.

Robots need to learn in a similar way. But while chatbots can learn from billions of pages of text on the internet, robots need to learn from physical interactions—holding objects, handling soft wires, moving in tight spaces, and reacting to unexpected situations. And this type of data doesn't readily exist online. It must be generated, collected, and processed.

Gao Yang, co-founder of Spirit AI, made no secret of this difficulty: "The brain is truly the weakest link in the entire robotics tech stack."

That is why Chinese companies are pursuing various approaches to collect training data. Spirit AI has hired about 1,000 people to record daily and industrial activities. Another company, Robbyant, stated it trained its model using 160,000 hours of footage, including 50,000 hours of human-guided robot movements. Some other developers use computer simulations to teach robots industrial tasks—though simulations cannot perfectly replicate real-world conditions.

The Names Leading the Race

The robot brain race is not a playground for amateurs. It requires a combination of advanced artificial intelligence, high-quality data, and real-world testing capabilities. And several Chinese companies are emerging as heavy contenders.

Spirit AI: When Robots Learn from Humans

Spirit AI is one of the most watched companies in China's embodied AI sector. With its Moz1 model, the company is developing software that allows robots to learn new tasks through observation and imitation. With a data collection team of about 1,000 people, Spirit AI is building one of the largest human movement datasets in the industry. The company is valued at approximately $3 billion USD.

X Square: From Code to Common Sense

X Square is another notable name. The company published research showing that robots can acquire object manipulation skills from just a single human video, without needing to retrain the model for each task. This is a crucial step forward because it allows robots to learn faster and more flexibly.

Yang Qian, co-founder and COO of X Square, shared the company's vision: "We want robots to shift from being controlled by code to being controlled by common sense." X Square is also valued at approximately $3 billion USD.

Galaxea: Another Formidable Contender

Galaxea, alongside X Square and Spirit AI, forms China's top trio of embodied AI startups, with each company reaching a valuation of roughly $3 billion USD. The presence of multiple highly valued companies demonstrates that investors are placing massive bets on the sector's potential.

The Giants Aren't Sitting Out Either

Not just startups, but traditional robot manufacturers are also heavily investing in AI. UBTech, a Hong Kong-listed company, posted a job opening for a chief scientist position heading AI model development with compensation reaching up to 124 million yuan (approximately $18 million USD), including stock options and benefits. Unitree, one of the world's largest humanoid robot manufacturers by shipment volume, reportedly has about 30 researchers developing AI to predict object behavior.

The US-China Race: Who Will Win?

The robot brain race is not just an internal Chinese story. It is a new front in the US-China tech rivalry, and both sides have distinct advantages.

On the US side, startups like Physical Intelligence, Skild AI, and Generalist are developing software that allows robots to learn various skills without excessive retraining. Generalist claims their GEN-1.5 model can complete about 60% of simple tasks after watching a demonstration lasting 3 to 12 seconds. US companies also benefit from advanced AI research and access to powerful computer chips—advantages that Chinese rivals are restricted from due to sanctions.

However, China is not without weapons. Chinese companies can work directly with robot manufacturers and industrial clients to test and improve their systems—an advantage that US companies struggle to secure. They also have access to massive amounts of real-world training data from factories and warehouses across China.

Georg Stieler, a robotics consultant, offered a balanced assessment: "Top US laboratories are setting a rigorous standard, especially regarding how fast robots can acquire and reuse skills." However, he also acknowledged that China's strongest startups are formidable competitors.

And there is one factor nobody can ignore: manufacturing scale. China has long been the world's factory in many fields, and humanoid robots may prove to be no exception. If Chinese companies can combine hardware manufacturing capabilities with sufficiently powerful AI brains, they could create a competitive advantage that rivals will find difficult to close.

Why This Race Matters to All of Us

You might be thinking: "This is a story about robotics companies and investors, what does it have to do with me?"

The answer is: a lot.

First, humanoid robots have the potential to change how we work. From factories and warehouses to hospitals and even households, robots can take on dangerous, boring, or high-precision jobs. When robot brains are smart enough, they won't just follow orders but will also be able to self-learn and adapt to new situations—something current robots cannot do.

Second, this race will reshape global supply chains. The country that leads in smart robotics will gain a major advantage in production automation and logistics. This could impact employment, goods prices, and the economic standing of many nations.

Third, this is a story about the limits of technology. We are accustomed to AI writing text, drawing pictures, and answering questions. But teaching a machine to understand and interact with the physical world is an entirely different challenge. Success in this field will usher in a new era of AI—an era where machines don't just talk to us but work alongside us.

Final Thoughts: When "Muscle" Is Achieved, "Intelligence" Is the Final Stretch

I have watched the robotics industry go through many developmental stages over the past two decades. From bulky industrial robots to humanoid robots that can walk and dance, and now the race for the AI brain. Each step forward brings new opportunities, but also raises new questions.

The robot brain race is not just a story about technology. It's a story about patience, long-term investment, and vision. China spent years perfecting hardware, and now they are pouring billions of dollars into software. The US is leading in AI research but risk falling behind in commercialization. And in this race, the winner will be the one who best combines mind and body.

But one thing is certain: humanoid robots will no longer just be acrobatic performances on stage. They will become real working tools, and the race to build brains for them is happening right now. The question is no longer "can robots work?" but "who will teach robots how to work?"

Question for you: When humanoid robots can self-learn and adapt to new jobs after just a few seconds of observation, how do you think that will change your industry—and what will you prepare to become the person who controls the robot, rather than being replaced by it? Share your views below, because in this intellectual race, everyone's preparation is a piece of the puzzle to ensure we aren't left behind!