China is betting that putting artificial intelligence into robot bodies will give the nation a decisive edge in the race toward artificial general intelligence (AGI), a theoretical AI system with human-level reasoning abilities. The country's strategy centers on what experts call "embodied AI," where robots and other hardware interact with the physical environment to learn and improve, rather than relying solely on text-based language models like the West does.

What Is Embodied AI and Why Does China Think It Matters?

Embodied AI represents a fundamentally different philosophy about how machines learn. Instead of training artificial intelligence exclusively on vast amounts of text data, as large language models (LLMs) do, China's approach integrates robots, drones, and vehicles with AI systems. The theory is that by physically interacting with the world, these machines can develop a more complete understanding of how things work.

"China believes you can only go so far with artificial intelligence if you are dealing only with text and other abstractions of language," said William Hannas, lead analyst at the Center for Security and Emerging Technology at Georgetown University.

William Hannas, Lead Analyst at the Center for Security and Emerging Technology at Georgetown University

This contrasts sharply with the United States and Europe, which are primarily pursuing AGI through increasingly powerful language models trained on text. China's policymakers see embodied AI as a more direct path to machines that can reason and act like humans across any task.

How Big Is China's Humanoid Robot Industry Right Now?

The scale of China's humanoid robot push is staggering. The country is home to an estimated 140 manufacturers that control more than 90 percent of global humanoid robot sales. Shipments are expected to more than triple this year to 50,000 units, compared with roughly 13,000 units last year, according to financial firm Morgan Stanley. To put this in perspective, Chinese buyers are snapping up the vast majority of those robots domestically, making humanoid robotics a massive industry within China itself.

AgiBot, one of China's top robotics firms, shipped more humanoid robots than any other manufacturer in the first half of this year, with 9,700 units according to research firm Counterpoint. The company's X2 model, a 4-foot-tall humanoid robot, recently helped AgiBot win more than a dozen gold medals at the World Humanoid Robot Games in Beijing.

How Are Chinese Companies Using Robots to Build Better AI?

The real innovation lies not just in building robots, but in using them to gather data. Chinese manufacturers are deploying humanoid robots in factories and other settings to collect real-world information that can train better AI systems. UBTech Robotics, one of China's largest humanoid robot firms, deployed more than 1,000 industrial robots in factories last year and aims to deploy 10,000 this year. Each robot in the field generates valuable data about how to move, manipulate objects, and navigate real environments.

"More human robots used in different industries means more data coming in," said Michael Tam, chief brand officer at UBTech Robotics, adding that physical data gathered by robots is "crucial" to achieve AGI.

Michael Tam, Chief Brand Officer at UBTech Robotics

Currently, UBTech's robots operate at about 30 percent of human efficiency when performing tasks like picking up boxes and placing them on shelves, with a goal of reaching 50 percent efficiency this year. While this may sound modest, every task completed generates data that helps train the next generation of robots to be faster and more capable.

What Role Is the Chinese Government Playing?

Beijing's policymakers are actively supporting this embodied AI strategy. In June, the government issued a directive requiring local authorities and state-owned firms to test and incorporate embodied AI in manufacturing, logistics, retail, and other economic sectors. The government is also promoting large-scale "data collection factories" to accelerate AI model training.

These facilities are sprouting up across Chinese cities and provinces, including Beijing, Shanghai, Wuhan, Tianjin, Anhui, and Fujian. At these centers, local governments and manufacturing firms purchase humanoid robots and employ human workers to use controllers, cameras, and other tools to direct the robots in specific actions. The data collected is then sold back to manufacturers, creating a feedback loop that continuously improves robot capabilities.

Steps to Understanding China's Embodied AI Advantage

• Data Collection Strategy: Chinese firms are deploying thousands of robots in real-world settings to gather physical interaction data, which they use to train AI systems that understand the actual world rather than just text.
• Government Support: Beijing's policy directives mandate that state-owned firms and local authorities test embodied AI across multiple economic sectors, creating a coordinated national push rather than scattered private efforts.
• Manufacturing Scale: With 140 domestic manufacturers controlling over 90 percent of global humanoid robot sales, China has built an ecosystem that allows rapid iteration and cost reduction through mass production.
• Feedback Loop: Each robot deployed in factories or data collection centers generates information that improves the next generation of robots, creating a self-reinforcing cycle of improvement.

What Challenges Does China's Strategy Still Face?

Despite the aggressive push, significant obstacles remain. The United States imposed a ban in July on imports of new humanoid and quadruped robots made in China and other foreign countries, citing national security concerns. This regulatory barrier could slow China's ability to export robots and gather international data.

The commercial viability of humanoid robots remains unproven. Alan Fern, director of the Dynamic Robotics and AI Laboratory at Oregon State University, noted that "there are tons of companies trying to stay alive" in the humanoid robotics space, and many will likely fail given the risk of overproduction and price wars. He also observed that the current business model for humanoid robots is "based on hope" rather than proven profitability.

Within China itself, industry leaders acknowledge that progress is uneven. Zhang Jianxin, a technical director at AgiBot, pointed out that while "the hardware is good, the software isn't there yet," highlighting that AI technology improvements remain the critical bottleneck. A data shortage coupled with high training costs represents the biggest obstacle in the humanoid robot industry, according to China's state-run Xinhua News Agency.

Still, if China's humanoid robots continue to advance and the data collection strategy pays off, the nation's embodied AI approach could fundamentally reshape the global race toward artificial general intelligence. The next few years will reveal whether physical interaction with the world truly offers a faster path to AGI than text-based language models alone.