ITjuzi: Chinese embodied AI companies raised ~$13B across 322 deals in H1 2026, up 5x YoY, as robot makers shift focus from hardware to AI (Reuters)

Embodied AI is artificial intelligence designed to operate through a physical body, usually a robot. It must sense the surrounding world, understand what is happening, choose an action, and control movement. This matters because physical environments are unpredictable. A robot must cope with changing objects, people, surfaces, and distances. Software-only AI handles digital information, such as text, images, or databases, without directly needing to move through the real world.

A humanoid robot running, dancing, or performing a backflip shows the physical side of the challenge. Its AI must combine sensing, balance, timing, and movement. The hardware supplies motors and sensors, while the AI turns information into actions. Better software can therefore change what the same machine can do.

Chinese companies are now directing resources toward this AI challenge after years of building capable humanoid hardware. ITjuzi recorded about $13 billion raised across 322 Chinese embodied-AI deals in the first half of 2026. The shift suggests intelligence is becoming as important as the robot body itself.

Chinese embodied-AI companies raised approximately $13 billion during the first half of 2026. The money came through 322 deals, according to ITjuzi. That is a striking scale for a sector still developing its commercial and technical foundations. The total shows that investors are treating embodied AI as more than a robotics hardware niche.

The comparison is equally important. Funding was up fivefold year over year. In other words, companies raised about five times as much as they did in the comparable period of 2025. The available article excerpt does not provide the exact 2025 dollar total, so the fivefold comparison is the clearest reported measure.

The surge also reflects a change in emphasis among Chinese robot makers. After years spent building humanoids that could run, dance, and perform backflips, companies are pouring resources into AI. The funding figures suggest that teaching robots to understand and act may now be viewed as the sector’s central challenge.

Embodied AI covers systems in which artificial intelligence is linked to a physical machine. The robot may sense its surroundings, interpret what it detects, make decisions, and move through the world. The defining feature is the connection between digital reasoning and physical action. This makes the sector broader than software models that only generate or analyze digital information.

The source gives humanoid robots as its clearest example. Chinese companies have built machines capable of running, dancing, and performing backflips. Those abilities require AI to coordinate sensing, balance, timing, and movement through mechanical hardware. The sector therefore includes both the robot platforms and the AI systems that help them perform tasks. The excerpt does not name specific commercial, industrial, or household categories.

The article frames the sector around a transition from impressive physical demonstrations toward more capable AI. Its funding figures cover Chinese embodied-AI companies generally, while the background focuses on humanoid robot makers. That combination indicates a field where hardware remains essential, but intelligence increasingly determines usefulness.

Hardware gives a robot a body, but AI determines how that body responds to the world. A machine may have strong motors and advanced joints yet remain limited if it cannot recognize objects, understand instructions, or choose safe actions. This is why the shift matters: better intelligence can unlock more uses without requiring an entirely new robot design.

The article describes years of work on humanoids that could run, dance, and perform backflips. Those feats demonstrate physical capability. The newer challenge is different. AI must help a robot interpret its surroundings and decide what to do. In practical terms, software connects sensors, decisions, and movements. It can make existing hardware more adaptable rather than merely more athletic.

The excerpt does not give a specific statement from a company explaining its investment decision. It does show a broad industry trend: Chinese robot makers are pouring resources into AI. The roughly $13 billion raised across 322 deals in H1 2026 suggests that investors also see this intelligence layer as strategically important.

A robot’s usefulness depends on more than its motors, joints, and appearance. AI can help the same hardware recognize objects, interpret instructions, predict movement, and select suitable actions. It can also coordinate several joints more smoothly. As a result, software improvements may expand a robot’s capabilities without replacing its physical components. This is one reason the move toward AI matters.

Consider a humanoid robot already able to walk or perform a backflip. With stronger AI, it might better understand where it is, identify a target, plan a route, and adjust its movements when conditions change. The mechanism is a loop: sensors provide information, the AI processes it, and controllers send commands to the motors. Feedback then helps the robot correct its behavior.

The article does not provide a detailed task demonstration beyond running, dancing, and backflips. It does report that Chinese robot makers are redirecting resources from hardware toward AI. The broader implication is clear: future progress may come less from adding physical features and more from making existing machines flexible, aware, and useful.

Funding for physical robot manufacturing pays for tangible production. Companies need components, motors, sensors, factories, tooling, testing, inventory, and workers. These costs rise when firms build and ship large numbers of machines. AI software funding is usually directed toward research, engineers, data, model training, computing infrastructure, testing, and deployment. It can be reused across many machines once developed.

The article reports one combined figure: Chinese embodied-AI companies raised about $13 billion across 322 deals in the first half of 2026. It does not separate that total into hardware and software funding. The source does, however, describe robot makers shifting their focus from building hardware toward AI. That signals a change in investment priorities, not a precise split in spending.

Software can improve a robot’s behavior without requiring a new body, while manufacturing creates the physical platform itself. In practice, the two remain connected. AI needs robots for testing and deployment, and robots need AI to become more capable. The reported fivefold year-over-year funding increase shows strong momentum, but not how much went to either category.

An embodied AI system must learn a continuous physical-world loop. First, it needs perception: identifying objects, people, locations, motion, and changes around it. Next, it needs understanding and decision-making. It must connect what it senses with a goal, predict possible results, and select an action. Finally, it needs control: turning that decision into coordinated movement through the robot’s motors and joints.

For a humanoid robot running, dancing, or performing a backflip, timing and balance are essential. The system must combine sensor information with a movement plan, send commands to the body, and use feedback to correct errors. Safety adds another requirement. The robot must recognize hazards, limit force and speed when appropriate, and stop or change course when conditions become unsafe.

The source does not list a formal training curriculum or specific algorithms. It does show why these capabilities matter. Chinese companies are moving resources toward AI after years of developing physically capable humanoids. The reported investment surge reflects the challenge of making robots understand and act, not merely move impressively.