A major Chinese robotics company just secured $735 million by betting that the future of humanoid robots doesn't need two legs. AI² Robotics, valued at roughly $2.8 billion, is leading a quiet but significant rethinking of how humanoid robots should actually move and work. Instead of copying human bipedal locomotion, the company builds wheeled mobile manipulators with humanoid torsos and five-fingered hands, a design choice that's reshaping investor expectations and deployment timelines across the industry.

What Makes Wheeled Humanoids Different From Bipedal Robots?

The distinction between wheeled and bipedal humanoid designs reflects a fundamental trade-off in robotics engineering. Bipedal robots like Tesla's Optimus and Boston Dynamics' Atlas can theoretically navigate any environment humans can, including stairs and uneven terrain. But that versatility comes at a cost: mechanical complexity, durability challenges, and regulatory hurdles when deploying into public spaces. AI² Robotics' wheeled approach eliminates those friction points.

The company's AlphaBot features over 34 degrees of freedom and a custom waist-leg lifting mechanism that elevates the upper torso through multiple positions, allowing the robot to reach high shelves and work at different heights without needing to climb. While the robot cannot traverse stairs or rugged terrain, it is significantly cheaper to produce, mechanically more durable, and faces much lower regulatory barriers when deploying into logistics warehouses, manufacturing plants, biotech facilities, retail spaces, and public service environments.

How Does AI² Robotics Make Its Robots Actually Work?

Hardware alone doesn't make a functional robot. Every AlphaBot operates on Alpha Brain, AI² Robotics' proprietary vision-language-action (VLA) model, a type of artificial intelligence that combines visual perception, language understanding, and physical action planning into a single system. This foundation software handles real-time spatial reasoning, environmental understanding, and complex, multi-step task planning, allowing the robot to adapt to changing conditions on factory floors and warehouse environments.

The funding announcement reflects broader momentum in the humanoid robotics sector. AI² Robotics' $735 million raise comes alongside other recent funding events, including pending initial public offerings by robotics companies like Agility, funding rounds for Apptronik and Neura, and strategic acquisitions reshaping the competitive landscape.

Why Are Investors Backing This Strategy?

The diverse investor base in AI² Robotics' latest funding round reveals the strategic importance now placed on physical AI technology. The company secured commitments from government funds like the National Small and Medium Enterprises Development Fund, corporate players ranging from Sino Biopharmaceutical to the Moutai Group, and financial firms like CICC Capital and GSR Ventures. This mix of state-backed entities, industrial corporations, and financial institutions signals confidence that wheeled humanoid robots will solve real labor shortages in structured environments.

Meanwhile, on the other side of the world, Japan's construction industry is pursuing a different but complementary approach. Shimizu Corporation has begun demonstration testing of autonomous patrols using a bipedal humanoid robot at the Torch Tower construction site, Japan's tallest building currently under construction in Tokyo. The robot patrols the site at 1.0 meter per second, analyzing on-site video using multimodal large language models to streamline management tasks.

What Are the Real-World Applications Emerging Now?

The contrast between AI² Robotics' wheeled approach and Shimizu's bipedal experiments highlights how different industries are solving different problems. Shimizu's focus on humanoid robots stems from the unique characteristics of construction sites, where the environment changes daily as work progresses and manual labor still overwhelmingly dominates. The company determined that a humanoid robot has high potential to effectively substitute for human workers in such environments, particularly for tasks like site patrols and painting work.

Central to Shimizu's efforts is the construction of an "AI ecosystem" specialized for the construction industry. This refers to a mechanism that steadily cycles through on-site data collection and analysis, simulation, AI learning model construction, and implementation into robots. The modeling and archiving of skills that progress through this process are also expected to serve as a means of passing down advanced techniques Japan's construction industry has cultivated over many years to the next generation, amid the rapid aging of its skilled workforce.

How to Evaluate Humanoid Robots for Your Industry

• Environment Type: Assess whether your workplace has stairs, uneven terrain, and changing layouts daily. Bipedal robots excel in dynamic construction sites; wheeled robots dominate structured warehouses and manufacturing floors.
• Task Complexity: Determine whether you need general-purpose manipulation or specialized single-task performance. AI-equipped robots with vision-language-action models can handle multiple tasks through learning, while conventional robots require dedicated machines for each task.
• Deployment Timeline: Consider regulatory requirements and mechanical durability. Wheeled humanoids face lower regulatory hurdles and require less maintenance, enabling faster real-world deployment than bipedal systems.
• Workforce Integration: Evaluate whether your goal is replacing labor or preserving expertise. Shimizu's approach focuses on archiving skilled worker techniques through imitation learning, creating a bridge between human craftsmanship and automation.

For developers and researchers seeking accessible platforms, the landscape is also shifting. Unitree Robotics' G1, a compact bipedal humanoid priced from $13,500, represents a new category of affordable developer robots. With about 35 kilograms of weight, 23 to 43 joint motors depending on configuration, and approximately 2 hours of battery life, the G1 is positioned as a humanoid research and education platform rather than a finished industrial worker.

The G1's quick-release 9,000 milliamp-hour battery design allows developers to swap power packs during testing rather than waiting for long charging breaks, making it practical for iterative development. The robot can walk, move dynamically, carry small loads up to about 2 kilograms, and serve as a platform for artificial intelligence and robotics experiments. Its small size and transparent pricing make it more realistic for university labs and AI teams than many enterprise-only humanoids that require seven-figure pilots.

What's emerging across these three distinct approaches,AI² Robotics' wheeled manipulators, Shimizu's bipedal construction robots, and Unitree's compact developer platforms,is a maturation of the humanoid robotics market. Rather than a single "winning" design, the industry is discovering that different physical forms solve different problems. Wheeled robots excel at logistics and manufacturing; bipedal robots adapt to construction sites; compact platforms accelerate research and education. The $735 million funding round signals that investors now understand this diversity, and they're backing companies that solve specific problems rather than chasing the consumer robot fantasy that has dominated headlines for years.

" }