The evolution from robots twirling handkerchiefs at last year's Spring Festival Gala to various robots performing together on stage this year highlights how major robotics manufacturers are gaining mainstream attention. Over the past year, robots have walked catwalks, sparred with founders, and even performed energetic dances at marathon concerts, significantly raising their public profile. The only disappointment was the stock market closure during the holiday period.
In contrast to the increasingly visible and tangible Chinese robotics sector, the American robotics industry appears comparatively quiet. While ambitious players like Tesla have even paused Model S/X production to focus on mass-producing humanoid robots, the current timeline remains "starting deliveries in the second half of 2026." The long-dormant Boston Dynamics, another notable name, recently unveiled a robot with a large USB port on its face at CES.
Although prominent American robots are few, the number of robotics companies is substantial. AI pioneer Fei-Fei Li chose robotics for her first venture, with her company World Labs focusing on data synthesis and robotic models. Physical Intelligence, backed by OpenAI, also specializes in models, with its founding team hailing from OpenAI and Google's DeepMind. NVIDIA is actively involved too, not only producing chips but also developing the Isaac platform specifically for humanoid robots, which has been updated three times this year, more frequently than its GPU releases.
This indicates that while Chinese and American industries share a consensus on the potential of robotics, their technological development paths diverge significantly. China emphasizes hardware, with robots capable of performing intricate movements and even working in factories. The U.S. focuses on software, producing numerous algorithms, research papers, and patents, often making product launches resemble academic seminars. For instance, Figure AI's release of a VLA model boosted its valuation, and Tesla's third-generation robot demonstrated "end-to-end" video learning, shifting Silicon Valley's focus to formulas and code.
China's robust manufacturing capabilities complement America's strong foundation in computer science, setting the stage for another competitive race in robotics.
This divergence represents the two pillars of humanoid robot development: hardware and software. Unlike traditional industrial robots, humanoid robots aim for "general-purpose" functionality, enabling them to perform any task a human can. Since human tools are designed around human proportions, robots must adapt to these specifications, requiring human-like forms, flexible limbs, and dexterous hands. Simultaneously, software algorithms are essential for robots to understand real-world physics, such as applying the right force to hold an egg without breaking it. Thus, hardware serves as the body, and software as the brain, both indispensable and potentially mutually constraining.
For Chinese manufacturers, robot dancing is not just for show but a demonstration of hardware prowess. Robotics is a nascent industry with limited component commonality. For example, "electronic skin" requires sensors with sensitivity and precision beyond current standards, while joints—critical for complex tasks—determine capabilities like carrying weight or threading needles. Even natural walking and dancing depend on advanced joint solutions.
In contrast, American firms concentrate on software, aiming to bridge the gap between large language models and spatial intelligence—understanding physical rules. While language models can describe phenomena (e.g., an apple falling), spatial intelligence simulates real-world physics, predicting environmental changes. Tasks like opening jars or cans are often more challenging than dynamic movements. Companies like Google and NVIDIA, along with startups like World Labs and Physical Intelligence, are developing world models to teach AI physics.
China is building the body, America the brain—a分工 reminiscent of the new energy vehicle industry. Autonomous driving can be seen as a precursor to humanoid robots; if electric vehicles are wheeled robots, humanoid robots are legged EVs. Both rely on cameras and sensors for data, compute chips and models for decision-making, and motors for execution. Humanoid robots require more precise data, complex decisions, and diverse tasks, but share core technologies like micro motors, battery systems, and control algorithms. Tesla's Optimus, for instance, reuses FSD algorithms from its cars.
This explains why companies successful in EVs are venturing into robotics, such as Tesla, Xiaopeng, and Li Auto. The supply chain reflects this分工: Chinese firms, rooted in automotive manufacturing, supply components like joints and batteries, while American companies provide software and chips. NVIDIA's Jetson Thor, derived from its Drive Thor autonomous driving chip, and Waymo's sensor technologies adapted for robotics, exemplify this synergy.
Tesla's Shanghai Gigafactory, built rapidly with support from China's automotive supply chain, exemplifies this collaboration, handling over half of Tesla's deliveries while software development remains in the U.S. Similarly, Google shifted from hardware to algorithm-focused robotics research.
Top talent flows to industrial hubs: the U.S. leads in computer science, attracting global experts, while China's manufacturing strength fuels innovation. As noted in a report, China's cost-effective components are key to Tesla's Optimus targeting a $20,000 price point. American firms leverage software for high value, outsourcing production to China.
However, China is narrowing the gap in software. While trailing in smartphones, advancements in autonomous driving show reduced disparity, thanks to talent from China's internet sector. In robotics "brains," China is catching up quickly; for example, VLA technology, pioneered by Google and OpenAI, saw Chinese versions from Zhiyuan and Xiaopeng with unique improvements shortly after.
Elon Musk, acknowledging this trend, ranked Tesla's Optimus first but expressed concern that spots two through ten would likely be Chinese companies. The competitive dynamics seen in EVs may soon replay in humanoid robotics.