Are the three giants of Silicon Valley igniting a mass production revolution, and will China's embodied intelligence claim the C position on the global stage?

Introduction: Predefined actions are today’s ticket to entry, while generalization capability is tomorrow’s ticket to the finals.

Editor: Jingcheng

Author: Jiang Jing

As the first quarter of 2026 comes to a close, a globally synchronized action in the tech industry officially declares a historic turning point for the humanoid robot industry.

The three major Silicon Valley giants, Google, Amazon, and Tesla, are making simultaneous efforts, sprinting from technological empowerment and scenario layout to mass production realization, pushing humanoid robots from tech showcases to industrial arenas.

Meanwhile, there have been more actions from China. On March 26, the China Academy of Information and Communications Technology, in collaboration with over 40 organizations, released the first industry standard in the field of embodied intelligence. Coupled with continuous policy support, accelerated enterprise implementation, and heightened capital enthusiasm, China is making a leap from following to running alongside, even beginning to challenge for the lead in several fields.

Will China occupy center stage in this revolution that disrupts future business rules and industrial ecology?

Global Surge: Silicon Valley Giants Move to Mass Production, Restructuring Future Productivity

No one considers humanoid robots merely a science fiction concept anymore.

Recently, the synchronized actions of the three Silicon Valley giants have made the footsteps of the mass production era clearly audible. Every step they take targets the reconstruction of future productivity, and the follow-up by global capital and local enterprises further escalates the heat of this track.

Google took the lead by creating a “smart brain” for robots, launching two new AI models, Gemini Robotics and Gemini Robotics-ER. The former allows robots to understand new contexts without specific training, while the latter can “comprehend complex and dynamic worlds,” empowering robots to land in real-world scenarios from a technological perspective.

Amazon focused on scenario implementation, acquiring humanoid robot startup Fauna Robotics and logistics robot company Rivr in quick succession within a week. Their strategy isn’t just about optimizing delivery; it’s about building a “capillary network of robotic services” from factory assembly lines to living rooms, creating the next-generation labor system.

Tesla’s mass production efforts have drawn the most attention. On March 25, the Optimus robot issued a talent recruitment announcement, clearly stating it will change the landscape of labor and manufacturing economies, aiming to achieve large-scale production as soon as possible. This summer, it will launch the world’s first humanoid robot production line with an annual capacity of one million units, pushing mass production into a substantive phase.

The layout in Silicon Valley goes far beyond this; American local enterprises are also accelerating implementation. On the same day, the Figure03 humanoid robot developed by Figure AI entered the White House, becoming the first American-made humanoid robot in the White House, capable of multilingual communication and autonomously completing household chores. The company secured over $1 billion in funding six months ago, with giants like NVIDIA and LG endorsing it, showcasing global capital’s enthusiasm for the humanoid robot track.

Yuan Shuai, deputy director of the investment department at the China Urban Development Research Institute, stated that the mass production actions of Silicon Valley giants and the release of China’s embodied intelligence industry standards jointly signify that the humanoid robot industry is moving from the deep waters of technological research and development to a golden period of commercialization. Breakthroughs in core technologies support large-scale manufacturing, while industry standards delineate technical specifications and reduce disorderly competition.

However, Gao Heng, an expert from the China Science and Technology Journalism Society, offers a cautious judgment, suggesting that the industry is currently entering the eve of commercialization and a partial realization period, rather than a golden period of full commercialization. The core change in the current industry is that all parties are beginning to jointly verify whether robots “can continuously work in real scenarios and whether costs are controllable,” rather than merely achieving breakthroughs in technological research and development.

China Breakthrough: Multiple Advantages Establish a Firm Footing, Core Shortcomings Urgently Need to Be Addressed

As Silicon Valley giants ride the wave of mass production, China is not passively following but has already laid the groundwork. Leveraging multiple advantages in standards, scenarios, markets, and capital, it has established a firm position in the global embodied intelligence track. However, compared to Silicon Valley giants, there remains a gap in core technologies and capabilities, which constrains further development of the industry.

In terms of advantages, China’s layout demonstrates distinct local characteristics and first-mover effects. Firstly, it has seized the power of standard-setting. On March 26, the China Academy of Information and Communications Technology, together with over 40 organizations, released the first industry standard in the field of embodied intelligence, establishing a unified benchmark testing framework and taking the initiative in standard-setting during the early stage of industrial development.

Secondly, it leads in scenario implementation. The development of embodied intelligence in China has never been confined to demonstration stages but has genuinely achieved practical applications. For instance, the Yushu quadruped robot has been deployed in various industrial inspection projects, including substations in Zhejiang, underground utility tunnels in Hangzhou, and the Guangdong Petrochemical base.

At the same time, China possesses a vast market scale and an active capital environment. By 2025, there will be over 140 domestic embodied intelligence companies, releasing more than 330 humanoid robot products, with shipments around 17,000 units. The market sizes for embodied intelligence and humanoid robots are projected to reach 5.295 billion yuan and 8.239 billion yuan, respectively.

On the capital side, Yushu Technology’s IPO has been accepted, making it the first humanoid robot stock in A-shares. Since the beginning of the year, there has been a surge in significant financing for the embodied intelligence industry, accelerating the capitalization process. In the first nine months of 2025, Yushu Technology’s sales revenue for quadruped and humanoid robots grew by 182.22% and 6.42 times year-on-year, respectively, providing direct evidence of market potential.

Despite the rapid development momentum, China’s shortcomings in the global competition are equally apparent.

Multiple experts point out that the core difference between Chinese and foreign humanoid robots lies not in hardware manufacturing but in data accumulation, model generalization capabilities, and foundational technological depth, which is manifested in the robots’ lack of flexibility in movements and generalization ability.

Yuan Shuai believes that the gap between Chinese and foreign humanoid robots appears to be differences in movement flexibility and generalization ability, with roots in foundational technologies, data accumulation, and research and development philosophies. For example, Google’s RoboCat can achieve flexible generalized movement due to long-term technological accumulation, particularly through sustained investment in areas such as large model algorithms, sensor fusion, and robotic dynamics control, relying on vast amounts of multi-scenario training data that enable robots to possess autonomous learning and environmental adaptation capabilities.

He points out that domestic products currently remain at the stage of predefined actions and fixed scene reproduction. The core shortcomings are a lack of high-quality, large-scale real-scene training data and insufficient algorithm generalization capabilities, as well as reliance on imported core components like high-precision servo motors and force sensors, which restrict motion precision and perception levels.

Gao Heng adds that the true gap lies in whether data, models, system engineering, and scenario closed-loop capabilities can form a synergy. The goals of leading foreign companies are to create intelligent robots that can understand their environments and autonomously complete tasks, fundamentally treating robots as sustainable iterative data products in research and development. Generalization capability is inherently a compound ability; domestically, the issue is not just single-point technological lag but rather a failure to form an iterative flywheel between data and scenarios, making it difficult for robots to grow increasingly intelligent beyond single tasks.

Renowned financial writer and head of the Tiaoyuan Influence Research Institute, Gao Chengyuan, states that the core gap is concentrated in data accumulation and model generalization capabilities. Foreign entities have distinct advantages in transfer learning from simulation to reality and multi-task general strategies, establishing cross-scenario data closed loops and foundational model R&D capabilities through long-term investments. Domestically, there remains a focus on predefined actions, fundamentally due to a lack of high-quality embodied data, along with a generational gap in the computational power and algorithm engineering capabilities needed for end-to-end large models.

Yushu Technology also acknowledges that the key technologies awaiting breakthroughs for large-scale commercial applications in industrial and household scenarios primarily include the embodied large model capabilities at the “brain” level and the finesse and durability of the “dexterous hand.” The most significant technical challenge is that globally, embodied large models are still in the early stages of development, with insufficient generalization capabilities.

Path to Breakthrough: Multi-Dimensional Paths to Enhance Capabilities, Balancing Current and Long-Term Development

Against the backdrop of insufficient data and scenario accumulation, how to enhance the flexibility of robot movements and generalization capabilities has become a core issue for domestic enterprises in catching up.

Multiple experts, considering the current industry situation, have proposed development paths that are both practical and forward-looking, while emphasizing that enterprises must balance short-term implementation with long-term R&D, using predefined actions as a ticket to entry and generalization capabilities as a core barrier.

Wang Peng, a researcher at the Beijing Academy of Social Sciences, suggested that domestic enterprises could catch up through two paths: “scenario anchoring + technology reuse.” On one hand, they should focus on vertical scenario data closed loops, first locking in standardized scenarios like industrial welding and material handling, obtaining exclusive data sets through small-scale implementations, and then training vertical field-specific embodied models. On the other hand, they should leverage open-source ecosystems for collaboration, using industry standards released by the China Academy of Information and Communications Technology to promote cross-enterprise data sharing and engage in joint training of general models based on a unified format of operational data.

Yuan Shuai recommends a multi-path approach, advocating for collaboration with universities and research institutions to use simulation and digital twins to generate virtual data for training and transfer to real scenes. They should also open interfaces to engage scenario stakeholders in pilot projects to collect real data for algorithm iteration, while promoting anonymous training data sharing among enterprises to break down data silos and increase self-research investments in core components to support flexible robot movements.

Gao Heng provided four practical paths: Firstly, obtain data from real scenarios, deeply integrating with factories, warehouses, and other environments to embed robots in real workflows to accumulate data. Secondly, prioritize simulations followed by real machines, training strategies in simulated environments before fine-tuning in real scenes to reduce training costs. Thirdly, focus on task generalization, concentrating on single-type tasks like picking and moving to achieve generalization and first realize commercial value. Fourthly, establish an industry-sharing data and standards system to address the lack of uniformity in interfaces and evaluation systems, forming industry-level iterations.

Experts unanimously agree that predefined actions and generalization capability are equally important for enterprise development.

Wang Peng believes that in the short term, robots with predefined actions can meet most industrial scene demands, and their costs are lower than those of robots with generalization capabilities. However, in the long term, generalization ability will be the core barrier that determines whether enterprises can transcend industrial cycles. As non-standard scenes like home services and emergency rescue expand, robots that can autonomously adapt to environments will gradually become mainstream.

Gao Heng also agrees that predefined actions are today’s ticket to entry, while generalization capability is tomorrow’s ticket to the finals. For enterprises, they should not abandon long-term investments in generalization capability just because they can earn money now through predefined actions; conversely, they should not neglect currently implementable scenarios in pursuit of generalization. Securing orders first and then developing intelligence is a more realistic approach.

Currently, China’s embodied intelligence market has already captured half of the global market, and practical applications have been realized in scenarios such as industry and emergency response. In the future, which types of scenarios will become the breakthrough points for the large-scale commercial use of Chinese embodied intelligence robots?

Gao Chengyuan believes that industrial manufacturing will be the breakthrough point for large-scale commercial use in China, especially in scenes like automotive manufacturing, 3C electronic assembly, and warehousing logistics. To uncover the demand for scenarios, it is essential to delve into the front lines of the industry and collaborate with leading manufacturing enterprises to co-establish joint laboratories, starting with the replacement of individual processes and gradually expanding to whole-line automation. The key to promoting the integration of technology and scenarios lies in establishing a reverse-driven mechanism of “scene-defined technology,” allowing real production line demands to drive hardware iteration and algorithm optimization, rather than the technology leading the way in search of scenes.

From “running alongside” to “global leadership,” China still needs to break through core bottlenecks in policy, technology, and industry ecology.

Yuan Shuai suggests that at the policy level, support and funding should be strengthened, and intellectual property protection should be improved; in terms of technology, the focus should be on tackling large model algorithms and core components to enhance the autonomous learning and generalization capabilities of robots; within the industrial ecology, upstream and downstream collaboration should be strengthened, accelerating the localization of components, deepening the integration of production, education, and research, and promoting the transformation of achievements. At the same time, actively engage in international cooperation and participate in global standard-setting to enhance industry discourse power, ultimately building a comprehensive embodied intelligence industrial ecology to achieve leadership goals.