Embodied intelligence is igniting a transformative wave of integration across industries such as automotive, manufacturing, and the internet. From established giants with advanced supply chains to emerging startups in niche sectors, companies are increasingly “crossing boundaries” to invest in this field. This movement is driven by the pursuit of new growth avenues and strategic positioning in the race to merge artificial intelligence with the physical world. However, significant challenges remain, and the future landscape of platforms and enterprises in embodied intelligence is still unfolding. Industry insiders emphasize that this competition lacks a predefined blueprint; only those who continuously experiment and deeply cultivate practical scenarios can navigate the uncertainties and achieve tangible industrial breakthroughs.
At its core, embodied intelligence relies on a “body” to interact with the environment, generating intelligence and forming a closed loop of perception, decision-making, and action—much like how humans must enter the water to learn swimming, rather than practicing on land alone. Embodied intelligence represents a “system-level” intelligence encompassing hardware, software, and algorithms. In 2025, humanoid robots have emerged as the most prominent application form of embodied intelligence, already being deployed in various areas of smart manufacturing.
- Enterprises Crossing Boundaries for New Growth Curves
The automotive industry is regarded as one of the most critical and promising fields for embodied intelligence technology. At the recent International Motor Show Germany for Automotive and Smart Mobility, Chinese new energy vehicle manufacturer Xpeng showcased its IRON humanoid robot, which has entered factory training, attracting widespread attention. Xpeng representatives indicated that the domestic new energy vehicle manufacturing sector is gradually reaching a consensus: autonomous driving and robotics share极高的 technical commonality. The physical world models and foundational AI capabilities essential for autonomous driving can be repurposed for robotics applications.
In July of this year, the Hong Kong University of Science and Technology and BYD established a joint laboratory focused on embodied intelligence. BYD plans to invest tens of millions of Hong Kong dollars over the coming years to support the lab’s operations, concentrating on frontier research in robotics and smart manufacturing. As the “ultimate form” of embodied intelligence, the hardware iteration and cost control of humanoid robots have become focal points for the industry. Swedish IT giant Hexagon’s humanoid robot, AEON, is undergoing application testing in sectors like automotive and aerospace, primarily performing tasks such as operational work, inspection patrols, reality capture, and job assistance. Executives from Hexagon noted that humanoid robots, as key carriers integrating AI with the physical world, are becoming a core competitive domain in the new global industrial revolution.
Meanwhile, some manufacturing giants not yet involved in developing embodied intelligence technologies have formed extensive partnerships with robotics firms. In June, Bosch Group’s investment platform, Bosh Capital, and Beijing Galaxy General Robot Co., Ltd. jointly announced the establishment of a合资 company named “Boyin Hechuang.” The new entity will focus on core manufacturing scenarios like complex assembly and intelligent quality inspection, developing dexterous robots and advancing the large-scale implementation of embodied artificial intelligence in industrial settings.
Internet giants are also active participants in the embodied intelligence sector. At last month’s World Robot Conference, JD.com unveiled its “Intelligent Robot Industry Acceleration Plan,” committing to invest over ten billion resources in the smart robotics field. Additionally, JD.com announced the launch of its embodied intelligence brand, JoyInside, with dozens of leading brands in陪伴, education, and industrial sectors already integrated. Experts highlight that whether robotics companies, automakers, or internet conglomerates, all are vigorously entering the embodied intelligence arena, motivated by both short-term and long-term logic. In the short term, the market imagination for embodied intelligence far surpasses that of the automotive industry; merely announcing布局 can quickly boost capital market expectations and valuations. Long-term, the supply chains and technology stacks for new energy smart vehicles and embodied intelligence robots高度 overlap, from motors and sensors to computing platforms, allowing shared R&D outcomes and economies of scale. Thus, crossing boundaries is not only logical but also helps reduce costs and accelerate technological iteration. On a deeper level, the impetus stems from strategic positioning, as embodied intelligence is widely seen as a crucial direction for AI’s expansion into the physical world. Those who first capture this high ground may gain a competitive edge in future industrial rivalries.
Analyses from research institutions suggest that the primary objectives and mindsets of various players entering the embodied intelligence industry can be summarized into three points: first, seeking second growth curves and seizing future industrial entry points, as embodied intelligence is viewed as the next trillion-dollar disruptive intelligent terminal following personal computers, smartphones, and new energy vehicles; second, technology-driven strategic defense and ecosystem positioning, where in the AI era, rapid technological iterations pose颠覆 risks, making embodied intelligence布局 both a bet on the future and a defensive strategy; third, intrinsic needs for industrial upgrading and cost efficiency, especially for automotive and manufacturing firms, where introducing embodied intelligence, particularly humanoid robots, is key to advancing factory intelligence, flexible production, and addressing labor cost increases and demographic shifts.
- Challenges in Embodied Intelligence Development
As one of the most cutting-edge technologies, the practical implementation of embodied intelligence must overcome numerous obstacles. For instance, in agriculture, companies like XAG, a leading Chinese agricultural drone manufacturer founded in 2013, emphasize that embodied intelligence in farming faces significant barriers. One major threshold is the “digital agricultural infrastructure,” essentially high-definition farmland maps. When XAG first entered the agricultural sector, Chinese farmlands lacked navigation networks, and field boundaries were unclear. The company pioneered the introduction of北斗 IT navigation into Chinese farmlands, establishing over 2,000 self-operated base stations nationwide to cover most agricultural areas, enabling drones, unmanned vehicles, and future farm robots to connect to navigation systems upon activation. XAG has spent over a decade refining agricultural large models and accumulating farmland data to address real-world issues like fruit tree pest identification and drought monitoring through specialized models trained on field data collected during operations.
In the smart wearable industry, which has been evolving in tandem with embodied intelligence, products are协调融合 in both software and hardware. Companies like Liangliang Vision, a smart glasses firm, describe smart wearables as focusing on “human functional extension,” while embodied intelligence aims for “machine subject evolution.” Smart glasses enhance human perception through augmented vision and computing, whereas embodied intelligence requires full closed-loop perception, action, and learning. Currently, smart wearables have not deeply integrated with embodied intelligence but are poised to become key participants. Technically, smart glasses achieve basic perception and learning but rely on humans for action, lacking autonomous mobility. This limitation allows them to concentrate on human-computer interaction optimization, such as visual displays and voice interactions, rather than mechanical development. Future synergy could involve smart wearables acting as information hubs in mobile scenarios or data bridges in virtual-real integration, potentially combining with embodied intelligence in areas like industrial inspection and medical assistance.
Startups like Shenzhen-based Shengjing Technology, which specializes in spatial generation for embodied intelligence training, point out that core bottlenecks in embodied intelligence development include insufficient data quality and diversity. Traditional methods relying on physical scene capture or synthetic data are costly and lack authenticity. By generating highly realistic 3D virtual scenes, they provide standardized, reusable digital environments for robot training. For example, a virtually generated kitchen can simulate precise spatial layouts and material reflections under various lighting conditions, offering a realistic training ground for robots performing tasks like object retrieval, obstacle avoidance, and delivery. Industry practitioners note that embodied intelligence development is not reliant on a single “point technology” breakthrough but involves multi-level, multi-link integration of complex hardware and software systems. Achieving human-like intelligence in robots requires deep fusion of multi-modal perception and tight coordination between algorithms, sensors, motors, and computing platforms. Consequently, embodied intelligence cannot be accomplished by a single company; it resembles a full-chain industrial collaboration project. Mature large tech enterprises have greater opportunities, as automakers bring powertrain expertise and production scale, internet companies hold large models and vast data, and chip and sensor firms provide foundational support, creating complementary advantages.
Experts caution that the path to developing embodied intelligence will encounter more difficulties than previous technological cycles. Companies must progress through a “crossing the river by feeling the stones” phase, where frequent failures and continuous experimentation are essential for accumulating experience and vitality. Unlike past trends, there is no template from companies like Tesla in the U.S.; Chinese firms must lead the charge, exploring and paving their own way. This presents both pressure and a rare opportunity to form a distinct Chinese embodied intelligence system through rigorous refinement.
- Diverse Approaches to Embodied Intelligence Worldwide
Globally, companies are adopting varied strategies to跨界 into embodied intelligence. In the United States, tech giants are leveraging their unique strengths to build competitive advantages. Tesla, under Elon Musk, has shifted focus from electric vehicles to robotics and AI innovation. Analysts believe Tesla’s integration of hardware and AI technologies gives it potential advantages in solving supply chain and manufacturing challenges plaguing the robotics industry. Its associations with SpaceX and xAI could position it as a leader; recent videos show Tesla’s Optimus robot now equipped with xAI’s Grok AI assistant. Beyond hardware-focused firms, computing platform companies like Nvidia are advancing embodied intelligence competition. At this year’s GPU Technology Conference, Nvidia announced the launch of its latest Isaac GR00T N1 foundation model, an open-source, pre-trained, customizable model aimed at accelerating humanoid robot development and evolution. OpenAI has adopted a分散 investment strategy in humanoid robotics, collaborating with startup Figure AI while also supporting Norwegian robot startup 1X, with reports suggesting OpenAI may be developing its own humanoid robot hardware.
In South Korea,跨界 into embodied intelligence and robotics is concentrated among giants. Automotive leader Hyundai Motor Group acquired an 80% stake in U.S. robotics firm Boston Dynamics in 2021, planning to deploy the humanoid robot “Atlas” in Hyundai production plants by 2025. Earlier this year, electronics giant LG purchased an additional 30% stake in U.S. robot startup Bear Robotics, giving it a majority 51% ownership. LG aims to integrate Bear Robotics with its commercial robotics division to strengthen home and industrial robot businesses. South Korean media report that the government designated humanoid robots as a national cutting-edge strategic technology last year, and in April, the “K-Humanoid Robot Alliance” was formed with the goal of commercializing Korean-style humanoid robots by 2030. The alliance will collaborate with major universities to develop common AI models and core components equivalent to a robot’s “brain.”
In Europe, numerous companies are跨界 developing embodied intelligence, beyond traditional robotics firms like ABB and KUKA. For example, Germany’s BMW Group partnered with U.S. humanoid robot company Figure to pilot humanoid robots for assembling auto components in BMW’s U.S. factories. Additionally, U.K. online retailer Ocado acquired two U.S. robotics companies, Kindred Systems and Haddington Dynamics, with its technology chief stating that the move aligns with decade-long investments in autonomous robotics and sorting technology to expand business more efficiently and rapidly. European embodied intelligence development often focuses on specific domains: U.K.’s Neural Foundry creates autonomous robots for industrial environments; Switzerland’s Resmonics uses ambient AI and sensors for real-time respiratory risk monitoring; Norway’s Zygizo offers AI-based snow water equivalent detection tools for optimizing hydropower, flood prediction, and alpine infrastructure.
- Future Ecosystem: Super Giants and Application-Oriented Firms
Looking ahead, the commercial potential of embodied intelligence extends far beyond manufacturing, services, or exploration—it promises to reshape nearly every industry with disruptive force, potentially becoming the largest商业 opportunity in human history. From smart assistants in factories to home care, education, transportation, healthcare, and disaster rescue, embodied intelligence will gradually permeate daily life. The future societal norm will involve humans and robots coexisting, collaborating, and evolving together. Those who adapt first to this symbiotic格局 will grasp the initiative in global competition.
In terms of smart wearables and embodied intelligence synergy, potential integration scenarios include acting as “information transfer hubs” in mobile settings, where smart glasses relay real-time visual data for remote control of embodied robots, or establishing data bridges in virtual-real fusion via AR devices to align virtual training with real tasks. Smart wearables’ innate “human-machine symbiosis” can address pain points in交互实时性 and instruction precision, driving combination in scenarios like industrial inspection and medical assistance. Strategic choices by companies in this space reflect these trends.
Industry experts predict a future ecosystem of “few super giants + numerous application-oriented firms.” Giants would control upstream core elements like computing power, sensors, and actuators, while application companies leverage open underlying platforms to create diverse scenario values across sectors. For China, this represents a significant opportunity to promote global standards and transition from a manufacturing powerhouse to an innovation leader. At the recent World Robot Conference, dynamic demonstrations of domestic humanoid robots, such as Fourier Intelligence’s GR-1 and Ubtech’s WalkerS, drew attention. Experts generally believe that with accelerated localization of core components and economies of scale, hardware costs for humanoid robots could enter a critical decline within 3 to 5 years, clearing obstacles for early applications in industrial manufacturing and logistics.
Regarding industrial development patterns, some argue that the sector will exhibit “hardware convergence, software supremacy.” As Chinese supply chains advance and consumer electronics mature, hardware components like sensors and robotic arms in embodied intelligence, particularly humanoid robots, will become通用, with true competitive barriers emerging in software algorithms. Others envision a more complex “platform ecosystem”格局, featuring a handful of “platform-type giants” from current internet behemoths or new embodied intelligence-era leaders. Their core competitiveness would lie in developing embodied intelligence “operating systems + large model brains,” profiting through ecosystems that serve hardware makers and application developers. A specialized layer of “component and solution” suppliers would thrive around these giants, while整机 manufacturing would see “branded” and “differentiated” competition.
However, the current industry fervor also breeds “pseudo-demand traps,” such as overemphasizing humanoid forms while neglecting scenario complexity. Enterprises and investors must distinguish between “flashy tricks” and “scenario-specific skills.” Academics caution that despite the hype, practical limitations persist, with many embodied intelligence technologies facing “overhype” and widespread losses among invested firms. Recommendations include building “layered intelligence” systems tailored to different scenario needs, implementing human-machine collaboration, cross-border integration, and co-creation sharing as advocated for “AI+” initiatives.
In summary, embodied intelligence represents a pivotal frontier in the fusion of AI and the physical world, driving cross-industry innovation globally. While challenges in data, integration, and cost remain, the collaborative efforts of diverse enterprises—from automotive and internet giants to specialized startups—are shaping an ecosystem where embodied robots could redefine productivity and daily life. The journey demands continuous experimentation and scenario深耕, with those embracing this approach likely to lead the next industrial revolution.