At the "Intelligence Ignites a New Journey" Capital Markets Summit held in Shanghai from May 6th to 8th, 2026, Professor Wang Tianmiao, Honorary Director of the Beihang University Robotics Institute and founder of the Zhiyou Yari Science and Technology Innovation Platform, delivered a keynote speech. He shared his insights on the three major trends anticipated for the next phase of embodied artificial intelligence (Embodied AI) development over the coming 3-5 years, drawing from his experience in technology incubation, investment, and discussions with entrepreneurs, experts, and investors.
**Trend Forecast One: Innovation Ecosystem – National Strategy Takes Precedence, Focusing on Autonomy, Security, and Global Leadership** Embodied AI, as a new quality productive force, requires a clear understanding of new production relations. The prioritization of national strategic logic is now an inevitable trend, centered on national rejuvenation, propelling China's technology industry onto the global stage while ensuring autonomous and secure development. In this context, breakthroughs in hardcore technologies and "bottleneck" technologies, along with achieving domestic substitution, form the core thrust of China's development.
This trend is driven by two major challenges: severe geopolitical impacts on global supply chains, leading to their restructuring, and the "large but not strong" status of many of China's key industries. Consequently, semiconductors, AI, robotics, energy autonomy, and secure communications have become core strategic industries. Development must rely on a comprehensive intelligent, autonomous, and secure system, combined with civil-military integration. Encouraging major corporations and state-owned enterprises to increase R&D investment, deepening the integration of technological and industrial innovation, and combining organized innovation with free, curiosity-driven exploration are crucial. IPO resources should be allocated as a national strategic asset.
Observing current unicorn development, the embodied AI innovation ecosystem is coalescing around four key elements: 1) Balancing national strategy with market logic; 2) Defining China's global positioning in advantageous industries from a long-term hard-tech perspective, and identifying and nurturing leading chain-master and specialized "little giant" enterprises; 3) For sustainable entrepreneurial development, integrating technological leaders with business resource leaders to form "dual-core partnership teams"; 4) Considering the importance of future deployment scenarios, it is essential to partner with leading or key clients, secure important data resources, and strive for support from both state and social capital.
It is noteworthy that amidst the current high enthusiasm in the embodied AI sector, irrational valuation bubbles and sentiment-driven pricing are inevitable growing pains. These must be resolved through development itself—by achieving breakthroughs in foundational technologies, entering through edge applications, and adhering to business fundamentals. This will allow the industry to navigate the bubble cycle. As a disruptive new species combining hardware and software, embodied AI has the potential to propel China to a leading position in global intelligent manufacturing. Ultimately, leading enterprises will define the landscape, set standards, and rationalize supply chains, forging a uniquely competitive Chinese industrial ecosystem.
According to the Hurun Global Unicorn Index 2025, there are 1,523 unicorns globally, with 758 in the US. China, with 343, ranks second worldwide and has seen its number rise for three consecutive years. Approximately 50% of domestic unicorns have received early-stage investment from state-owned assets, and 30%-40% stem from organizational design, incubation, or spin-offs by large corporations or top-tier institutions. China's entrepreneurial ecosystem has evolved from grassroots startups to a stage of organized, multi-factor resource incubation and empowerment.
Recent capital deployment is also telling. In 2025, embodied AI-related financing involved 300-400 deals totaling 300-500 billion yuan. In just the first four months of 2026, there were 151 deals with a total value exceeding 500 billion yuan, matching the full-year 2025 figure. Top-tier VCs are making preemptive, full-industry-chain investments; internet corporate venture capital (CVC) is making strategic plays; national mega-funds are anchoring directions; and local state-owned capital is providing collaborative empowerment—all driving the high-end transformation of manufacturing.
Industrial clustering is becoming more pronounced. Survey data indicates 70% of unicorns originate from three major clusters: Beijing-Tianjin-Hebei (focused on original innovation), the Yangtze River Delta (focused on industrial traction and ecosystem incubation), and the Greater Bay Area (focused on global expansion and supply chain utilization). This pattern aligns with the policy direction emphasized during the Two Sessions: "comprehensively enhancing independent innovation capabilities, strengthening original innovation and breakthroughs in key core technologies, and building world-class sources of scientific and technological innovation."
**Trend Forecast Two: Convergence of Technology and Application – Embodied Large Models, Multimodal Fusion, and On-Device Chip Integration are Converging, Approaching the Eve of Large-Scale Application** First, it is essential to revisit the core concept of "Embodied AI." Distinct from traditional robotics, embodied AI centers on an intelligent agent that relies on a physical body to interact with the environment in real-time, forming an integrated system of perception, cognition, decision-making, and action. Through human-machine interactive cognition, perception-control loops, and iterative learning, it achieves generalization, transferability, and adaptation to a time-varying physical world, serving as a key technological bridge between the digital and physical realms. Embodied AI can be seen as the ultimate form of AI transitioning from "cloud algorithms" to "physical deployment." Its core technological elements—perceptual data, model algorithms, and computing chips embedded in carriers—enable the intelligent iteration of systems, equipment, devices, or terminals. Macroscopically, its strategic development direction can be understood as encompassing power, computing power, intelligence, data power, and productive application deployment.
Embodied AI can be broadly categorized into: Robots (AI + traditional robots, quadrupeds, humanoids, dexterous hands, bionic robots, emotional companion robots, special-purpose robots, etc.); Intelligent Carriers (unmanned vehicles, drones, unmanned surface vessels, eVTOLs, underwater robots, space robots, etc.); and Smart Terminals (wearables, smart home devices, information appliances, intelligent assistants, etc.).
The reason the embodied AI industry can evolve and expand is that all data is being activated, all cognitive decisions involve AI, all devices are becoming intelligent, and all software is being reconstructed.
The current embodied AI industry faces a "three highs, one difficulty" situation: high popularity, high investment, high trial-and-error, coupled with challenges in scenario generalization, non-transferable data operations, significant error tolerance risks, and high maintenance costs. Therefore, in 2026, industry focus has shifted from superficial metrics like manufacturing cost competition, large model parameter comparisons, and dataset sizes to the core foundational capabilities of embodied AI: embedding cloud and edge-side models into carriers, achieving multimodal fusion of vision, spatial intelligence, and tactile/force sensing, and gradually perfecting these through interactive iteration to overcome challenges in local scenario generalization, transferability, and system reliability.
The embodied large model is the core breakthrough point. Globally, a definitive, converged technological path and architecture have not yet emerged. However, over the next 3 to 5 years, both China and the United States are expected to produce "dark horses" in this area.
Furthermore, critical breakthroughs are needed in related ecosystem components like data, modules, chips, and tools. This includes, from a systems perspective, redefining systems, tasks, agents, and the collaborative allocation of roles and tasks between humans and digital workers using tools like OpenClaw. It also involves the quality mining, collection, and distillation of data value, as well as upgrades to core components like dexterous hands, edge-side perception chips, inference chips, and motion control chips. These are vital foundations for the scaled development of embodied AI.
Application scenario entry should follow the principle of "intelligent boundary definition," which involves delineating deployment scenario boundaries and, from a systems perspective, determining requirements for environment, objects, operational processes, real-time demands, error tolerance, safety, and cost to achieve corresponding intelligent functions. Disruptive innovation often breaks through from edge scenarios rather than directly entering mainstream battlegrounds. Three trends are currently evident: 1) Starting with commercial scenarios that are easy to deploy, cost-calculable, and have high error tolerance; 2) Prioritizing layout in edge industrial and commercial scenarios like factory assistance and in-store shopping guides; 3) Gradual penetration into public livelihood scenarios, from agricultural product processing to home services and elderly care. Market data further corroborates this trend.
At the industrial ecosystem level, embodied AI is far from a "zero-sum game for 3-5 companies." It is poised to create a market worth hundreds of billions to trillions over the next decade, primarily due to the formation of a "3+2" ecosystem: 3 Chain Leaders (embodied large models customized for different verticals; system integrators building universal platforms; AI RaaS providing operation, maintenance, service subscription, and leasing) + 2 Supporting Layers (50-100 specialized embodied AI application companies in细分fields; a full-chain supply chain), supporting the industry's breadth and depth.
**Trend Forecast Three: Serving the Intelligent Economy – Three Emerging Industries Converge to Drive Development, Building China's Advantageous Aggregated Intelligent Agent Industrial Cluster in Preparation for the Next Cycle of the Intelligent Economy** In the future, intelligent driving, embodied AI, and the low-altitude economy (eVTOL) will serve as emerging industrial drivers, integrating to form an aggregated intelligent agent industry that leverages China's supply chain advantages, preparing for the next cycle of the intelligent economy. This forecast is based on the high degree of correlation among these three industries: 1) **~80% Technological Commonality**: Data, models, and chips are the three core driving elements; the disruption, rise, and development of these three industries are largely based on the iterative advancement of these three technological engines. 2) **~60% Industrial Base Overlap**: This includes foundational manufacturing equipment, testing equipment, and core component supply chains. 3) **Synergistic Development**: As a hardware-software integrated实体industry, embodied AI does not develop in isolation but advances in synergy with related industries of the past, present, and future.
The 2026 International Auto Show illustrates this. China's new energy vehicles already hold 70% of the global market. Level 3 autonomous driving has achieved mass production scale, while RoboTaxi and RoboVan are entering trial operation phases. Automakers and commercial insurers are negotiating risk coverage for accidents. Automotive hardware is evolving from distributed 3-electric systems to centralized architectures. Chassis, autonomous driving, and cockpit systems are becoming more modular, intelligent, and networked. Supporting key technologies like cost-effective LiDAR sensors, safe solid-state batteries, and 3-5 minute flash-charging stations are also iterating simultaneously. Most automakers are preparing to enter embodied AI R&D or supply chain system standards, creating synergistic effects.
Sino-US industrial layouts differ markedly: China places greater emphasis on combining foundational infrastructure (data, power, computing power, supply chain) with vertical applications of large models, while the US heavily invests in foundational large models and underlying hardware/software tools. This differentiated competition and integrated development will significantly propel the industry's evolution from the "Stone, Iron, Electrical" ages toward the "Intelligent" age, forming a "1+3" aggregated intelligent industry chain—"1" being the foundational layer of models, data, chips, energy, and intelligent manufacturing; "3" being整机manufacturing, core components, and operational services.
In the long term, over the next 10-20 years, China will, through investment in disruptive technologies, reconstruct its infrastructure system. This will encompass areas like AI large models, computing power centers, AI for Science, fusion nuclear energy, space-based computing, life sciences/genetics, and foundational material purification. The goal is to establish general and specialized intelligent agent manufacturing platforms and corresponding supply chains for market applications, ultimately enhancing China's voice in shaping international order, economic restructuring, asset definition, AI ethics, and civilizational discourse.
**Summary: Understanding the development trends of the next phase of embodied AI and boldly embracing the new wave of the future intelligent economy.** We must recognize current opportunities: 1) The "Dual Super" engine driving change (supercomputing centers,超级large models & Agents); 2) The rise of three new industries (intelligent driving, embodied AI, low-altitude economy); 3) The redefinition of strategic resources (rare metals, computing power chip tokens, new energy support). Future emerging technologies like quantum computing, controlled fusion, space-based computing and mining, life sciences/genetic coding, and AI for Science also warrant attention.
The core trend forecasts for the next 3-5 years in embodied AI can be summarized in three points: First, national strategic logic takes precedence, with autonomous and secure development as the core guiding principle. Second, technology and application are converging; embodied large models, multimodal data fusion, and on-device chip integration are gradually converging, beginning the march towards large-scale deployment. Third, intelligent driving, embodied AI, and the low-altitude economy (eVTOL) will serve as three emerging industrial drivers, developing synergistically to build China's most advantageous aggregated intelligent agent industrial cluster, supporting the next new cycle of the intelligent economy.
It is believed that AI will continue to catalyze the development of embodied AI and aggregated intelligent agents, unleashing a new wave of intelligence across all industries, propelling China from its advantages in industry and supply chains towards the high end of global technology.
Comments