When a technology moves from concept validation to scaled deployment, the questions that matter most change completely. For embodied intelligence, the shift is already underway. On August 20, during the 2026 World Robot Conference, the Embodied Intelligence Industry-Finance Practice and “Venture Star” Roadshow event brought together some of the country’s leading robotics scientists, senior executives from core industry chain enterprises, investors from frontline hard-technology investment institutions, and a group of highly promising innovation and entrepreneurship teams. Together they examined the critical moment in which embodied intelligence is stepping out of the laboratory and into the real economy.
The gathering was organized around two thematic threads that linked two separate conversations on the industrial and financial development of embodied intelligence. The first thread, “Piercing the Fog of Technology, Seeing the Truth of Industry,” tackled the technical fundamentals and the realistic boundaries of embodied intelligence. The second thread, “The ‘Super Scenario’ of Embodied Intelligence: A Manufacturing Revolution,” moved the discussion onto the factory floor and examined what it actually takes to move embodied intelligence from a demonstration to a daily production asset. Across both conversations, participants attempted to clarify the direction of the industry amid competing viewpoints and to identify a pragmatic landing path for the second half of the embodied intelligence era.
1. Piercing the Fog of Technology to See the Truth of Industry
The first conversation was conducted between Xi Ning, Chair of the Robotics Society of the Chinese Institute of Electronics, Chair Professor of Robotics and Automation and Head of the Department of Data and Systems Engineering at the University of Hong Kong, and Director of the Institute of Emerging Technologies, and Wang Tianmiao, Honorary Director of the Robotics Institute at Beihang University, Dean of the Zhongguancun Zhiyou Research Institute, and Initiator of the Zhiyou Yari Science and Technology Innovation Platform. The two scholars engaged in a technical dialogue that ranged from the fundamental nature of embodied intelligence to the practical conditions under which it can be deployed at scale.
Xi Ning observed that although the country’s robotics technology sector started relatively late, after years of catching up and sustained effort it is no longer separated from the United States by a difference in “quality” but by a difference in “quantity” in terms of basic theory and methodology. In areas such as the supply chain, the country has even moved ahead of the United States. That assessment carries significant weight for the embodied intelligence community, because it suggests that the foundational science is no longer the primary bottleneck. Instead, the challenge lies in accumulation, engineering depth, and the volume of practical deployments that turn laboratory achievements into dependable industrial products.
Turning to artificial intelligence models designed for the physical world, Xi Ning argued that such models are four-dimensional. They must handle not only logical relationships but also spatial, temporal, and interactive relationships. This is a critical distinction for embodied intelligence. A language model that operates purely in the domain of text can afford to ignore physics. An embodied intelligence system cannot. Every action it takes unfolds in space, consumes time, and changes the state of the world around it, which means the model must reason about consequences as well as correlations.
Xi Ning further noted that as an emerging technology, the deployment of embodied intelligence will inevitably begin with rigid-demand, high-value-added industries before radiating outward into other fields. He expressed strong conviction that in the future there will be more types of embodied intelligence robots than there are mobile phones today. That prediction reframes the entire debate about form factor. Rather than converging on a single dominant design, embodied intelligence is likely to fragment into a wide spectrum of specialized embodiments, each tuned to the demands of a particular task environment.
Wang Tianmiao, drawing on nearly two decades of incubation and observation, offered a historical perspective on how the robotics industry reached its current state. He pointed out that four companies — FANUC of Japan, Yaskawa of Japan, ABB of Switzerland, and KUKA of Germany — made historic contributions to the development of the global industrial robotics industry. Their dominance shaped the standards, the supply chains, and the customer expectations that define the sector even today.
Against that backdrop, Wang Tianmiao highlighted that domestic robot manufacturers have climbed a steep slope over 15 years of struggle, and their share of the domestic market has now reached 57 percent. For him, this is not merely a commercial milestone but evidence of a deeper structural transformation. He stated that the emergence of world-leading embodied intelligence enterprises in China is a necessity of historical logic, and that such enterprises will occupy an important place in related industries globally.
Looking ahead, Wang Tianmiao predicted that over the next five to ten years, embodied intelligence will not be a winner-take-all market but rather a landscape of a hundred flowers blooming. Enterprises valued at the level of tens of billions of yuan and hundreds of billions of yuan may number in the dozens or even approach one hundred. That projection implies a far more distributed value chain than many observers assume, and it has direct implications for investors who are trying to decide where to place their bets.
He also rejected the idea that embodied intelligence can be reduced to a single narrative. According to Wang Tianmiao, the field encompasses far more than general-purpose humanoid robots. It also includes outstanding artificial intelligence large-model companies, ecosystem development tools and simulation and edge-side chip companies, companies that customize and iteratively root data models for vertical scenarios, dexterous hand and actuator companies, upstream core component little-giant companies, and even application integration, leasing, and service companies. Each of these categories represents a distinct opportunity within the broader embodied intelligence economy.
Participants at the event observed that the two scholars moved from speculation about the essence of technology to a discussion of landing paths, stepping beyond black-and-white arguments and presenting a balance between technological idealism and industrial reality. In doing so, they clarified the coordinates for the industry’s next stage of progress. For companies working on embodied intelligence, the message was clear: the question is no longer whether the technology is real, but where and how quickly it can be made useful.
2. Walking the Difficult but Correct Road
If the expert dialogue charted the long-term technical direction for embodied intelligence, the industrial conversation that followed pulled the focus back to the production line and dissected every practical detail of the journey from validation to deployment. Under the theme “The ‘Super Scenario’ of Embodied Intelligence: A Manufacturing Revolution,” the organizers invited You Wei, Chairman and General Manager of EFORT and Qizhi Robot, and Xu Yong, Chairman and Chief Executive Officer of Xizhun Robotics, to engage in a dialogue that reconstructed the truth of landing embodied intelligence in manufacturing scenarios from a frontline, hands-on perspective.
Drawing on EFORT’s many years of deep cultivation in industrial scenarios, You Wei argued that addressing the flexible operation problem of industrial robots requires a deep combination of artificial intelligence and robotics technology, adopting an embodied operation paradigm based on data-driven methods and neural networks. That paradigm shift is not cosmetic. It changes how robots are programmed, how they learn, and how they adapt when the task or the environment changes.
However, You Wei also identified a hard constraint. Under conditions of limited cost, it is extremely difficult to balance three properties simultaneously: inference speed, accuracy, and generalization capability. These three requirements pull against one another, and satisfying all of them at once remains out of reach with current architectures. In his view, two core problems must first be resolved — energy and computing architecture — before this currently impossible triangle can be balanced.
His practical guidance for companies working on embodied intelligence was notably candid. In actual deployment, he said, generalization performance can be sacrificed to a moderate degree. That is a significant statement in an industry that often treats generalization as the ultimate measure of intelligence. You Wei’s point is that reliability and speed in a narrowly defined task may matter far more to a manufacturing customer than the ability to handle an open-ended range of situations.
You Wei further observed that the current application of embodied intelligence in industrial scenarios is relatively homogenized, and that this homogeneity stems in part from a shortage of efficient tools available on the market for partners to carry out secondary development. Only by becoming open, and by cooperating with vertical partners who understand scenarios and understand end-user requirements, can effective division of labor across the industry chain and in commercial terms be achieved. That, in turn, is what will resolve such pain points and satisfy the diverse needs of thousands of industries.
The argument has an important corollary for the embodied intelligence ecosystem. A platform company that attempts to serve every vertical market on its own will inevitably spread itself too thin. A platform company that provides a powerful, open base and then empowers domain specialists may be able to reach far more applications than it could ever address alone. The economics of embodied intelligence may therefore reward openness rather than vertical integration.
Xu Yong approached the question from the perspective of landing precision manufacturing technology. He encouraged embodied intelligence enterprises not to fear new and unknown domains, and to bravely confront the multiple challenges of entering the embodied era — for example, changes in the core competencies of the team and the integration of different types of talent. He urged companies to walk the “difficult but correct road,” moving from short-term business toward long-term business.
That phrase captures a tension that runs through the entire embodied intelligence sector. Short-term business is easier to win: pilot projects, demonstration lines, and customized one-off deployments can generate revenue and credibility. Long-term business requires something harder — repeatable products, standardized interfaces, and a support model that scales without a proportional increase in engineering headcount. Companies that choose the difficult road are effectively betting that the embodied intelligence market will eventually reward durability over novelty.
Xu Yong also argued that for industrial embodied intelligence to truly land, deep cooperation across thousands of industries is indispensable, while embodied intelligence enterprises themselves should make good use of base platforms and focus on their own core capabilities. This division of responsibility mirrors the structure of mature technology industries, in which a small number of platform providers support a much larger population of application specialists.
Finally, Xu Yong reminded his audience that robots in the future are not necessarily humanoid, and will instead take diverse forms. This observation aligns closely with the earlier prediction that the variety of embodied intelligence robots will exceed the variety of mobile phones. It also serves as a caution against over-indexing on a single form factor when the underlying intelligence is what determines value.
3. Comparing the Perspectives Across Both Dialogues
The two conversations addressed the same subject from complementary angles, and the contrast between them is instructive for anyone trying to understand where embodied intelligence is heading. The table below summarizes the participants, their roles, and the central arguments they advanced.
| Dialogue | Participant | Role | Central Argument |
|---|---|---|---|
| Piercing the Fog of Technology, Seeing the Truth of Industry | Xi Ning | Chair of the Robotics Society of the Chinese Institute of Electronics; Chair Professor at the University of Hong Kong | The gap with the United States in basic theory and methodology has shifted from a difference in quality to a difference in quantity, with the supply chain already ahead; models for the physical world are four-dimensional; deployment starts with rigid-demand, high-value-added industries; the variety of embodied intelligence robots will exceed that of mobile phones. |
| Piercing the Fog of Technology, Seeing the Truth of Industry | Wang Tianmiao | Honorary Director of the Robotics Institute at Beihang University; Dean of the Zhongguancun Zhiyou Research Institute | Four global industrial robotics leaders made historic contributions; domestic manufacturers reached a 57 percent share of the domestic market after 15 years; world-leading embodied intelligence enterprises emerging in China is a necessity of historical logic; the next five to ten years will be a hundred flowers blooming rather than winner-take-all. |
| The “Super Scenario” of Embodied Intelligence: A Manufacturing Revolution | You Wei | Chairman and General Manager of EFORT and Qizhi Robot | Flexible industrial operations require data-driven, neural-network-based embodied operation paradigms; inference speed, accuracy, and generalization form an impossible triangle under cost constraints; energy and computing architecture must be solved first; generalization can be moderately sacrificed in real deployment; openness and vertical partnership are essential. |
| The “Super Scenario” of Embodied Intelligence: A Manufacturing Revolution | Xu Yong | Chairman and Chief Executive Officer of Xizhun Robotics | Embodied intelligence enterprises should not fear unknown domains; teams must confront changes in core competencies and the integration of different talent types; companies should walk the difficult but correct road from short-term to long-term business; deep cooperation across industries is required; robots will take diverse forms rather than necessarily being humanoid. |
Read together, the four viewpoints describe a market in which the technological foundation for embodied intelligence is largely in place, the supply chain is strong, and the principal obstacles are now economic and organizational rather than purely scientific. That is a very different diagnosis from the one that dominated the earlier phase of the embodied intelligence conversation, when the central question was whether the technology could work at all.
4. What the Manufacturing Super Scenario Demands of Embodied Intelligence
The choice of manufacturing as the first super scenario for embodied intelligence is not arbitrary. Manufacturing environments offer structured tasks, measurable outcomes, and a clear economic rationale for automation. They also impose constraints that many laboratory systems are not designed to satisfy. The dialogue between You Wei and Xu Yong surfaced several of these constraints, and they can be organized into a practical checklist for companies preparing to deploy embodied intelligence in production settings.
- Cost discipline: Inference speed, accuracy, and generalization compete for limited hardware budgets, and any embodied intelligence solution must be designed around a target cost rather than optimized in isolation.
- Energy and computing architecture: These two foundational problems must be addressed before the trade-offs among speed, accuracy, and generalization can be resolved rather than merely managed.
- Deliberate generalization trade-offs: In real deployments, narrowing the scope of generalization may be the fastest route to dependable performance.
- Open development tools: The relative homogeneity of current industrial embodied intelligence applications reflects a shortage of efficient tools for partners to build on top of.
- Vertical partnership: Companies that understand specific scenarios and end-user requirements are indispensable to effective industrial and commercial division of labor.
- Talent integration: Building embodied intelligence products requires merging teams with different core competencies, which is an organizational challenge as much as a technical one.
- Long-term orientation: Moving from short-term projects to long-term business requires standardization and repeatability, not just successful demonstrations.
- Form-factor agnosticism: Value in embodied intelligence derives from capability, not from resemblance to a human being.
Each of these items reflects a lesson learned in the field rather than a theoretical concern. Together they suggest that the second half of the embodied intelligence story will be decided less by breakthroughs in a single model architecture and more by the accumulation of engineering discipline across an entire ecosystem.
5. The Industry and Finance Dimension of Embodied Intelligence
The event was framed explicitly around industry-finance practice, and the presence of frontline hard-technology investors alongside scientists and operators reflects a broader recognition that capital allocation will shape which embodied intelligence paths are pursued. The roadshow format, which placed promising innovation and entrepreneurship teams in front of experienced investors, is itself a signal that the embodied intelligence sector is entering a phase in which execution quality and commercial traction matter as much as technical novelty.
Wang Tianmiao’s forecast that dozens or even close to one hundred enterprises at the tens-of-billions and hundreds-of-billions of yuan level may emerge over the next five to ten years implies a market structure very different from the platform monopolies that characterized the consumer internet. In embodied intelligence, value is likely to be distributed across model developers, simulation and edge chip providers, vertical data model specialists, dexterous hand and actuator makers, upstream component suppliers, and application integration and service businesses. An investor looking for exposure to embodied intelligence therefore has many distinct entry points, each with its own risk profile.
That distribution also has implications for the enterprises themselves. If the market is genuinely a hundred flowers blooming rather than a winner-take-all contest, then specialization is a viable strategy. A company that focuses narrowly on dexterous hands, or on simulation tooling, or on a single vertical scenario, may build a durable position without needing to become a full-stack embodied intelligence platform. Conversely, a company that attempts to compete on every front simultaneously may find itself outmatched by more focused rivals.
The emphasis on openness voiced by You Wei reinforces this logic. When a platform becomes open, it lowers the cost for vertical partners to build embodied intelligence applications, which expands the total market. When it remains closed, it limits the number of scenarios that can be addressed and slows the diffusion of the technology. For an industry that still needs to prove its return on investment across a wide range of use cases, openness is not merely a philosophical preference but a growth strategy.
6. Why Supply Chain Strength Matters for Embodied Intelligence
Xi Ning’s assessment that the country has moved ahead of the United States in supply chain terms deserves particular attention in the context of embodied intelligence. Embodied intelligence systems are hardware-intensive. They depend on actuators, sensors, reducers, motors, controllers, and structural components, in addition to the computational hardware that runs their models. A robust supply chain reduces costs, shortens development cycles, and makes rapid iteration possible.
Wang Tianmiao’s account of four global industrial robotics leaders and the 15-year climb of domestic manufacturers to a 57 percent share of the domestic market provides the historical context for that advantage. The industrial robotics supply chain that was built during that period is now being repurposed for embodied intelligence. Component suppliers that once served traditional industrial arms are now developing products for humanoid and other embodied platforms. That transition is one reason why the cost curve for embodied intelligence hardware may fall faster than many observers expect.
At the same time, supply chain strength does not automatically translate into product leadership. The dialogue between the two scholars and the dialogue between the two executives point in the same direction: the remaining work is about integration. Integrating artificial intelligence models with physical systems, integrating vertical domain knowledge with general-purpose platforms, and integrating teams with different skill sets into coherent organizations. These are the tasks that will determine which embodied intelligence enterprises become the world-leading companies that Wang Tianmiao expects to emerge.
7. A Roadmap for the Second Half of Embodied Intelligence
The conversations at the event collectively suggest a roadmap for the second half of the embodied intelligence era, one defined less by dramatic breakthroughs and more by systematic execution.
- Start where the demand is rigid and the value is high. Embodied intelligence will first take root in industries where the economic case is unambiguous, then radiate outward.
- Accept engineering trade-offs honestly. Speed, accuracy, and generalization cannot all be maximized at once under realistic cost constraints, and pretending otherwise delays deployment.
- Solve energy and computing architecture. These foundational constraints determine how far the trade-off frontier can be pushed.
- Build open tools for partners. The pace of embodied intelligence adoption depends on how easily vertical specialists can build on top of existing platforms.
- Specialize rather than monopolize. A distributed market rewards focused companies across models, components, tools, and vertical applications.
- Invest in long-term business design. Repeatability and standardization, not demonstration projects, create durable value in embodied intelligence.
- Remain form-factor agnostic. The value of embodied intelligence lies in capability, and capability can be delivered in many physical shapes.
Each of these steps is grounded in the experience of practitioners rather than in speculation. Taken together, they describe a transition from a period of technological exploration to a period of industrial consolidation, in which the winners will be determined by execution rather than by novelty.
8. Looking Ahead
The 2026 World Robot Conference’s embodied intelligence industry-finance practice session made clear that the sector has entered a more demanding phase. The excitement of demonstrating a new capability has given way to the harder work of making embodied intelligence reliable, affordable, and repeatable in real production environments. The two dialogues captured both sides of that transition: the long view from the scholars, who see a future populated by more varieties of embodied intelligence robots than there are mobile phones today, and the grounded view from the executives, who see the daily constraints of cost, energy, computing architecture, and integration.
What unites these perspectives is a shared conviction that the path forward runs through cooperation rather than isolation. Vertical partners who understand scenarios, platform providers who supply open foundations, component suppliers who drive down costs, and investors who are willing to fund long-term business building all have roles to play in the embodied intelligence ecosystem. No single company is likely to capture the entire opportunity, and no single form factor is likely to define it.
For the enterprises now competing in the embodied intelligence space, the message from the event is unambiguous. The technology has advanced far enough that the decisive questions are now about deployment discipline, open ecosystems, and sustained commitment to difficult but correct roads. Those who answer those questions well will shape the second half of the embodied intelligence era, and they will do so across a landscape of dozens of significant companies rather than a single dominant champion.
As the industry moves from concept validation to scaled landing, the value of gatherings that bring scientists, operators, and financiers into the same room becomes evident. Embodied intelligence will be built by many hands, and the conversations that took place at the 2026 World Robot Conference offer a working map for the journey ahead.