The "Physical AI and Embodied Intelligent Robot Innovation Ecosystem Exchange," a featured event of the China International Fair for Trade in Services, took place in Beijing on September 11, 2026. Yu Tianyu, Industry Partner and Head of the Industrial Research Institute at Kailian Capital, attended and delivered remarks during the event's dialogue session.

Yu began by introducing Kailian Capital, noting its Beijing headquarters and its decade-plus track record as a comprehensive equity investment firm primarily managing RMB funds. The firm's investment focus spans semiconductors, advanced manufacturing, automotive, and energy, alongside ongoing coverage of healthcare and consumer sectors. Yu has personally overseen embodied intelligence initiatives over the past two years, building a portfolio that includes Zhiyuan, Xingyuanzhi, Tashi Zhihang, Dijia Robotics, and Critical Point Technologies, as well as projects incubated by the Beijing Zhiyuan Research Institute. Kailian manages over 10 billion RMB in total assets, historically concentrating on growth and late-stage investments, though its embodied intelligence efforts have notably shifted toward earlier-stage opportunities in recent years.

Moderator Cui led the discussion by reflecting on the rapid passage of time in the robotics sector over the past five years. He recalled entering the field around five years ago, not as a professional robot investor but as a software engineer, and noted how companies like UBTech reached the secondary market during this period. Some startups advanced from angel funding rounds directly to Hong Kong IPO readiness, while others completed significant transformations over those 60 months.

Yu shared his personal journey, recalling his first WRC attendance in 2018 at a KOHLER forum, the year KOHLER went public. At that time, industry discussions centered on relatively defined products like floor-sweeping and inspection robots, with far less industry influence and participant scope than today. He also highlighted 2019 as a memorable year when, during his consulting days, he led industry research for CICC's team assisting UBTech in planning its domestic listing. UBTech, already a notable robotics company with CCTV exposure, had established meaningful revenue streams in entertainment performance and education scenarios by then. Yu's team conducted three-to-five-year and longer-term industry projections, including how market space might evolve as robots entered more human-machine interaction and industrial settings. While some technology and application directions have since materialized, commercial progress in industrial scenarios has lagged his 2019 expectations. He observed that predictions often rely on linear extrapolation, whereas reality unfolds through inflection points, nonlinear shifts, and accelerated phases, suggesting the industry may now be entering an acceleration period at such a turning point.

When pressed on whether his current outlook differs from his 2017-2018 expectations, Yu explained that past discussions of robotics remained within a relatively narrow scope. The vision of robots entering every industry and household has existed for decades, but around 2017-2018, commercial discussions were limited to specific robots in particular industries and processes. This explains why floor-sweeping and inspection robots dominated exhibitions—technological capabilities simply constrained the problem-solving boundary. Today's defining shift is the significant expansion of that technical capability frontier. Previously limited to narrow scenarios, robots can now address broader sets of use cases, and this boundary will continue widening. The key difference now is not renewed imagination about robots, but the industry approaching a genuine inflection point where those long-held visions begin translating into commercial reality.

Yu supplemented observations on the secondary market, outlining how domestic capital markets have mapped embodied intelligence in stages. The first phase followed Tesla's Optimus emergence, particularly post-2024, as product iterations and demos drove A-share investors to seek supply chain plays—driven largely by thematic speculation and industry expectations. The second notable milestone came as leading embodied intelligence firms like Zhiyuan established more direct connections with A-shares through capital operations and industrial consolidation, signaling a shift from pure technology storytelling to actual capital market and industry integration. This spawned recognizable "Zhiyuan concept" and "Unitree concept" sectors. The third phase, unfolding this year, sees the broader tech sector experiencing significant rallies and sector rotation, with more capital flowing into tech assets. Robotics, as a premier tech investment direction, has expanded its market influence and capital reach. Looking ahead, as representative companies like Unitree enter public markets, domestic robotics will develop clearer valuation anchors. Although UBTech already trades on the Hong Kong exchange, its impact mechanism differs for A-share investors. Should more leading companies access domestic capital markets, robotics asset pricing logic will inevitably undergo renewed transformation.

The conversation then shifted to commercialization closure, a frequently raised question. The moderator offered his perspective that robotics is an established industry with predetermined demand—citing a 2016 fund that invested in cloud computing, AI, and robotics, targeting applications in sanitation, cleaning, industry, and retail, which remain the same today. While technology has matured and robots grown more intelligent, the fundamental question remains: have commercialization closure approaches changed?

Yu argued for examining B2B and B2C scenarios separately. B2B contexts are more straightforward to delineate since enterprise customers ultimately calculate returns. He proposed three evaluation dimensions: capability, cost, and scalability. First, capability—whether robots can reliably execute tasks, meeting required success rates, cycle times, and consistency. Second, cost—not merely technical feasibility, but whether the economics work relative to existing labor, automated equipment, or alternatives, validating ROI. Third, scalability—whether solutions replicate across customers rather than demanding bespoke projects at each factory. While technological progress continues expanding solveable scenarios, few companies currently excel across all three dimensions simultaneously. Some satisfy two. Specialized robotics, for instance, shows promising commercial deployment because these fields tolerate higher costs, prioritizing functional achievement. However, these tend toward project-based models with significant customer variation, consuming substantial service and relationship resources.

When asked about the specialized sector's prospects, Yu expressed a measured view: quality companies can emerge, but platform-scale ambitions may face ceilings. Conversely, perceptive robotics companies are increasingly deliberately selecting scalable scenarios from inception. Manufacturing divides into dozens of sub-industries with vastly different standardization levels across processes. The most promising scenarios align where short-term technical capability is adequate, economics pencil out, and customer-to-customer demand variation remains manageable. When capability, cost, and scalability align, significant B2B robotics companies can emerge.

Yu cited a cross-sector example from his recent meeting with DJI's enterprise business team in Shenzhen. Broadly interpreted, drones constitute intelligent robots, and DJI's agricultural drones exemplify successful commercialization. They address a highly vertical function—crop protection and seeding—yet achieved enormous markets through superior efficiency versus manual labor and viable cost structures. DJI now expands toward heavy-lift transport, crane operations, and medium-short-distance logistics, with products like FC30, FC100, FC200 already commercially deployed, plus the newly showcased EV50 eVTOL drone for longer-range transport. The lesson: pursuing so-called general-purpose robots isn't mandatory. A seemingly narrow industrial scenario can birth extremely large products and companies if market size, efficiency gains, and cost economics align. Pan-industrial robotics likely follows similar commercial logic.

B2C scenarios differ fundamentally. Yu analogized to assisted driving, where quantifying economic value for consumers proves difficult—purchasers don't calculate annual savings but weigh experience, convenience, and desirability. Future household robots likely follow similar logic: performing household tasks, assisting activities, or providing companionship and emotional value. Consumers pay when they perceive utility and enjoyment. Thus, B2B hinges on ROI calculations, while B2C turns on experiential value—a crucial distinction in commercialization approaches.

When asked about his most preferred closed-loop scenario, Yu highlighted automotive wiring harness assembly, the current focus of portfolio company Tashi Zhihang. Flexible harnesses have resisted traditional automation due to demanding visual, force control, and fine manipulation requirements, yet represent genuine large-scale manual processes. If robot technology achieves stable performance, costs gradually undercut labor, and replicability holds across automotive and harness factories, this scenario satisfies all three conditions—capability, cost, and scalability. Such applications represent representative commercialization entry points for embodied intelligence at this stage.

The moderator closed by acknowledging the valuable multi-perspective discussion, expressing hope for new industry breakthroughs in the coming year, and concluding the session with thanks to all participants.