星海图全球开发者大会：首秀双足人形机器人 启动百万小时真实数据计划

At the recent Xinghaitu Global Developer Conference, the company unveiled several core breakthroughs in embodied intelligence, sending a clear signal on the sector's development direction. Key takeaways are as follows:

1. Xinghaitu showcased its bipedal humanoid robot Kengo for the first time and open-sourced its new generation G0.5 foundation model, which tops six global authoritative benchmarks. Ranking first domestically and among the top tier internationally, the model makes Xinghaitu China's only embodied intelligence firm that boasts both cutting-edge models and top-tier hardware. Its three-year "full robot + AI" strategy is now officially complete.

2. The industry has entered a new development phase, shifting the core competition from individual device performance to ecosystem building, marking the start of an open, collaborative era for the sector.

3. At the conference, Xinghaitu partnered with Beijing Yizhuang to launch the "Million Hours of Real-World Data" initiative, and co-launched the Star Road Program to invest in early-stage embodied intelligence startups. It also outlined a three-stage commercialization roadmap, as well as a penetration strategy that expands from factory applications to home use, clarifying the future direction for the entire industry.

This summary outlines industry trends and positioning strategies for brands in the embodied intelligence space. Key takeaways are as follows:

1. New industry trends: Embodied intelligence has shifted from competing on individual product performance to open ecosystem collaboration. Isolated point technologies cannot support end-to-end development; only by opening core capabilities and bringing together industry partners can brands build competitive advantages. Brands can adopt an open ecosystem strategy to differentiate from closed, in-house development models.

2. Building core moats: Unlike model architectures which can be easily replicated, large-scale real-world data is the hardest-to-copy barrier for embodied intelligence. Brands can start laying the groundwork for real-world data collection early, leveraging China's advantages in hardware and data supply chains to build a data flywheel and solidify their competitiveness.

3. Guidance for commercialization and product launch: Brands should prioritize entry into structured scenarios and expand gradually, and evolve their business models in three stages. In the long run, players should focus on AI-driven solution subscriptions and token-based sales, rather than blindly chasing high early-stage full-unit sales volume.

This summary clarifies opportunity directions and risk warnings for practitioners entering the embodied intelligence track. Key takeaways are as follows:

1. Startup and partnership opportunities: As the industry enters the open ecosystem collaboration phase, industry leader Xinghaitu has co-launched the Star Road Program, which provides capital, technology and industrial resources to early-stage embodied intelligence startups. It has already invested in 18 related companies over the past year, and plans to invest in 30 to 50 more over the next 3 to 5 years. Founders across the upstream and downstream can apply for support, and many portfolio companies have already become partners of the leading firm.

2. Guidance for go-to-market strategy: Players can prioritize entry into structured scenarios with fixed environments such as factories, warehouse sorting and delivery, which only require centimeter-level accuracy. After accumulating capabilities and data, teams can expand into unstructured scenarios such as high-precision manufacturing, agriculture and home use, lowering early-stage go-to-market barriers.

3. Risk warnings: Embodied intelligence is an industry that will penetrate gradually over the long term, where competitiveness depends on cumulative capabilities, data and scenario iteration. Players should not blindly chase a short-term inflection point, and must prepare for long-term布局.

This summary outlines clear directions and business opportunities for factories pursuing intelligent transformation and entering the embodied intelligence track. Key takeaways are as follows:

1. Production opportunities and demand: Hardware body is a core component of the fast-growing embodied intelligence industry, and China already has a globally leading hardware supply chain. Following Xinghaitu's launch of its bipedal humanoid robot, demand for body manufacturing and component production is growing steadily. Relevant factories can connect with the leading firm's industrial chain to secure stable orders.

2. Opportunities for in-house intelligent transformation: Factory environments are the first entry point for embodied intelligence deployment. These scenarios have fixed working environments and only require centimeter-level accuracy to meet operational requirements. Factories can prioritize introducing embodied intelligence to upgrade their production and sorting processes, boosting productivity and reducing labor costs.

3. Development insights: The full embodied intelligence supply chain requires collaboration across multiple players. Factories can leverage their manufacturing advantages to join the leading firm's open ecosystem, become an upstream or downstream partner, gain access to technical resources, jointly advance the industry's digital and intelligent transformation, and capture a share of the sector's growth dividend.

This summary outlines industry trends, customer pain points and growth opportunities for service providers in the embodied intelligence space. Key takeaways are as follows:

1. Industry development trends: Embodied intelligence has shifted from individual device competition to open ecosystem collaboration. The industry has a long value chain with increasingly refined division of labor, and more segments now require participation from professional service providers. The market will continue expanding, creating significant growth opportunities for service providers.

2. Core customer pain points: The core bottleneck holding back breakthroughs in embodied intelligence capabilities is the lack of systematically collected large-scale real-world data. No company has previously completed systematic large-scale collection of this type of data, and data is the core fuel for AI advancement. This represents a key market segment that service providers can enter.

3. Growth opportunities: China has globally leading advantages in data supply chains, with clear strengths in both scenario diversity and data cost. Service providers can leverage these advantages to build a system for large-scale real-world data collection and operation, develop hard-to-replicate competitive advantages, and connect with leading firms' ecosystem initiatives to access partnership resources.

This summary provides guidance on ecosystem building and operations for platform players active in the embodied intelligence space. Key takeaways are as follows:

1. Core industry demand: The embodied intelligence industry has an extensive value chain covering AI, manufacturing, sensors, data, scenarios and multiple other segments. No single company can build out the full value chain independently, so the entire sector needs an open platform as an ecosystem foundation to bring together global developers and upstream-downstream partners for collaborative development.

2. Ecosystem building strategy: Platforms can follow the open-source foundation model playbook: open core capabilities to developers, allowing continuous model iteration through developer development, testing and feedback, while accumulating developer resources into the platform's own technology stack to build the foundation of the ecosystem. This open approach enables differentiated competition.

3. Ecosystem operation direction: Platforms can use a "capital + technology + industrial resources" model to incubate and invest in upstream and downstream startups to complete the industry ecosystem. They can also partner with local governments to build data companies to advance large-scale collection of core real-world data, and build the industry's core barriers. Players should also prepare for long-term布局 and avoid blindly chasing short-term hype.

This article discloses the latest industry developments in embodied intelligence, providing first-hand key information for industry research. Key content is as follows:

1. Latest industry dynamics: The embodied intelligence sector has officially entered a new development stage, shifting from competition centered on individual device performance to open ecosystem collaboration. Leading player Xinghaitu has completed its three-year "full robot + AI" strategic cycle, becoming China's only embodied intelligence firm that masters both cutting-edge models and top-tier hardware, bringing new changes to the industry landscape.

2. New core industry challenges: The industry has now reached a clear consensus: model architectures and parameters can be replicated by competitors, but only large-scale real-world data is a core barrier that cannot be replicated in the short term. Currently, no player in the industry has completed systematic large-scale collection of this type of data, which is the core new challenge restricting industry capability breakthroughs.

3. New reference for business models and go-to-market paths: Leading firms have proposed a three-stage commercialization path that progresses from full-unit sales to solution subscriptions, and eventually to token sales, alongside a go-to-market strategy of gradual outward penetration. The open ecosystem collaboration model also provides the industry with a new alternative to the closed in-house development model, offering high research value.