蚂蚁为何加码具身智能？

This article clearly maps out Ant Group’s end-to-end strategy for scaling its presence in the embodied intelligence sector, as well as the industry’s current development stage, serving as a practical, accessible reference for general readers looking to quickly understand progress in the embodied AI space.

1. You will gain a clear overview of Ant’s core layout moves: In the second half of 2024, Ant’s management concluded that failing to invest in embodied intelligence would leave the company irreversibly behind in future development, and launched a dual-track strategy combining in-house R&D and external investment. As of September 2026, Ant has led more than eight public investment rounds, covering critical nodes across the full industrial chain from robot hardware and core components to data infrastructure and general-purpose large models. Its subsidiary Ant Lingbo focuses on building universal robot brains, with the goal of becoming a core infrastructure provider in the era of physical AI.

2. You will learn the real state of the industry: Embodied intelligence is currently in the early stage of mass production and commercial deployment, with the biggest bottleneck being insufficient generalization capabilities. Deploying a robot for a single scenario can cost as much as RMB 2–3 million, and universal robot brains are seen as the key breakthrough to resolve this pain point. Ant’s open-source embodied model already supports more than 20 robot models from 17 manufacturers, ranking among the most developer-friendly offerings in the industry.

3. You will receive clear, actionable industry signals: Embodied intelligence will eventually cover household, industrial, and commercial scenarios. As universal robot brains mature, robot deployment costs will continue to fall, and robots will gradually enter daily production and life. However, the sector remains in an early stage, so consumers should avoid purchasing unproven consumer-grade robot products and be wary of overhyped marketing claims in the space.

The embodied intelligence industry progress outlined in this article provides clear decision-making references for brands to track technology trends, plan future product roadmaps, and prepare for emerging consumer scenarios.

1. Brands should map commercial scenario trends for the next three years in advance: Embodied intelligence is approaching the mass deployment stage, and will ultimately reshape service models across household, industrial, and commercial offline scenarios. Real-world use cases have already emerged, such as three different types of robots collaborating to complete tasks in pharmacies. Brands need to anticipate how robot-enabled services will impact offline consumer experiences and brand touchpoints, and prepare to lay out presence across these new interaction points.

2. Brands can identify new brand interaction opportunities driven by maturing technology: The core industry bottlenecks at present are weak robot generalization capabilities and high deployment costs. As major players including Ant advance open-source universal robot brains and improve cross-hardware adaptability, the integration cost for brands to adopt robot services will continue to decline. Brands can explore new brand communication formats such as robot shopping guides and intelligent interactive experiences ahead of mass adoption.

3. Brands should recognize the emerging window for ecosystem collaboration: Ant is building an embodied intelligence ecosystem through investments and open-source models, and has also entered the consumer-grade small humanoid robot segment for household use. Brands can proactively engage with such ecosystem platforms early to secure premium brand exposure positions in robot service scenarios and on home intelligent terminals.

The breakdown of Ant’s embodied intelligence layout strategy and industry pain points in this article points sellers in relevant sectors toward new market opportunities, replicable business playbooks, and potential collaboration directions.

1. Sellers should capture clear incremental market opportunities: Competition in the embodied intelligence space is shifting from robot hardware itself to upstream positioning in data and model segments. Large technology companies are heavily investing in areas including real-world data collection, model evaluation infrastructure, multimodal data services such as haptic data, and world model technology projects, creating clear growth room for associated upstream and downstream supply chains and supporting services. At the same time, major players have already entered the consumer-grade small humanoid robot market, signaling early growth in the C-end consumer segment.

2. Sellers can learn replicable business playbooks: Ant has adopted a model of using investments to pave market entry, developing core technology in-house, and leveraging open source to aggregate ecosystem partners. It is moving up the value chain from downstream hardware to upstream core technologies in line with industry development rhythms, binding upstream and downstream partners through investments, and rapidly scaling its base of developers and partners via open-source models, with the end goal of becoming a core ecosystem infrastructure provider. This playbook is a useful reference for vertical-sector sellers building their own partnership networks.

3. Sellers should balance risks and collaboration opportunities: Technology roadmaps for embodied intelligence have not yet converged, and single-scenario deployment costs remain high. Large-scale commercialization will take time until universal robot brains drive an exponential drop in post-training data costs, so sellers should avoid overinvesting blindly in a single technology route. At the same time, sellers can proactively connect with open ecosystems such as Ant Lingbo to reduce their own model adaptation costs by leveraging large players’ technical capabilities.

The embodied intelligence industry chain demand and deployment progress disclosed in this article provide clear references for factories to identify new order opportunities and advance smart manufacturing upgrades.

1. Factories can tap clear supporting production business opportunities: Major players including Ant are deploying across the full embodied intelligence industrial chain, with 2025 investment priorities focusing on robot hardware and core components, including cable-driven humanoid robots, joint modules, dexterous hands, and humanoid robots. This will steadily release demand for high-precision components and standardized hardware assembly. Factories with precision manufacturing capabilities can connect with Ant’s portfolio companies to compete for spots in core supplier systems.

2. Factories can track new solutions for smart manufacturing upgrades: The core pain point of deploying robots in factories at present is the high cost of retraining robots for individual tasks, with a full deployment setup costing RMB 2–3 million per scenario. Universal embodied brains rolled out by players such as Ant support cross-hardware reuse, which can drastically cut training and adaptation costs for robot deployments. Factories can track the maturity of these universal brains, and plan robot replacement roadmaps for production lines in advance to reduce digital transformation costs.

3. Factories can explore new value-add opportunities: As of 2026, industry investment focus has shifted upstream to segments including data collection and world models. During supporting production processes, factories can simultaneously accumulate robot operation data from real production scenarios; this type of scenario-specific data will become a core, high-demand input for embodied model training in the future, and can develop into a new value-added business stream for factories.

This article’s overview of the embodied intelligence industry’s development stage, core pain points, and large players’ layout directions offers clear guidance for service providers to identify service entry points and develop tailored solutions.

1. Service providers should accurately track industry development trends: The embodied intelligence sector has moved past the 2025 stage of core hardware capability upgrades, and entered a critical mass deployment period in 2026. Competitive focus is shifting from hardware to intelligent capabilities, with real-world data, general-purpose models, and evaluation infrastructure representing the core segments where large players are concentrating investments. Service providers can design supporting service offerings around these core links.

2. Service providers should precisely target core customer pain points: Unmet demand exists across every link of the industrial chain: Robot hardware manufacturers face weak generalization capabilities and high costs for multi-scenario adaptation; enterprise customers in sectors such as manufacturing struggle with high single-scenario deployment costs and the need for full retraining for new tasks; model developers face high pre-training computing and data costs, as well as fragmented technology roadmaps. All of these pain points represent viable entry points for service providers.

3. Service providers can pursue collaboration opportunities by connecting with large players’ ecosystems: Ant is building an embodied intelligence ecosystem via an investment plus open-source model, and its universal brain model already supports more than 20 robot models from 17 manufacturers. Going forward, it will serve the industry by exporting foundation model capabilities and opening up its toolchains. Service providers can build vertical-scenario deployment solutions on top of such open-source foundation models to reduce investment in underlying technology R&D.

This article’s breakdown of Ant’s platform-style playbook for building its embodied intelligence business provides practical references for all types of platforms launching new business lines and entering frontier technology sectors.

1. Platforms can reference a proven playbook for building platform ecosystems in emerging sectors: When entering embodied intelligence, Ant did not carry the business burden of tying the new sector to its existing, relatively shallow offline scenario business, instead adopting a dual-track approach to build its ecosystem. On one hand, it invests in high-quality companies across critical industry nodes in line with sector development rhythms to bind core partners; on the other, it develops core universal brain capabilities in-house, and uses open-source licensing to attract developers and hardware manufacturers to join, rapidly scaling ecosystem size as it targets the position of core infrastructure for physical AI. This “investment-led partner binding + open-source core capabilities + ecosystem collaboration” model is a strong reference for building platforms in new technology tracks.

2. Platforms should clarify core priorities for merchant recruitment and operations: Core participants in the embodied intelligence sector currently include robot hardware manufacturers, core component suppliers, data and model technology service providers, and scenario deployment customers. Platform recruitment should cover critical nodes across the full value chain, with particular focus on onboarding service providers for current bottleneck segments including data collection, model evaluation, and world models. On the operations side, platforms can reduce partner onboarding costs via open-source tools and ecosystem matchmaking events.

3. Platforms should mitigate operational risks: Technology roadmaps for embodied intelligence have not yet converged, there is no industry consensus on the final form of core models, and single-scenario deployment costs remain high. Platforms should avoid overcommitting to a single technology route too early, instead building multi-layered positions across key nodes, while balancing long-term R&D investment with commercialization timelines to avoid operational risks from overconcentrated resource allocation.

This article presents a full picture of Ant Group’s embodied intelligence layout path and current industry development status, providing a vivid research sample for scholars studying how digital platform companies expand into hard technology, as well as the evolution of the embodied intelligence industry.

1. Researchers can observe clear new industry trends: Unlike other major technology players, whose embodied intelligence layouts primarily focus on serving their existing ecosystems and building customer pipelines for cloud businesses, Ant leverages its lack of heavy offline scenario baggage to pursue a dual strategy of in-house universal brain R&D and full value chain investment. It is moving upstream from downstream hardware to data and model segments to build “one brain, multiple machines” cross-hardware adaptation capabilities, with the goal of becoming a new type of infrastructure in the physical AI era. Its model of aggregating ecosystems via open source and binding partners via investments represents a typical new pathway for technology platforms entering frontier technology sectors.

2. Researchers can focus on emerging core industry challenges: The embodied intelligence sector is currently in the early stage of mass deployment, facing two core bottlenecks. On the technology side, model generalization capabilities remain insufficient, technology roadmaps have not converged, and multiple routes including vision-language-action (VLA) models and world models are being explored in parallel. On the commercialization side, single-scenario deployment costs run as high as RMB 2–3 million, with high post-training data costs restricting large-scale rollout. Both issues represent important directions for industry research.

3. Researchers can track the evolution of frontier business models: Ant Lingbo’s future commercialization directions include exporting foundation models and providing data toolchain capabilities to industry peers, positioning itself as the “foundation model behind foundation models.” This model, which avoids tying itself to a single hardware provider and instead positions itself as a provider of cross-hardware universal capabilities, offers a new research sample for commercialization pathways in the embodied intelligence sector.