"DeepSimple AI" Closes Hundreds of Millions in Pre-A Series Led by DiDi, Linear Capital Follows On | Linear Portfolio

Today, Simple AI announced the completion of a Pre-A funding round worth several hundred million yuan, led by DiDi — marking the company's first introduction of strategic capital with dual expertise in industrial scenarios and robotics supply chains.

Previously, Simple AI partnered with CTS Hotels Group to launch the industry's first proof-of-concept validation of embodied robotics in a hotel setting at Beijing Lido Vison Hotel. The deployment tests the robot's autonomous perception, task execution, and multi-scenario coordination capabilities in complex, dynamic real-world hotel operations.

Linear Capital was a co-lead investor in Simple AI's angel round and increased its stake in this round.

Recently, general-purpose embodied intelligence robotics company Simple AI announced the completion of a Pre-A funding round worth several hundred million yuan. The round was led by DiDi, with Plum Ventures and Keli Sensing as co-investors. Existing shareholders CCV, Linear Capital, and Puhua Capital continued to increase their stakes. To date, Simple AI has completed four consecutive funding rounds within a year, securing comprehensive support from top-tier investment institutions, internet platforms, and industrial capital. This round marks Simple AI's first introduction of strategic capital with dual backgrounds in industrial scenarios and robotics supply chains. The entry of industrial investors represents not only strong recognition of Simple AI's technical capabilities and commercial potential in embodied intelligence, but also means the company has gained access to real service scenarios, supply chain resources, and industrial synergy capabilities — laying a solid industrial foundation for the scaled production and scenario deployment of embodied robots.

As the embodied intelligence industry gradually shifts from "model capability validation" to "real-world deployment," the core of competition is moving from parameter scale toward robots' understanding of, generalization across, and continuous learning from the physical world. As one of the earliest domestic companies focused on lightweight world models for general-purpose embodied intelligence robotics, the team proposed a unified foundation model called Simple-World, combined with a hierarchical memory-enhanced agent architecture that significantly improves robots' complex long-horizon task planning, cross-embodiment zero-shot adaptation, and operational generalization and instruction following in non-standard scenarios — translating commercially to shorter adaptation cycles for new scenarios and lower deployment costs. On the data front, the company developed a self-built, low-cost, high-precision pure-visual UMI + Ego collection system, enabling large-scale, low-cost collection of multimodal data in real scenarios. At the same time, the company has built a complete real-machine operational data feedback loop, driving autonomous robot evolution through efficient continuous learning, forming a dual flywheel of "smarter with use, more efficient with delivery" for both technology and business. Building on these foundations, Simple AI has achieved a series of breakthroughs in the international academic and technical community: strong performance in the authoritative international BEHAVIOR 1K embodied intelligence challenge, demonstrating leading complex planning and operational generalization capabilities in long-horizon tasks; multiple original research contributions on world models and embodied manipulation submitted to top international conferences including NeurIPS, CoRL, and RSS, with concurrent filings for multiple core technology invention patents; and co-organization of the World Models Workshop at ECCV 2026, a top international computer vision conference, focusing on core topics including "Trustworthy World Models" and "Embodied AI," with deep participation in global frontier research collaboration on embodied intelligence and world models.

Simple AI is also continuing to expand its core technical team, further strengthening the critical capability loop between frontier research and industrial commercialization. Dr. Weitao Zhou currently serves as Co-Chief Scientist at Simple AI, responsible for embodied foundation model training and continuous learning system development. He holds a Ph.D. from Tsinghua University and completed postdoctoral research as a "Shuimu Scholar," with long-term focus on physical AI in long-tail scenarios, reinforcement learning, and continuous learning. His research has been published in top international journals and conferences including Nature Machine Intelligence, T-ITS, ICML, and IROS. He has won honors including the global Mcity Challenge autonomous driving competition championship and the National Invention Exhibition Gold Award. His research has been validated at scale in real-world environments, forming a complete technical system spanning data feedback loops, model training, and online iterative optimization. His arrival further strengthens Simple AI's core capabilities in embodied foundation model training, continuous learning, and physical world intelligent evolution, creating synergy with the team's existing foundational manipulation models and robotics deployment capabilities to accelerate the development and commercialization of next-generation embodied intelligence foundation models. Dr. Jinming Ma currently serves as Senior Researcher at Simple AI, primarily responsible for foundational manipulation model training, post-training, and physical world deployment. He earned his Ph.D. from the University of Science and Technology of China, with long-term expertise in foundational manipulation models, reinforcement learning, and multi-agent systems, having published more than ten papers in international journals and conferences including T-ITS, RA-L, ICRA, AAMAS, IROS, and IJCAI. He previously led UMI system development and foundational manipulation model algorithm research at a leading internet platform company's robotics lab. His arrival further strengthens Simple AI's capabilities in foundational manipulation model training, post-training optimization, and physical world execution feedback loops. Competition in embodied intelligence is not merely about model capabilities, but a comprehensive contest of systems engineering capability, scenario understanding, and long-term productization ability. As the core team continues to expand, Simple AI is building out a complete R&D system covering world models, robot control, multimodal interaction, systems engineering, and scenario operations, further solidifying the company's long-term technical moat in general-purpose embodied intelligence robotics.

Simple AI partnered with CTS Hotels Group to launch the industry's first proof-of-concept validation of embodied robotics in a hotel setting at Beijing Lido Vison Hotel. In real hotel operations, Simbot has begun handling item delivery, public area patrols, guest room services, and basic cleaning tasks, continuously validating autonomous perception, task execution, and multi-scenario coordination capabilities in complex, dynamic environments. Meanwhile, at the Huawei Cloud INSPIRE 2026 conference, Simbot demonstrated natural human-robot interaction, high-precision motion execution, and scenario-adaptive operation capabilities, and became one of the first companies to join Huawei Cloud's Industry AI Dream Factory embodied intelligence zone. The two parties jointly launched the "Hundred Models, Thousand Forms, Cloud Convergence for Mutual Success" ecosystem partnership plan, collaborating with Huawei Cloud to build CloudRobo, a full-loop development platform spanning "data — training — deployment — execution." This partnership marks Simple AI's transition from single-point validation to industrial-grade ecosystem collaboration. Following this funding round, Simple AI will further accelerate product iteration and scaled commercial deployment of embodied intelligence robots in real scenarios, with focus on Simple-World Model iterative upgrades, embodied intelligence data system development, core talent recruitment, and deployment in general household scenarios — continuously advancing robots from single-point capability validation toward long-term service capability building in the real world.