Octopus Power Raises Nearly RMB 1 Billion in 3 Months to Accelerate Development of "Hand-Brain Integrated" Embodied AI Platform

On June 17, SynapX (Octopus Dynamics) announced the completion of a new $50 million funding round. Over the past three months, the company has raised nearly RMB 1 billion in total. This round was co-led by Jinqiu Fund, Xinlian Capital, and Huangpujiang Capital, with participation from Jafco Asia, Houxue Capital, Hong Kong Top Talent Venture Fund (HKTT), and Panga Capital. Existing investors, including Linear Ventures, also continued their support. Additionally, the company’s next funding round—targeting RMB 500 million—is nearing completion. Founded in early 2026, SynapX focuses on building a sustainable, self-evolving embodied intelligence productivity system. The founder previously established Jianzhi Robotics and was among the founding team members of Horizon Robotics’ “Horizon 6” project and Baidu’s Institute of Deep Learning (IDL). All co-founders have backgrounds at either Horizon Robotics or Jianzhi Robotics. The team, now approaching 100 members, is primarily composed of talent from leading tech companies such as ByteDance Seed, Baidu, Xiaomi, Huawei, NIO, XPeng, Li Auto, and Meta. Centered around world foundation models, SynapX has built a closed-loop system spanning models, data, and real-world execution, leveraging its strategy models, SYNData—a multimodal visual-force-tactile data system—and a dexterous manipulation hardware platform. The company has proposed a three-stage evolutionary roadmap for world models: WM1 (Generation/Interaction), WM2 (Physics/Action), and WM3 (Universal Optimization). Currently, SynapX is focused on the WM2 stage, which emphasizes robust macro-scale physical modeling to enable models to predict changes in physical states and outcomes of actions. SynapX has also defined a five-level progression path for physical AI productivity—from L1 (Rule-Driven Systems) to L5 (Superhuman Physical Intelligence)—and introduced the SYNTH DeepThought architecture based on this framework. SYNTH integrates SYNData (real-world operational data accumulation), SYNWorld (physical causal learning and reasoning), and SYNAction (strategy-to-execution transformation) to achieve general-purpose action capabilities for embodied intelligence in real environments. Currently, SynapX’s physics-based action world model and its “integrated hand-brain” platform, both developed around the SYNTH architecture, have already garnered collaboration and validation from customers and industry partners globally. The company plans to launch its foundation model and debut core products in the second half of 2026.