Deta AI Secures Five Funding Rounds in Six Months, Focusing on Foundational Models for Humanoid Robots

Delta Intelligence was founded in January 2026 as a spin-off from the Beijing Institute for General Artificial Intelligence (BIGAI). The company possesses a complete capability loop spanning foundational models, system architecture, and end-to-end deployment. Within less than six months of its inception, Delta Intelligence has completed five consecutive rounds of financing. In the Seed+ round, investments were jointly led by Oriza Holdings and Fosun RZ Capital. The Angel round was co-led by Meridian Capital, Cowin Capital, and Lenovo Capital & Incubator, with Oriza Holdings and Fosun RZ Capital participating in an oversubscribed follow-on investment. The Angel+ round received backing from financial investors including Zhuhai Science & Technology Industry Group, Haiwang Capital, and Plum Ventures, alongside deep participation from semiconductor industry investor Jiaoyuan Asset and an undisclosed automotive industry player, with Meridian Capital continuing to increase its stake. The two earlier funding rounds included participation from BIGAI, Leju Robotics, Agibot, Xinghaitu, and Hillhouse Capital. Ma Xiaojian, Founder and CEO of Delta Intelligence, previously worked at Google Robotics and NVIDIA Research, focusing extensively on large-scale machine learning and foundational models for embodied intelligence. Co-founder Liu Hangxin is currently an Assistant Professor at Peking University; he previously contributed to U.S. DARPA, ONR, and NSF robotics projects and, after returning to China, led multiple industrial and specialized applications of humanoid robots at BIGAI. Co-founder and Chief Scientist Huang Siyuan currently serves as Director of the Embodied Robotics Center at BIGAI and formerly worked at DeepMind and Meta. The company specializes in developing native foundational models for humanoid robots, centered around its proprietary 3D world engine, driving the evolution of humanoid robots from fragmented capabilities toward full-body coordination. Its model treats the entire humanoid body as a unified modeling object, building a closed-loop system around Loco-manipulation that integrates 3D scene understanding, task reasoning, whole-body control, and force-position interaction—enabling robots to coordinate their physical execution while comprehending tasks. To date, Delta Intelligence has achieved cross-platform adaptation and performance validation across multiple humanoid robot platforms, including Unitree, Agibot, and Leju Robotics. It has also successfully demonstrated real-world task execution in industrial and energy scenarios such as collaborative operations on FAW Hongqi production lines, substation inspection for China Southern Power Grid, and SMT reel sorting. The company has made significant progress in task success rates, continuous operation duration, and cross-scenario generalization capabilities. Delta Intelligence aims to deliver transferable and generalizable foundational models for humanoid robots across diverse hardware platforms, accelerating the transition from model training to real-world deployment through joint validation with robot manufacturers.