Embodied intelligence startup 眸深智能 raises 300 million yuan in Pre-A funding for robot brain platform

Shanghai-based embodied intelligence startup 眸深智能 (MOTION BRAIN) has raised 300 million yuan (approximately $42 million) in a Pre-A funding round, the company announced on May 14, 2026. The round was led by state-backed investors including Green Technology Capital (绿技行资本) and Fujian Provincial Investment Capital (闽招基金), with participation from listed companies such as Leo Group (利欧股份), an automotive parts supplier, and financial investors including Huafu Capital, Yingjue Glory, and Gengxin Capital. The company, founded in 2025 by three Fudan University alumni - serial entrepreneur Mu Zelin (born 1995), Professor Chen Tao, and former Intel China chief scientist Zhang Yimin - has now completed five funding rounds in six months, with two subsequent rounds already oversubscribed and in closing.

This funding surge reflects a broader shift in the embodied intelligence market from hardware-centric competition to software-defined intelligence. 眸深智能's core product is a "robot brain" platform built on its proprietary Motion GPT foundation model, which enables cross-robot deployment across humanoid, cleaning, logistics, and industrial robot forms. The company claims its T²MB embodied large model, released in March 2026, is the world's first autonomously evolving model that improves absolute performance by 25% after offline deployment. This positions 眸深智能 in the increasingly contested "brain" layer of the robotics stack, competing with firms like Agility Robotics, Figure AI, and domestic players such as Zhiyuan Robotics. The company has already achieved tens of millions of yuan in 2025 revenue and expects over 100 million yuan in orders and 1,000-unit product shipments in 2026, with a recent partnership to deploy 100,000 home-care robots with Yijia Elderly Care.

The practical test for 眸深智能 is whether its "one brain, many forms" approach can deliver reliable, repeatable performance across diverse hardware platforms and real-world environments. The company's rapid fundraising cadence - five rounds in six months - signals strong investor conviction, but the embodied intelligence sector remains early-stage, with few companies having demonstrated sustained production-scale deployments. The key execution milestones will be the 2026 shipment target and the ability to maintain model performance improvements post-deployment, which the T²MB model claims to address. For buyers evaluating robot brain platforms, the critical question is whether the Motion GPT architecture can generalize across different robot hardware and task domains without requiring extensive per-deployment retraining.