Chinese tech startup Galaxea (星海图) has raised over US$100 million in its A4 and A5 strategic financing rounds. The funding was led by Meituan Long-Z Investments, with contributions from Capital Today and existing investors such as MiHoYo and Wuxi Venture Capital Group.

The company focuses on embodied intelligence, integrating AI with physical systems for applications in robotics and automation. The funds will be used to enhance Galaxea’s research in “machine + intelligence.” This includes improving end-to-end models for visual-language-action systems, expanding production capacity, and developing joint modules.

Galaxea’s CEO, Jiyang Gao, a former senior engineer at Waymo, leads a team specializing in AI and embodied intelligence algorithms. Their research encompasses reinforcement learning, imitation learning, and full-body motion control. The company has developed a dual-system model for embodied intelligence, EFM-1, which combines visual language and action models for closed-loop decision-making.

Since launching its A-series financing in 2025, Galaxea has raised nearly 1.5 billion yuan (approximately US$210 million). Its valuation has more than tripled since the beginning of the year.

🔗 Source: AI Base

🧠 Food for thought

1️⃣ Embodied AI funding grows exponentially despite lengthy commercialization outlook

Xinghai Map’s $100+ million funding round reflects a broader surge in embodied AI investments, with 114 financing events recorded just in the first half of 2025 1.

This growth contrasts with earlier investment patterns, where pioneering companies like Embodied Intelligence (co-founded by AI expert Pieter Abbeel) secured just $7 million in 2017 2.

The sector’s financing dynamics have shifted, with early-stage companies like Xinghai Map raising $210 million in 2025 alone, while acknowledging the industry remains in an “infrastructure building phase” without widespread application 1.

Investors are showing willingness to make significant bets despite understanding commercialization timelines are lengthy, particularly as major tech companies like Meituan, JD, Tencent, and Huawei strategically position themselves in the space 1.

This investment pattern highlights a market shift from purely technological proof points toward securing competitive positioning in what many consider a foundational future technology.

2️⃣ Technical integration of vision, language and action marks new frontier in robotics

Xinghai Map’s dual-system EFM-1 architecture integrating visual language models with action models represents a significant evolution in robotics development approaches.

This approach aligns with advanced research pioneered by experts like Pieter Abbeel, whose work has focused on teaching robots complex skills through reinforcement learning and imitation learning 3.

The technical focus on end-to-end models for visual-language-action (VLA) systems addresses historical bottlenecks in embodied AI development, particularly around sensor accuracy and computing power limitations 1.

Recent Meituan developments