Xinghaitu Raises $280M at $2.8B Valuation

Post LinkedIn

🐼Read original on Pandaily

#embodied-ai#robotics-funding#vla-modelsxinghaituxinghaitu

💡$280M raise fuels embodied AI robotics scaling + VLA models

⚡ 30-Second TL;DR

What Changed

Raised nearly RMB 2B ($280M) in funding

Why It Matters

The massive funding positions Xinghaitu as a leader in embodied AI, accelerating robotics innovation and potentially disrupting automation sectors.

What To Do Next

Evaluate Xinghaitu robotics APIs for embodied AI prototypes in your projects.

Who should care:Founders & Product Leaders

Key Points

•Raised nearly RMB 2B ($280M) in funding
•Valuation surpasses RMB 20B ($2.8B)
•Scaling robotics deployment efforts
•Advancing vision-language-action and world models

🧠 Deep Insight

Background and context from public sources — not the original article. 3 sources cited.

🔑 Enhanced Key Takeaways

•Xinghaitu (also known as Galaxea AI) employs a 'whole machine + intelligence' strategy, developing both the R1 series of wheeled humanoid robots and the EFM (Embodied Foundation Model) software stack.
•The company has transitioned from serving a developer-focused market—with approximately 150 customers including academic institutions like Stanford—to targeting industrial productivity applications in manufacturing, logistics, and commercial services in 2026.
•Xinghaitu's technical approach includes the 'Robotic Spatial Intelligence Engine' (RSR), described as a Real2Sim2Real robotics engine, and the G0/G0 Plus VLA models, which are designed for cross-scenario task generalization.

📊 Competitor Analysis▸ Show

Competitor	Key Focus	Notable Hardware/Model
Unitree Robotics	Humanoid/Quadruped Hardware	H1/G1 Humanoids, Unitree Go series
Galaxy Universal	Embodied AI Brains	General-purpose embodied models
Agility Robotics	Bipedal Logistics	Digit
Apptronik	Humanoid Hardware	Apollo

🛠️ Technical Deep Dive

•Hardware Platform: R1 series (Lite, Standard, Pro) featuring wheeled bases, dual-arm configurations, and up to 26 degrees of freedom.
•Compute: Powered by NVIDIA Jetson AGX Orin 32GB (8-core CPU, 200 TOPS GPU).
•Model Architecture: Transformer-based VLA (Vision-Language-Action) foundation models (EFM/G0 series) designed for end-to-end control.
•Data Strategy: Utilizes a Real2Sim2Real (RSR) engine to bridge the gap between simulation and physical deployment; collected ~100,000 hours of real-machine data in 2025.
•Capabilities: Supports isomorphic and VR remote control, autonomous task planning, and multi-modal sensor integration (7 HD cameras, LiDAR).

🔮 Future ImplicationsAI analysis grounded in cited sources

Xinghaitu will achieve 'result verification' in industrial settings by late 2026.

The company is shifting its primary focus from developer-centric POCs to end-to-end manual task substitution in manufacturing and logistics.

The company will diversify its hardware portfolio beyond wheeled platforms.

Management has publicly announced plans to release dexterous hands and bipedal humanoid robots in May 2026 to expand their physical form factor capabilities.