- Hangzhou startup secures second major round in 30 days, bringing total financing to 3 billion yuan
- Real-world data strategy and early industrial deployments draw investor backing
Spirit AI, an embodied AI and robotics company, today announced that it had raised 1 billion yuan ($145 million) in a new Series A+ funding round, underscoring investor appetite for embodied AI startups racing to translate large-model advances into real-world robotics applications.
The announcement comes barely a month after the Hangzhou-based company completed a nearly 2 billion yuan Series A financing, lifting total capital raised within 30 days to about 3 billion yuan.
The latest round was jointly led by Shunwei Capital and YF Fund, with participation from investors including Fortune Capital, Galaxy Securities, Turing Fund, Xinding Capital, and Glacier Capital.
Notably, Shunwei is the venture capital firm established by Xiaomi founder Lei Jun and YF Fund counts Alibaba founder Jack Ma as one of its two co-founders.
The rapid back-to-back fundraising highlights growing confidence in companies pursuing data-driven scaling strategies for physical AI — an area increasingly viewed as the next frontier after generative AI software.
At the core of Spirit AI’s approach is what it calls a “diversified data” framework aimed at extending the Scaling Law into embodied intelligence.
In the context of AI, the Scaling Law refers to the empirical observation that a model’s performance improves predictably and consistently as you increase the three key factors of compute, data, and model size.
The company said it has accumulated more than 200,000 hours of real-world interaction data spanning internet video parsing and learning, teleoperation, and wearable-based collection systems, with total datasets expected to exceed 1 million hours by 2026.
A key enabler has been its self-developed wearable data-collection device, now in its fifth generation. Utilizing a universal manipulation interface (UMI) approach, the system has reduced data acquisition costs to roughly one-tenth of traditional methods, helping remove a major bottleneck to large-scale physical-world training, Spirit AI said in a press release.
To support the expansion, the startup plans to grow its data engineering team to about 1,000 people by April 2025, aiming to continuously feed higher-quality interaction data into model iteration.
Commercial deployment is advancing alongside data development and fundraising efforts. In March, the company signed a strategic partnership with JD.com, an e-commerce titan, integrating its Moz robot into JD MALL smart retail environments to perform precision coffee-making tasks.
Separately, Spirit AI said it has launched what it describes as the world’s first humanoid embodied-intelligence production line at CATL’s Zhongzhou base in northern China’s Henan Province, deploying the Moz robots to help install power batteries.
According to the company, Moz robots operating on battery PACK assembly lines have already produced nearly 1,000 battery units with zero failures, maintaining insertion success rates above 99% while matching the operating tempo of skilled human workers.
Spirit AI added that the robots can autonomously adapt to uncertainties such as shifting component positions and changing connection points — capabilities investors increasingly see as critical for scaling embodied AI into industrial settings.
