Key facts
- Chinese AI startup PsiBot has raised 2 billion yuan ($280 million) in angel and Pre-A funding rounds.
- The company's valuation has exceeded $1 billion.
- PsiBot focuses on embodied robotics and the AI models and services needed to train and operate them.
- The funding announcement occurred during China's 'Two Sessions' political meetings, aligning with the government's emphasis on AI and robotics in its 15th Five Year Plan.
- PsiBot employs a 'small full-stack' approach, controlling key hardware design elements while outsourcing component production and manufacturing.
Chinese AI startup PsiBot, also known as Lingchu Intelligence, has secured 2 billion yuan ($280 million) in its angel and Pre-A funding rounds, achieving a valuation exceeding $1 billion. The significant capital infusion highlights China's increasing investment in embodied artificial intelligence and the supporting infrastructure.
PsiBot's strategy centers on developing both the physical robots capable of performing practical tasks and the AI models, training systems, and deployment services required for their operation. This 'small full-stack' approach involves controlling crucial aspects of hardware design, such as structure and motion, while relying on specialized suppliers for component production and manufacturing.
The timing of PsiBot's funding announcement, during China's annual 'Two Sessions' political meetings, aligns with the government's stated priorities in its 15th Five Year Plan, which emphasizes AI and robotics for industrial applications. This suggests a broader national push to integrate AI technologies into the real economy.
Unlike some competitors focused on creating highly theatrical robots, PsiBot emphasizes the practical aspects of embodied AI, particularly the collection of high-quality data necessary for real-world reliability. The company aims to build scalable, repeatable learning pipelines by gathering data through simulation, teleoperation, and field use, addressing the challenge of data acquisition for robots, which lacks the vast digital resources available to large language models.