Key facts
- Global investors are using perpetual futures, a type of derivative, to gain exposure to Chinese technology stocks.
- Unitree, a Chinese humanoid robot maker, and memory chip producer CXMT saw significant investor interest before their IPOs.
- Unitree's IPO on the Shanghai STAR Market recorded an oversubscription multiple of 8,288 times.
- Unitree's IPO was priced at ¥150.80 per share, valuing the company at approximately $9 billion.
- AI firm DeepSeek invested $20.9 million in Unitree's IPO, with both companies planning to collaborate on AI humanoid models.
Global investors are increasingly utilizing highly leveraged derivatives known as perpetual futures to gain exposure to Chinese technology stocks, particularly those involved in artificial intelligence. This trend is highlighted by the significant investor interest in companies like Unitree, a leading Chinese humanoid robot developer, and memory chip producer CXMT, ahead of their recent initial public offerings.
Unitree Robotics, officially registered as Yushu Technology, experienced an overwhelming response to its IPO on the Shanghai STAR Market. The company opened public subscriptions with an unprecedented oversubscription multiple of 8,288 times, indicating massive demand. Unitree priced its IPO at ¥150.80 (approximately $22.35) per share, resulting in a valuation of about ¥61 billion (approximately $9 billion). The sheer volume of applications led to ¥800 billion (approximately $118.6 billion) in subscriber funds being frozen.
Adding to the investor excitement, AI firm DeepSeek invested ¥140.8 million (approximately $20.9 million) in Unitree's IPO through a strategic placement, receiving shares under a three-year lock-up agreement. The collaboration between Unitree and DeepSeek aims to develop joint AI humanoid models, leveraging DeepSeek's AI expertise and Unitree's strengths in mechanical engineering and embodied intelligence. Unitree reported delivering over 5,500 humanoid robots globally in 2025, capturing a 32.4% market share.
However, a joint filing between Unitree and DeepSeek also publicly acknowledged a significant industry challenge: the Vision-Language-Action problem. This refers to the difficulty in creating AI systems that can perceive unfamiliar environments, understand natural language instructions, and translate that understanding into adaptive physical actions, rather than pre-scripted movements. Researchers have noted the brittleness of current robot systems lacking joint perception, making intelligent autonomous behavior a distant goal.