Two months after DeepSeek invested RMB 141 million in Unitree, founder Liang Wenfeng is reported to be leading the seed round of Winverse AI, an embodied-AI startup only months old with no product shown yet. According to Bloomberg, DeepSeek is leading a seed round of tens of millions of dollars at a post-money valuation of about US$400 million, with Meituan’s Longzhu Capital committed to follow on.
The more intriguing thread is the people. Winverse founder Hu Han was formerly head of Tencent’s Hunyuan multimodal model and left Tencent in July. Tencent is also a potential investor in DeepSeek’s new financing round. The talent and capital networks of the big platforms are interweaving in subtle ways.
In July, Liang laid out DeepSeek’s AGI roadmap across six rungs: language model, chain of thought, Agent, continual learning, self-iteration, and only last, embodied intelligence. His logic was to train the brain before growing the limbs. He even poured cold water on world models at the time. Less than half a year later, he has apparently come around.
From Unitree to Winverse, Liang’s embodied-AI map is clarifying into a “jump-step” strategy. Rather than fight in the crowded hardware race, he is placing early bets on the brain layer.
More from the round
For Europe’s robotics scene, the signal is that China’s leading AI capital is crowding the model-and-brain layer of embodied intelligence, not only the metal.
Editor’s note: This is an adapted translation of the original OFweek report. It has been trimmed and restructured for readability for an international business audience. The full original (in Chinese) is at https://robot.ofweek.com/2026-10/ART-8321205-12003-30707783.html.