To date, DeepSeek has made public investments in just two companies. One is Unitree, a well-established enterprise, and the other is Winverse AI, an emerging startup specializing in embodied AI and world modeling, where DeepSeek is set to take the lead in investment. Winverse AI's founder, Hu Han, boasts an illustrious background. After graduating from Tsinghua University, he went on to work at Baidu's Institute of Deep Learning and Microsoft Research Asia. As a key researcher, he contributed significantly to several groundbreaking achievements in the field of computer vision, including the development of Swin Transformer. Subsequently, he joined Tencent, where he led pre-research efforts on multimodal models, large language models, and world models associated with Hunyuan. In July 2026, he left Tencent to embark on his entrepreneurial journey.

DeepSeek's decision to invest in Hu Han's venture stems primarily from the trust forged through Hu Han's former subordinate at Microsoft Research Asia, who is now a core member of DeepSeek. This relationship has facilitated efficient information exchange, keeping communication costs low. Moreover, Hu Han's team brings deep expertise in visual multimodality and large-scale model training, which perfectly complements DeepSeek's existing strengths.

By opting to invest rather than directly venture into the embodied AI field, DeepSeek aims to sidestep the complexities of hardware and supply chain management, areas where it lacks familiarity. Through a strategic division of labor—where Unitree supplies physical robot platforms, Winverse AI develops embodied AI models, and DeepSeek provides computational power and foundational models—DeepSeek can rapidly enhance the capacity of its language models to interact with the physical world. This approach positions DeepSeek to effectively counter competitive pressures from OpenAI in the embodied AI sector.