DeepSeek founder Liang Wenfeng contacts embodied intelligence entrepreneurs, reportedly weighing entry

leaderobot.com reported on Oct. 13, 2026, that DeepSeek founder Liang Wenfeng has been in contact with entrepreneurs in the embodied intelligence sector. The report said Liang may be considering DeepSeek's entry into the field. Options reportedly include forming an in-house team or investing in related companies.

The embodied intelligence sector is seeing increasing specialization not only in hardware development but also in model training and data collection. Reports that DeepSeek may enter the field have drawn attention to whether model companies could play a role in the robotics industry's structure. DeepSeek had not responded to the report about entering embodied intelligence as of the original article's publication. The article also reported that DeepSeek made a strategic investment of 141 million yuan in Unitree Robotics in August.

The new element in this report is that DeepSeek is not treating embodied intelligence merely as a peripheral topic: the company's founder is said to be personally involved. The original article reported that possible ways of entering the field are still under consideration, while also saying DeepSeek made a strategic investment in Unitree Robotics in August and signed a cooperation memorandum. For now, it is more reasonable to view this as an early exploration of possibilities than as a settled expansion plan. Related stories from Physical AI News include [Lumi Robotics and ModelBest form strategic partnership on embodied data collection, model training and scenario applications](/news/shika-mei-kiki-nin-to-menpeki-chinou-ga-senryaku-teikei-gu-m-d899b5) and Geek+ to invest 50 million yuan in Beijing Shenyi. They illustrate how hardware, models, data and capital are developing along distinct tracks in embodied intelligence. This report suggests that DeepSeek, a company closer to the upstream model layer, may be exploring how to get involved, whether through external investment or in-house development. However, the information is based on reports; see the original article for details. The industry's competitive landscape is increasingly divided by who handles not only robot performance but also training data, model training and connections to the real world. If DeepSeek enters the field, a key question will be how it applies its large language model development expertise to the physical world. The report does not specify how it might do so. Points to watch include whether it establishes an internal organization, makes investments, and decides what to develop in-house versus with external partners.