DeepSeek is planning to kick off its first round of funding with a valuation of at least $10 billion to address the core talent drain and the V4 development delay.

According to Sentinel Beating monitoring, the AI startup DeepSeek is in talks for its first external funding round, planning to raise at least $3 billion at a valuation of at least $100 billion. This marks a significant shift in fundraising strategy for the startup, which was previously fully supported by the hedge fund Quantum Quantitative (High-Flyer) and once held onto technological idealism and rejected external capital.

This adjustment comes as DeepSeek faces the dual challenges of technological iteration pressure and poaching from competitors. While the R1 model had achieved high performance at an extremely low cost that shook the global market, the release schedule for its next-generation flagship model, V4, has been repeatedly delayed. The delays are attributed to engineering challenges, as well as DeepSeek engineers attempting to natively adapt V4 to Huawei chips. Since the previous models were mainly developed based on Nvidia chips, this switch in underlying architecture has significantly extended the development cycle.

Meanwhile, key team members are experiencing attrition. Lofly, a core contributor to the V3 model, has joined Xiaomi to lead its AI department, while another key researcher, Guodaya, has been poached by ByteDance with a higher salary.