50 billion yuan (about 6 billion euros) in investments will give DeepSeek a valuation of approximately 50 billion euros. This is the first funding round for the Chinese AI lab, which surprised the world early last year with its capable DeepSeek-R1 model.
The largest investors outside the company itself will be Tencent and CATL, alongside NetEase and JD.com. This is reported by various media outlets, including Reuters and Bloomberg. The plans were still confidential and are subject to change.
U-turn
It is notable that DeepSeek is seeking outside funding. Until now, founder Lian Wenfeng has financed the AI startup through his quantitative hedge fund, High-Flyer. In this way, the Chinese company managed to build several impressive AI models. DeepSeek-R1 was released in early 2025, a “reasoning” LLM in the vein of OpenAI-o1, which was state-of-the-art in AI at the time. It delivered performance that came somewhat close to the best AI models elsewhere. However, this LLM was significantly smaller and open-source.
Since then, Anthropic, Google, and OpenAI have regained their lead with proprietary models. Anthropic and OpenAI, in particular, expressed concerns about the danger of “distillation,” a process in which a smaller AI model “learns” from a larger LLM by training on its responses.
DeepSeek needs to regroup
Stunning results from R1 offered no guarantees for the future. DeepSeek-R2 has still not been released, while DeepSeek-V4 was not as competitive against current rivals as R-1 was against earlier LLMs.
Nevertheless, research by the team behind DeepSeek is yielding breakthroughs that suggest significant improvements. For instance, it found ways to build “highways” within LLMs, something that, combined with Google’s TurboQuant, appears to be shaping up as a response to the ever-increasing requirements for running AI.
Nvidia or Huawei?
A Washington restriction on the sale of Nvidia chips to China has since been eased, even though the U.S. is attempting to block shipments to Chinese companies outside of China. Without Nvidia, DeepSeek has a domestic hardware alternative in Huawei. Beijing intends to end its dependence on the West, but the current performance level of non-Western AI processors is not yet competitive.
This means that DeepSeek must once again build models with fewer resources than its AI competitors. That is a territory in which the Chinese AI lab should now be comfortable.