GuShiio.com 
GuShiio.comAGI reported on February 7 thatMarket rumors are rumored that DeepSeek is considering a new round of financing at a valuation of US$10 billion, and Alibaba has plans to invest US$1 billion to subscribe for DeepSeek’s equity. Currently, the two teams are communicating specific implementation details, and Alibaba Cloud will be the first reasoning computing power choice.
Affected by the news, Alimei shares (NYSE: BABA) surged more than 6%.
As of press time, Ali people denied rumors of investing in DeepSeek.Yan Qiao, vice president of Alibaba Group, said that we are both companies in Hangzhou, China, and we applaud DeepSeek, but the information that Alibaba has invested in DeepSeek is false news.
GuShiio.comAGI further learned from sources that DeepSeek is currently valued at around US$8 billion. The news was initially spread among investment circles and quantitative groups, and many investment institutions were very interested in it.
According to Tencent Technology, Zhu Xiaohu, managing partner of Jinshajiang Venture Capital, said earlier that once DeepSeek opens up financing, he will definitely invest. ldquo; I will definitely vote! I will definitely vote! This price is no longer important, the key is to participate in it. It is very meaningful to truly witness the emergence of human AGI and witness the emergence of human AI consciousness.& rdquo;
Zhu Xiaohu emphasized that he believes that DeepSeek should still be open to financing, because moving forward requires burning money, and the current main resource demand is still computing power cards.
It is reported that DeepSeek (Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co., Ltd.) was established in 2023 and is headquartered in Hangzhou. It was founded by Chinese hedge fund Magic Square. The founder and CEO of DeepSeek is Liang Wenfeng.
On October 28, 2023, DeepSeek released the first large-scale model for in-depth exploration, DeepSeek-Coder, and released DeepSeek-LLM on November 29. As of December 13, 2024, DeepSeek will release DeepSeek-VL2, an expert hybrid visual language model for advanced multimodal understanding. On the 26th of the same month, DeepSeek released and open-source DeepSeek-V3, which attracted attention.
DeepSeek said that the training system for the large model is based on 2048 blocksNvidiaIt was completed in 55 days on a GPU cluster, and the training cost US$5.576 million. The evaluation results of DeepSeek-V3 surpassed open source models such as LLaMA 3.1- 405B (Meta Self-Developed Large Model) and could compete with closed-source models such as GPT-4o.
On January 20, 2025, DeepSeek released and open-source the inference model DeepSeek-R1 model. The cost was lower than expected, but the model’s performance in mathematics, code, natural language reasoning and other tasks is comparable to the official version of OpenAI o1.As of January 27 this year, DeepSeek’s smart assistant will be available in the USAppleOvertaking ChatGPT on the App Store download list and reaching the top of the App Store free app list.
According to reports, DeepSeek has fewer than 140 employees. Liang Wenfeng once said that the main way to retain young talents is to pay high salaries and manage enough computing power.
“What we see is that China AI cannot always be in a following position. We often say that there is a year or two gap between China’s AI and the United States, but the real gap is the difference between originality and imitation. If this does not change, China will always be a follower, so some explorations cannot escape.& rdquo; Liang Wenfeng said.
