DeepSeek, the AI startup whose low-cost model tanked Nvidia’s stock over a year ago, is raising funding at a valuation of as much as $50 billion, according to The Wall Street Journal.
One interested investor is reportedly China’s National Artificial Intelligence Industry Investment Fund, a one-year-old fund with $8.8 billion to invest and backed by the Chinese government.
DeepSeek aims to raise several billion dollars in funding. The capital would be used for R&D and expand computing infrastructure. The Journal also said DeepSeek wants to prove its worth and retain top employees through valuable equity.
China has been treating DeepSeek as a national AI champion to hedge against U.S. leadership in the field, according to the paper.
DeepSeek’s low-cost and high-performing V3 model tanked Nvidia’s stock by 17% and pressured other U.S. tech stocks on Jan. 27, 2025, on fears that it will attract sales away from American tech companies. Later, DeepSeek also released its R1 family of reasoning models.
DeepSeek’s latest model is V4, which was trained on Nvidia’s chips. However, it is working with Huawei and other Chinese chipmakers as it pivots away from dependence on U.S. companies.
The startup’s main models are open-source, which means they are free to modify and use depending on the license. DeepSeek earns revenue from clients using its API to access the models.
The Journal said DeepSeek had stayed away from external funding to keep its independence, relying instead on founder Liang Wenfeng’s hedge fund earnings. But now it is aligning itself with national interests to fight U.S. dominance.
In April, China scuttled the acquisition of AI startup Manus by Meta. While Manus is based in Singapore, it has Chinese roots.