DeepSeek, the Chinese AI company best known for completely disrupting the AI market, has engaged in preliminary talks with investors for a new funding round after completing its first external financing. The company seeks a pre-money valuation of US$ 71 billion, a 37 percent increase from the first US$ 52 billion valuation funding round which raised US$ 7 billion in late May 2026. The new round is expected to raise 10 billion yuan (US$ 1.4 billion), though the final amount is dependent on the investor demand.
In the previous round, DeepSeek founder and CEO, Liang Wenfeng was the largest investor, contributing US$ 3 billion. Other investors included Tencent, Chinese multinational technology conglomerate, CATL, a battery maker and technology company, JINGDONG (JD), a supply-chain technology and service provider, NetEase, a technology company, several venture capital firms, and a state-backed national AI investment fund.
According to the Financial Times, DeepSeek’s management has informed potential investors that the company intends to prioritize long-term AI research and development on open-source models and continue AI general intelligence research instead of near-term commercialization.
DeepSeek has also started preparations for an initial public offering (IPO) in which the company reportedly began discussions with accounting firms, investment banks and a mainland China listing. These IPO documents could be filed in 2026 with a potential public debut in 2027, although the timeline remains subject to market conditions and company progress.
The AI start up company received international attention after releasing its open-source R1 reasoning model which demonstrated performance comparable to Western systems such as Open AI and Anthropic while using fewer computing resources and caused major disruption in the AI industry and intensified competition.
The fundraising reflects DeepSeek’s growing capital to expand its computing infrastructure, data center construction, and AI chip acquisition to develop AI models that require more computing power.