Financing Pace Accelerates Sharply
Just over a month after its previous round, DeepSeek has reportedly launched a second funding round. According to the Financial Times, it is in talks with new investors at a pre-money valuation of about $71 billion, up 37% from the $52 billion post-money valuation after the first round.
The first round raised over RMB 50 billion, with founder Liang Wenfeng contributing RMB 20 billion personally, and participants including Tencent, CATL, and the National AI Industry Fund. The rapid launch of the new round signals a complete shift in DeepSeek’s capital strategy.
Computing Bottleneck Forces Strategic Pivot
Liang previously expressed a desire to focus on AGI research and avoid capital markets. But this year’s AI competition has fundamentally changed, it is no longer just about model intelligence, but also the scale of computing power behind it.
Agent systems demand continuous reasoning, long‑context processing consumes massive VRAM, and nearly all technology trends are accelerating compute consumption. To keep up infrastructure‑wise, DeepSeek cannot rely solely on internal funds; it needs external capital to secure future computing supply.
Clear Use of Funds
The new funds will go to three areas: building its own data centers, purchasing more AI chips, and developing next‑generation AI systems like Agents. Job postings confirm this expansion — the company has listed "IDC Design Planning Engineer" roles to build GW‑level data centers covering site selection, design, and construction.
The Harness unit is also hiring researchers, engineers, and product managers, while Infra teams are expanding for training frameworks, operators, and supercomputing clusters. More notably, Reuters reports that DeepSeek has started a secret project to develop its own AI inference chips and has approached chip design firms, foundries, and memory suppliers.
IPO Plans Emerge
Another key signal is that DeepSeek is preparing for a 2027 IPO, with the earliest filing possible by year‑end, though timing depends on market conditions and performance.
After the first round, Liang’s personal net worth rose from $16.7 billion to about $36 billion, but he still invests first in each round, mainly to reassure other investors.
What This Means for Regular Users
The accelerated fundraising and IPO preparations will not directly affect the C‑end user experience in the short term. But expanded computing power will eventually translate into smarter conversations, longer context, and more complex tasks.
If the IPO succeeds, DeepSeek will become a publicly tradeable AI asset for ordinary investors. Before that, however, it must prove it can sustain heavy spending while maintaining technological pressure on OpenAI and Anthropic.