DeepSeek’s valuation in its first venture funding round reportedly climbed from $20 billion to $45 billion within a matter of weeks, a rise of 125%. The lead investor is said to be China’s Integrated Circuit Industry Investment Fund, widely known as the “Big Fund,” while Tencent and Alibaba have also been reported to be in talks to participate.
The company did not appear to be under pressure to raise outside capital. According to the report, founder Liang Wenfeng already had steady cash flow from quantitative hedge fund High-Flyer, and DeepSeek had not previously taken external funding since its founding. That makes this round look less like a cash-raising necessity and more like a valuation-setting event.
An external price tag could support employee equity plans
The report points to talent retention as a likely reason for opening the round. With core researchers vulnerable to poaching by rivals offering higher pay, equity becomes a more practical retention tool only when the shares carry a market-recognized value. For an AI company that has long stayed away from outside investors, that pricing reference matters.
Liang may not have needed fresh cash. He may have needed a number the market would accept, one that turns employee stock into something concrete rather than an internal estimate.
The “Big Fund” gives the round a strategic dimension
The lead investor is not an ordinary financial backer. China’s “Big Fund” was set up with central government support to back the country’s semiconductor and technology industries, rather than to chase short-term returns. Its presence in DeepSeek’s shareholder roster suggests the financing has moved beyond a standard commercial deal.
The Financial Times also reported that Tencent and Alibaba were in discussions to join the round. A state-backed fund and two of China’s biggest tech companies appearing in the same AI financing process is a rare alignment. The report also cited a Bloomberg story from April saying China’s State Council had been considering a ban on AI unicorns accepting U.S. capital; if adopted, that would narrow DeepSeek’s funding options to domestic channels.
Lower compute costs and open-source distribution drew attention
When DeepSeek launched its language model in early 2025, the surprise was not limited to model performance. The report said the training compute and cost required were only a fraction of those associated with frontier U.S. models from OpenAI or Anthropic.
DeepSeek also released open-weight versions on Hugging Face for developers around the world to use at no cost. That decision helped the technology spread quickly and brought the company notable attention inside U.S. AI circles. As more developers build on the same architecture, the model’s position in the broader ecosystem becomes harder to displace.