DeepSeek founder Liang Wenfeng informed potential investors that the company prioritizes Artificial General Intelligence (AGI) development over rapid product monetization. At the same time, he confirmed the intention to continue releasing open-source models while maintaining a research model for the laboratory. This is reported by Ixbt.com reports .
According to Bloomberg, this is the first external funding round in the company's history. DeepSeek plans to raise at least $300 million, with the target valuation for the round potentially reaching approximately 70 billion yuan (nearly $10 billion). Some sources estimate that the total capital raised could grow to $7 billion in the long term.
Until now, DeepSeek was fully funded by High-Flyer Quant, an investment firm founded by Liang. Avoiding external investors was previously seen as a way to maintain independence in product and research agendas. The current round indicates a change in scale: the costs of training modern models have exceeded even the capabilities of a profitable hedge fund.
In its strategic positioning, the company continues to rely on open models and research. In April, DeepSeek introduced the trillion-parameter V4-Pro and V4-Flash model families. These systems are optimized for hardware platforms such as Huawei Ascend, Cambricon, and NVIDIA. Previously, the R1 model released in early 2025 caused a major market stir, forcing a reassessment of expectations regarding the costs of training advanced models.
The main parameters of the deal—the composition of investors and the final valuation timeline—have not yet been disclosed. Sources note that the scale of the fundraising will inevitably attract the attention of Chinese regulators, as they are tightening control over developers of foundational models. The company marks a new phase of transitioning to institutional capital while maintaining its research and open-source strategy.
Read “Zamin” on Telegram!
