DeepSeek, the Hangzhou-based AI startup, is looking to raise at least $300 million at a valuation of over $10 billion, reports Reuters. This is DeepSeek’s first external funding round since its founding in 2023 by Liang Wenfeng.
For two years, DeepSeek turned down all outside investment, even from top venture capital firms and major tech companies.
Reasoning models, agent systems, and the infrastructure require a lot of money, more than even a successful hedge fund can provide forever. High-Flyer, DeepSeek’s parent company, reportedly posted a 56.6% return in 2025, which has funded DeepSeek’s research so far, says Bloomberg.
The technology that emerged from this challenged people’s assumptions about the costs of AI development. DeepSeek’s R1 reasoning model, released in January 2025, matched OpenAI’s top model but reportedly cost only $6 million to train, which had a big impact on the market.
The model uses a mixture-of-experts design and reinforcement learning, rather than the more expensive supervised fine-tuning used by Western labs. It usually activates only 37 billion of its 671 billion parameters. Running the model costs about $0.55 per million input tokens, roughly 96% less than OpenAI’s version. The models are fully open-source, released under the MIT license, and can be used by anyone.
Since R1 was released, OpenAI, Anthropic, Google, and Meta have accelerated their own reasoning model projects in response. The same labs that DeepSeek disrupted are now using its efficiency methods and investing more in infrastructure and talent.
The funding round will show if DeepSeek can take outside money without losing the research-focused culture that produced R1. Liang has previously said that China’s AI industry needs confidence as much as capital, and now he is looking for both from the market.
