DeepSeek Prioritizes AGI and Open-Source with Landmark 70 Billion Yuan Funding Round

DeepSeek, the Chinese artificial intelligence firm that gained prominence in 2025 with its cost-effective R1 large language model, is nearing the completion of a landmark funding round. The company is in the final discussions for a deal that could inject 70 billion yuan (roughly $10 billion USD) and potentially value the AI lab at approximately $45 billion before the investment. This substantial financing represents a pivotal moment for DeepSeek, as it marks the first time the company will accept external capital, having previously been entirely funded by High-Flyer Quant, the quantitative trading firm founded by DeepSeek's CEO, Liang Wenfeng. During discussions with prospective investors, DeepSeek's management has emphasized a clear strategic direction: prioritizing groundbreaking AI research, particularly the pursuit of artificial general intelligence (AGI), over short-term commercialization goals. Founder Liang Wenfeng has reportedly pledged to continue developing and releasing open-source AI models, a stance that sets DeepSeek apart in an industry increasingly focused on proprietary solutions and immediate revenue generation. This commitment underscores DeepSeek's belief that it can scale its operations without compromising its core principles of advanced research and open collaboration. The list of potential investors in this high-profile round includes significant players such as China's National Artificial Intelligence Industry Investment Fund, a state-backed vehicle designed to propel strategic AI projects. Other major Chinese tech giants like Tencent Holdings, IDG Capital, and Monolith Capital are also reportedly close to finalizing their participation. The state-backed fund alone is considering an investment of around 10 billion yuan. If finalized, this fundraising could set a new record for a first-time financing round by a Chinese tech startup, highlighting Beijing's strategic interest in fostering domestic AI leadership. DeepSeek's technical prowess, including its DeepSeek Sparse Attention system for handling long inputs efficiently and its models reportedly matching frontier systems like Google's Gemini-3.0-Pro on some reasoning tasks, has contributed to strong investor interest. The company has also been noted for optimizing its latest V4 model for inference on Huawei's Ascend 950PR chips, showcasing a trend of coordinated software and hardware development. Despite its "open" model approach, some of DeepSeek's models utilize custom licenses with usage limitations, which helps build a developer community while maintaining control over the technology. This massive capital infusion is expected to accelerate DeepSeek's pace of revenue generation and commercialization, with plans to increase the release frequency of its AI models.