DeepSeek recently concluded its first external equity financing round, securing approximately 51 billion yuan in funding, with the company’s valuation now estimated between 325 billion and 350.9 billion yuan. During a four-hour dialogue with investors, founder Liang Wenfeng articulated the company’s strategic vision: prioritizing Artificial General Intelligence (AGI) as its central mission, adhering to open-source principles, pursuing only reasonable profits, and positioning itself as a technology-driven organization committed to long-term AGI investment rather than a purely profit-oriented entity focused solely on revenue expansion.

The company exercises rigorous scrutiny in selecting financing partners, emphasizing alignment of long-term interests with investors. Notably, Liang Wenfeng himself emerged as the largest individual contributor in this funding round. He emphasized that DeepSeek fosters team cohesion through a shared vision, prioritizing core interests such as maintaining stability within the team, enabling agile management, minimizing overtime for employees, and allocating half of their working hours to independent exploration and innovation.

Liang Wenfeng argued that open-source principles and commercialization are not mutually exclusive; in fact, the most robust models are likely to remain open-source. API pricing is structured to achieve a reasonable profit margin, with the goal of recovering server costs within 10 months. He also expressed hope that more companies would leverage their models to launch new ventures.

DeepSeek remains steadfast in its focus on AGI, refraining from diverting resources toward short-term commercial gains. The company believes that following the development of AI agents, large language models must overcome the bottleneck of continuous learning to ultimately progress toward embodied intelligence—a form of AI that interacts with the physical world. Currently, the company’s most significant challenge lies in computing power constraints, though it asserts that the Scaling Law (the principle that model performance improves predictably with increased scale) has not yet reached its limits. Additionally, DeepSeek is optimistic about the substitution potential of domestic AI chips, anticipating that Chinese foundational model companies will eventually converge in terms of technological advancement and market positioning.