DeepSeek has recently completed its first external financing round since the company’s founding, raising over RMB 50 billion (approximately US$7.4 billion) and valuing the company at roughly US$54.3 billion. This massive influx of capital, which included investments from Tencent (RMB 10 billion), CATL (RMB 5 billion), NetEase, JD.com, IDG Capital, and the National Artificial Intelligence Industry Investment Fund, marks a significant shift for a company whose founder, Liang Wenfeng, previously maintained a strict “no fundraising, no IPO” policy.
A leaked four-hour investor meeting transcript published by The China Academy now reveals what Liang told investors about the company’s open-source commitment and its roadmap to AGI. The fundraising followed a period of intense public scrutiny, with DeepSeek subsequently pausing a second funding round after Liang’s viral investor remarks circulated widely.
Despite this formal entry into the capital markets, a recently leaked transcript of a four-hour investor meeting reveals that Liang remains steadfast in his commitment to open-source development and his long-term vision for Artificial General Intelligence (AGI). The transcript, published by The China Academy and sourced from Tencent Technology’s edited version of Liang’s remarks, provides the most detailed public account yet of DeepSeek’s philosophy and strategic direction. It arrives at a moment when the global AI community is intensely debating whether open-source AI is a public good or a national security threat, a debate in which DeepSeek sits at the very center.
The Open-Source Commitment as Strategic Restraint
During the meeting, Liang addressed concerns that external funding might force DeepSeek to abandon its open-source ethos in favor of aggressive commercialization. He firmly dispelled these notions with a statement that has since reverberated across the global AI community: “We will continue to open-source our models, and I believe that even our most advanced models will eventually be open-sourced.”
Liang framed this restraint as a deliberate strategy rather than a commercial burden. “Restraint is a strategy. Sometimes you give up certain things in order to gain something much greater,” he explained. “Open source is the same. You can see it either as a burden we take on or as value that we’re choosing to give away.”
By remaining open-source, DeepSeek aims to empower the broader developer community and accelerate the overall progress of the AI industry, which Liang believes will ultimately increase the company’s probability of achieving AGI. He noted that this approach has not cost the company anything: “We haven’t lost out because we’re open source, because we’ve acted in good faith, or because we’ve helped others. On the contrary, it may even have worked in our favor.”
This philosophy stands in stark contrast to the approach of U.S. frontier labs, which have increasingly moved toward closed, proprietary models. As Liang previously stated, AGI is the only goal and commercialization can wait, a position that has earned DeepSeek enormous goodwill in the global developer community while simultaneously alarming U.S. national security officials.
The Roadmap to AGI: Continuous Learning as the Missing Link
The transcript also provided a detailed look at DeepSeek’s technological roadmap. Liang identified “continuous learning” as the critical missing link between current AI capabilities and true AGI. While today’s models excel when provided with complete context and instructions, they lack the ability to learn continuously over long periods, as a new employee would over their first months at a company.
“The development of AI can be understood as climbing a staircase,” Liang explained. “Last year’s step was Chain of Thought. This year’s step is the Agent. After Agents, the next problem we need to solve is continuous learning.” He argued that once continuous learning is achieved, AI will be capable of developing its own next version and conducting its own research, a process he described as the “singularity,” though he cautioned it will be a gradual transition rather than an abrupt leap.
Liang was also notably candid about what DeepSeek will not pursue. He explicitly stated that the company will not enter video generation or world models, viewing them as commercially viable but not fundamentally related to the core problem of intelligence. “We focus only on the main path toward AGI,” he said. This focused, restrained approach, combined with the company’s open-source commitment, positions DeepSeek as a uniquely principled actor in a field increasingly dominated by commercial imperatives and geopolitical pressures.
Liang also addressed the question of competition directly. He argued that DeepSeek’s greatest risk is not being outcompeted by other Chinese labs or by U.S. frontier labs, but rather the possibility of the company itself losing its own focus, a concern he has repeatedly expressed in internal discussions about the company’s disciplined refusal to chase adjacent markets.
This self-awareness is unusual in an industry where companies routinely expand into adjacent markets at the first sign of success. Whether DeepSeek can maintain this discipline as its valuation grows and investor expectations evolve will be one of the defining questions of the Chinese AI industry in the years ahead, a question further examined by Storyboard18 in its coverage of the leaked transcript.