Woofun AI reports that DeepSeek has temporarily suspended its second round of financing, a decision directly attributed to founder Liang Wenfeng’s frustration over the leakage of content from a closed-door meeting, as disclosed by Bloomberg. The company has verbally informed potential investors that investment agreements will not be signed as previously expected, marking a significant disruption in its capital-raising timeline. This pause stems from the unauthorized circulation of sensitive internal discussions, which have now become public knowledge, forcing a reassessment of the company’s external communications and investor relations strategy.

The leaked material originates from a session held during the first round of financing, where Liang Wenfeng addressed potential investors for nearly four hours. The discussion covered critical operational aspects, including commercialization strategies, the model roadmap, computing power reserves, the integration of domestic chips, and projections for future revenue. Two months after the event, a transcription spanning tens of thousands of words began circulating online, exposing details that were intended to remain confidential. The scope of the disclosure was extensive, touching on both strategic vision and granular operational metrics, which had not been previously shared with the broader market.

Despite the current suspension, Bloomberg indicates that the financing round has not been entirely canceled, leaving open the possibility for negotiations to resume in the future.

Concurrently, DeepSeek’s preparations for a domestic IPO remain on track, with the earliest possible submission of the listing application targeted for later this year. Reuters reported on July 15 that the company aimed to raise funds at a valuation of approximately 500 billion yuan, with the potential size of the round reaching up to 50 billion yuan. This capital influx is viewed as essential for supporting computing power expansion and establishing a higher valuation benchmark ahead of the public listing, which is internally targeted for completion by 2027.

The financial context for this pause is rooted in the company’s recent capital activity. In June, DeepSeek completed its first external financing since its establishment, with multiple media outlets disclosing the amount at around 50 billion yuan. This round resulted in a post-investment valuation estimated between 3.5 trillion and 4 trillion yuan, a figure that solidified its position as one of the largest single-round financings in the Chinese AI industry. Shortly after the conclusion of this first round in mid-July, the company engaged with investors to plan the second round, aiming to secure additional resources before the IPO. The rapid succession of these rounds highlights the intense capital demands of large-scale AI development.

The breach of confidentiality is particularly egregious given the explicit instructions provided during the meeting. Held on May 20, the session lasted approximately 3 hours and 44 minutes, with Liang Wenfeng introducing the business philosophy, open-source logic, and AGI roadmap in the first half, followed by an investor Q&A. Before the Q&A began, the host specifically reminded attendees: 'Liang Wenfeng has shared a lot of sensitive information, so please do not disclose any numbers or details, including card issuance numbers. Also, please do not screen record or share externally.' From July 22 to 23, this material began spreading in domestic tech and investment circles, eventually being widely reposted by media and overseas communities, despite these clear prohibitions.

Liang Wenfeng continues to prioritize AGI as the company’s ultimate goal, viewing consumer-facing products and B2B business as transitional steps rather than end-states. He describes AI development as a series of ascending steps: starting with language models, progressing to the cognitive chain, and then to Agents, with continual learning identified as the most critical issue. DeepSeek aims for its next-generation models to assist researchers in developing subsequent iterations, focusing primarily on General Agents and Coding Agents. Vertical agents in sectors such as finance and healthcare are currently assigned lower priority, reflecting a strategic concentration on foundational capabilities rather than immediate commercial applications in specific industries.

Compute power and chip availability represent the most sensitive aspects of the leaked discussion. Liang Wenfeng estimated that DeepSeek may have used only one-twentieth of the compute power of its American counterparts in the past, resulting in an overall model lag of about 12 to 18 months. The strategic goal is to reduce this lag to 3 to 6 months and achieve breakthroughs in specific directions. The transcript mentions the scale of chips required for future models, the number of chips provided by Huawei, and the performance conversion between Huawei cards and NVIDIA cards. These figures were widely cited in subsequent discussions, although no public procurement contracts or formal confirmations from DeepSeek or Huawei have been released to verify these specific hardware allocations.

Woofun AI data shows that the company’s ability to pause financing without immediate operational impact is underpinned by a strong cash position and lack of acute cash pressure. Prior to this year, DeepSeek had not conducted external equity financing for an extended period, emerging in 2023 from Phantasm Quant’s AI research system. Early R&D investment was primarily funded by Phantasm’s operating income and R&D budget. Financing rumors began circulating in April, with the valuation and scale rising rapidly within weeks. The first round landed in June, providing a true valuation for employee stock options and significantly mitigating the risk of talent outflow, which Liang Wenfeng cited as a non-negotiable area for team stability.

The equity structure following the first round reveals a concentrated control mechanism. Liang Wenfeng personally contributed about 20 billion yuan, making him the largest individual investor, while Tencent invested about 10 billion yuan and CATL about 5 billion yuan. Other participants included NetEase, JD.com, IDG, and the National Artificial Intelligence Industry Investment Fund. Ningbo Cheng'en Enterprise Management Consulting Partnership holds about 60.19% of the operational entity, Liang Wenfeng directly holds about 31.

01%, and Hangzhou ChingLee Platform, housing external investors, holds about 8.52%. The National Artificial Intelligence Industry Investment Fund holds about 0.28% directly. Domestic media estimates that Liang Wenfeng’s direct and indirect control rights amount to about 84.29%, ensuring that main decision-making power remains with him. Most commercial investors enter through limited partnership platforms, facing a five-year lock-up period and lacking voting rights in the operating company.

Phantasm Quant serves as another pillar of financial confidence, with an average return rate of about 56.6% in 2025, ranking second among domestic large-scale quantitative institutions with assets under management exceeding 100 billion yuan. Its current asset management scale exceeds 700 billion yuan, providing legitimate R&D investment through management fees and performance rewards. DeepSeek’s current flagship model is the V4 preview version released in April, with the official version originally scheduled for mid-July.

The V4-Pro variant features 1.6 trillion parameters with 490 billion activated per token, while V4-Flash has 284 billion total parameters with 13 billion activated. Both support a context of 1 million tokens. As competition shifts toward agents and continual learning, the need for more chips and data centers drives the urgency for the second round, which remains paused as DeepSeek has not publicly responded to the suspension or the authenticity of the leaked transcript, as of the time of publication.