Now It's Just Awaiting Liang Wenfeng to Ring the Bell

The IPO steps are drawing closer.

According to Bloomberg, DeepSeek has begun preparations for an IPO, planning to list domestically, with a possibility of submitting an application as early as this year. As the discussions are still confidential, DeepSeek has not responded to requests for comment.

This is not unexpected. Recall on June 17th, the Shanghai Stock Exchange announced support for high-quality artificial intelligence large model enterprises that have not yet generated a certain scale of revenue to issue and list on the Sci-Tech Innovation Board. Subsequently, Zhipu and MiniMax have both accelerated their efforts to advance their IPOs on the Sci-Tech Innovation Board.

Around the same time, DeepSeek was reported to have begun preliminary discussions with new investors this week about initiating a new round of financing. Foreign media disclosed a pre-money valuation of approximately $71 billion (about 480 billion RMB).

"Investors all want to swipe their cards for Liang Wenfeng, but they can't find the POS machine." A partner at a well-established VC institution recently revealed to us that in July, they had an appointment to meet Liang Wenfeng in Hangzhou—the next round of players wanting a seat at the table are already on their way.

DeepSeek's Financing Secrets, Valuation Comparable to CXMT

"Many state-owned institutions and insurance capital couldn't get in." A person familiar with the matter previously revealed.

Having an allocation doesn't guarantee entry. A partner at a well-known institution recalled that during the first round of financing, he was informed that DeepSeek had a 20 billion RMB allocation available, but they needed to prepare proof of funding capability for 20 billion within a day. "The timeline was too tight, and we didn't participate."

Ultimately, we saw that the National Artificial Intelligence Industry Investment Fund became the main state-owned capital in the first round. Other participants included: founder Liang Wenfeng, Tencent, CATL, JD.com, NetEase, IDG Capital, Monolith Capital, Zhenxingu Capital, Shixiang Technology... gradually emerging.

Several of these names were surprising.

Investment Community noticed that during DeepSeek's first-round financing process, Monolith Capital and Shixiang Technology intensively established new entities in Tianjin, making them the youngest market-oriented institutions in this round.

First, look at Shixiang Technology. On May 13, 2026, Tianjin Shixiang Investment Partnership (Limited Partnership) was established with a registered capital of approximately 1.51 billion RMB. Information shows Shixiang Technology was initiated in 2019 by former Sequoia investor Li Guangmi. This active figure in the AI industry, in his 30s, once stated that "Coding is the key to AGI," which somewhat resonates with DeepSeek's ethos.

As for Monolith Capital, during this period, they successively established Tianjin Lisi Xingling Venture Capital Partnership (Limited Partnership), Tianjin Lisi Mingtang Enterprise Management Consulting Partnership (Limited Partnership), Tianjin Lisi Xingque Venture Capital Partnership (Limited Partnership), Tianjin Lisi Xinghan Venture Capital Partnership (Limited Partnership), Tianjin Lisi Mingjun Enterprise Management Consulting Co., Ltd., and other entities.

Under the leadership of Cao Xi, Monolith Capital, though young, has a unique character. Cao Xi previously revealed that internally, they do not pre-set investment directions for the team, nor even have formal titles; it's more about "doing what is reasonable."

"The most important direction for the next decade, I believe, is still the new generation of artificial intelligence. We focus relatively more on investing in model companies; 20% of our first fund was invested in Kimi." Cao Xi shared at the 2026 Investment Community SuperLink Conference.

Both Monolith Capital and Shixiang Technology have a common name appearing in their LP lineup—iHealth.

For example, in the aforementioned Shixiang Technology entity, Tianjin Shixiang Investment Partnership (Limited Partnership), SX Global Flagship Fund II L.P. and iHealth Hong Kong Co., Limited invested 760 million RMB and 750 million RMB respectively, with the latter being iHealth's investment platform. The fact that several of Monolith Capital's new entities are registered in Tianjin, the headquarters of iHealth, inevitably sparks speculation.

Battling Competitors, Establishing Employee Stock Ownership Plan

Extreme sincerity and extremely strict rules were Liang Wenfeng's attitude and conditions in the first round.

One detail is: In the first round of financing, investor funds needed to be injected into a limited partnership enterprise managed by Liang Wenfeng, rather than directly into the DeepSeek entity. This ensures Liang Wenfeng maintains absolute control over the company. Simultaneously, all investor shares have a five-year lock-up period, during which they cannot transfer their holdings.

Correspondingly, looking at the first-round financing of approximately 50 billion RMB scale, if calculated based on an overall post-money valuation close to 400 billion RMB, Liang Wenfeng released about 13% of shares externally. Liang Wenfeng alone invested 20 billion RMB, making him the largest investor this round. Therefore, the remaining allocation for external investors is about 7%. It is reported that after the financing was completed, DeepSeek established an employee stock ownership plan, distributing shares based on the actual valuation.

Times are different now.

For three years since its founding, DeepSeek had never raised external capital before, even during the hottest years for large model financing. The message Liang Wenfeng conveyed was that external investors could interfere with company strategy, force short-term commercialization and listing, and affect strategic independence. Therefore, he politely declined batches of investors knocking on the door.

Viewed in the present context, choosing to raise capital isn't a change of stance but more like an acknowledgment of a new competitive order.

China's large model industry has long passed the stage of technological showcases. The difficulty and importance of issues like R&D purity, cost curves, and talent competition are all increasing. Achieving AGI requires not only technological leadership but also a robust organization and a stable ecosystem—and financing can not only provide DeepSeek's internal core talent with a valuation anchor for "how much their options are worth in the market" but also help account for the heavy compute costs on the path to AGI.

Competitors have done this earlier and faster.

Zhipu and MiniMax raised over 8.3 billion RMB and $1.5 billion respectively before their IPOs. In January this year, Zhipu and MiniMax successively went public in Hong Kong, raising a combined total of nearly 10 billion HKD.

Looking at the non-IPO lineup, StepFun completed a new round of nearly $2.5 billion financing in May, with total financing for the year exceeding 20 billion RMB, bringing its IPO closer. There's also Moonshot AI (Kimi), often compared to DeepSeek externally, which was reported to have initiated a new round of financing, with its pre-money valuation rising to $31.5 billion.

A subtle scene: Moonshot AI founder Yang Zhilin and Liang Wenfeng are both from Guangdong, one from eastern Guangdong and the other from western Guangdong. These two "Guangdong AI Talents" are expected to occupy a trillion-yuan valuation territory in China's large model landscape.

The Sci-Tech Innovation Board Listing Window

Everything flows; the world has never been still.

Not long ago, MiniMax's Yan Junjie mentioned a past event in a panel discussion: Two years ago, he asked Liang Wenfeng if he wanted to do AI Coding? Liang Wenfeng said no—because the consensus at the time was that there might only be 1 to 2 million people in all of China who could write code, which didn't seem like a broad enough market.

Back then, everyone underestimated the imagination of AI to varying degrees and didn't anticipate the immense empowerment that the massive C-end user base's pursuit of productivity could bring to AI large model commercialization. Especially the explosion of OpenAI's ChatGPT at the beginning of the year was almost like an industry watershed: when large models moved from answering questions to executing tasks, the corresponding user boundaries, invocation depth, and commercial possibilities were also revalued together.

DeepSeek is also changing.

For a considerable period, DeepSeek resembled a research institution full of idealistic temperament, leaving the industry with the most profound label of model capability. However, with the release of DeepSeek V4 in late April this year or even a few months earlier, the outside world captured from hiring trends, technical papers, and other aspects that DeepSeek also began focusing on Agent, long context, Coding, and even the product side and AI infrastructure. The latest development: It is reported that DeepSeek is also developing AI inference chips and has privately intensified recruitment of chip design engineers in recent months.

Aiming for AGI, but also adapting to circumstances; open source and inclusive, yet not unaware of how to compete.

The opening of the capital market gate pushes competition to another new dimension.

On June 17th, the Shanghai Stock Exchange announced support for high-quality artificial intelligence large model enterprises that have not yet generated a certain scale of revenue to issue and list on the Sci-Tech Innovation Board. To this end, they formulated and released the "Shanghai Stock Exchange Rules for the Application of Listing Review Rules No. 10—Application of the Fifth Set of Listing Standards of the Sci-Tech Innovation Board for AI Large Model Enterprises," effective upon release.

The news landed extremely fast. On the same day, Zhipu's Sci-Tech Innovation Board IPO辅导 status changed to辅导验收, planning to raise no more than 15 billion RMB. Earlier, at the end of May, MiniMax was already preparing to list on the A-share market.

The times push all ambitious individuals forward, also forcing them to constantly re-answer the same question: How do you plan to continue leading, and what price are you prepared to pay for that leadership.

At this moment, Liang Wenfeng can hardly stay aloof.

This article is from the WeChat public account "Investment Community", author: Feng Yuchen