China is about to have a new richest person.
The first large model company to ring the bell on the Sci-Tech Innovation Board is brewing its IPO plan.
It is reported that DeepSeek has selected CITIC Securities to advance its Sci-Tech Innovation Board IPO, and plans to launch the listing process within this year. Reuters reported that Liang Wenfeng hopes the funds raised from the IPO will help DeepSeek strengthen incentives for core employees and researchers.
Two rounds of financing and the launch of IPO all took place in just four months.
At the moment, DeepSeek's valuation has reached 500 billion yuan, and it is exponentially expanding its team size and iterating its models rapidly. Earlier this year, when financing for large model companies was booming and the first IPO gate was opened, DeepSeek was still a relatively detached player in the capital market. Now, DeepSeek has caught up.
Aiming for a trillion-yuan valuation, China's new richest person may emerge
This year, apart from the gradual release of the DeepSeek V4 model, the two rounds of financing and the IPO plan can be described as a seamless process.
Foreign media reported that DeepSeek has hired CITIC Securities to prepare for its Sci-Tech Innovation Board IPO, and people familiar with the matter said the company is expected to officially submit IPO application materials within this year. However, the official listing counseling agreement has not been signed yet.
According to the current IPO rules of the Sci-Tech Innovation Board, if the total share capital of the company after issuance is not less than 50 million yuan, the number of publicly issued shares shall reach more than 25% of the total shares of the company; if the total share capital after issuance exceeds 400 million yuan, the proportion of publicly issued shares can be reduced to more than 10%.
Even so, Liang Wenfeng still has high control over the company.
In the previous first round of financing, DeepSeek released only about 10% of its shares to the public, with a financing scale of over 500 billion yuan and a valuation of 50 billion US dollars. At that time, Liang Wenfeng requested to verify the identities of all LPs participating in the financing; in the second round of financing launched in July, according to foreign media reports, the company plans to raise 500 billion yuan with a target valuation of 5 trillion yuan. Meanwhile, the Financial Times stated that Liang Wenfeng has begun to personally review some of the final investors, hoping to maintain control of the company and reduce potential governance risks before the IPO.
Even though the IPO still needs time to land, combined with the valuation of DeepSeek's second round of financing, the current market expectation has reached 1 trillion yuan or even 2 trillion yuan.
At that time, the reshuffle of the wealth ranking may also happen: on the Forbes China Mainland Rich List, Zhang Yiming maintained his position as China's richest person with a net worth of 69.3 billion US dollars in 2026. If DeepSeek's valuation really breaks through 1 trillion yuan, Liang Wenfeng is expected to become the new richest person in China due to his high shareholding ratio in DeepSeek.
It is worth mentioning that according to the current public information, no dollar funds, namely foreign capital, have participated in DeepSeek's two rounds of financing. In the eyes of the outside world, this will make DeepSeek's subsequent IPO process on the Sci-Tech Innovation Board relatively simpler and faster.
Liang Wenfeng is making full efforts to move forward
After the listings of Zhipu AI and MiniMax, among China's leading large model companies, DeepSeek and Moonshot AI have the highest valuations.
These two are also the companies waiting for their IPOs.
In the latest round of financing, it is reported that Moonshot AI's valuation has reached 50 billion US dollars. In March this year, Bloomberg reported that Moonshot AI, the company behind Kimi, has initially assessed the possibility of a Hong Kong IPO, and has contacted CICC and Goldman Sachs for potential listing matters. By September, the news further spread that the company is considering going public in Hong Kong as early as this year, with the maximum financing scale expected to reach 5 billion US dollars.
Now, DeepSeek is also keeping pace.
The team size is also expanding simultaneously. In June this year, DeepSeek released a recruitment notice stating that it is striving to expand the size of all departments by at least double. By September 8, the head of the DeepSeek Harness team unprecedentedly released about 150 recruitment positions, involving engineers for large model research platform, Agent framework components, R&D efficiency infrastructure, data engineering and other directions.
We learned that DeepSeek's internal team considered developing C-end businesses in the first half of the year, but the plan was adjusted a few months later. This to a certain extent shows DeepSeek's attitude of making rapid decisions and responses in the face of changes.
In terms of models, looking back at the past two months of booming financing, DeepSeek's update and iteration speed has also accelerated significantly:
On July 31, the official version of DeepSeek V4 Flash was launched; on August 13, the official version of V4 Pro was released, and DeepSeek announced a substantial adjustment to API prices on the same day. The developer preview version of DeepSeek Harness was open for testing and open sourced; by August 31, the weights of V4-Flash-Vision-Exp were open sourced; just today, DeepSeek will officially release the V4.1 Flash model, which comprehensively surpasses V4 Pro in all indicators.
At this point in time, whether it is the capital market environment, the focus of model competition or the industry narrative, it is very different from 2025 when DeepSeek R1 amazed the whole industry.
Only by continuous evolution to maintain the leading position can the company steadily go through the eve of human AGI.
This article is from the WeChat official account "PEdaily" (ID: pedaily2012), the author is Feng Yuchen, 36Kr publishes this article with authorization.