DeepSeek's valuation doubled overnight, but it's the next batch whose valuations haven't been revealed yet that are truly worth paying attention to.
On the evening of April 22nd, before the U.S. stock market opened.
Alibaba's stock price suddenly soared, with the increase once exceeding 2%.
The news came from the technology media The Information - Tencent Holdings and Alibaba are in talks to invest in DeepSeek.
The target valuation of DeepSeek has been raised from the initial at least $10 billion to over $20 billion.
It has doubled.
But the real question worth thinking about is - among the next batch of AI unicorns whose valuations haven't been revealed yet, who will be the next DeepSeek?
"Two Jumps" in the Valuation of Chinese AI in Two Months
From the first quarter of 2026 to the end of April, the Chinese large - model industry experienced two jumps in valuation.
The first jump occurred in January.
Zhipu and MiniMax, representatives of the "Six AI Dragons", listed on the Hong Kong Stock Exchange almost at the same time.
As of the close on April 22nd - Zhipu's market value was about HK$439.6 billion, and MiniMax's market value was about HK$291.1 billion.
Putting these two figures together made the industry realize for the first time -
AI large - model companies are no longer just financing tools. They can be truly priced by the capital market, and the pricing is not low.
The second jump was DeepSeek on April 22nd.
This "AI laboratory" that has never raised external funds before used one night's market news to pull its valuation from $1 billion to over $2 billion - equivalent to about 136.5 billion yuan.
On the same day, DeepSeek's official API quietly upgraded to a 1 - million - token context, up from the original 128k.
Subsequently, there was news that the new generation of the V4 flagship large model would be released in late April.
▸ Zhipu: Listed on the Hong Kong Stock Exchange in January 2026, with the latest market value of about HK$439.6 billion
▸ MiniMax: Listed on the Hong Kong Stock Exchange in January 2026, with the latest market value of about HK$291.1 billion
▸ DeepSeek: According to The Information's report on April 22, 2026, the target valuation jumped from $1 billion to $2 billion
▸ DeepSeek API: Upgraded to a 1 - million - token context on April 22nd (previously 128k)
▸ DeepSeek V4: According to public reports, it will be released in late April, with a MoE architecture and a total of about 1 trillion parameters
Within four months, the valuation coordinate system of Chinese AI has been completely rebuilt.
Who are the Next Batch: Five Candidates
If Zhipu, MiniMax, and DeepSeek are the first three to take the lead, who is in the next queue?
The so - called "Six AI Dragons" in the industry are Zhipu, MiniMax, DeepSeek, Baichuan Intelligence, Kimi, and Lingyi Wanwu.
Considering the newly entered Jieyue Xingchen, there may be five candidates in the new generation:
First: Kimi
One of the Chinese large - model companies with the largest C - end user scale.
Its Kimi intelligent assistant reached the top level in the industry in terms of cumulative monthly active users from 2024 to 2025. The founder, Yang Zhilin, is from the Tsinghua Institute for Interdisciplinary Information Science, and Alibaba is the largest external investor.
Its scarcity lies in being one of the very few Chinese large - model companies that have truly established a C - end brand image.
But the challenges are also concentrated on the C - end - the daily active user scale of Kimi lags behind the growth rate of DeepSeek after its popularity. It is still unclear whether it will adopt an advertising model or a subscription model in the long run.
Second: Baichuan Intelligence
The founder is Wang Xiaochuan, the former founder of Sogou.
Baichuan's positioning is relatively unique - it has almost given up the C - end consumer market and fully turned to the enterprise - level market, especially the medical industry.
The medical vertical large - model is one of the most difficult but also most valuable tracks in Chinese AI. Because it requires long - term data accumulation, compliance ability for regulatory approval, and in - depth access to hospital channels - these are things that cannot be bought with short - term capital investment.
The problem with Baichuan is that the return cycle of this path is slower than that of other large - model companies.
Third: Lingyi Wanwu
The founder is Li Kaifu.
Lingyi Wanwu has a style that lies between the academic and commercial camps - it not only adheres to the R & D of its own Yi series of large models but also has multiple product lines in the enterprise - level track.
Its overseas business started relatively early and has a certain customer base in markets such as Southeast Asia and the Middle East - this is an ability that many domestic pure B - end manufacturers do not have.
But Lingyi Wanwu also faces a common challenge: at the technical level, compared with open - source models such as DeepSeek and Qwen, how to prove its irreplaceability.
Fourth: Jieyue Xingchen
A relatively low - key company in Shanghai.
The founder is Jiang Daxin, an early core member of Microsoft Research Asia and the Microsoft Xiaoice business line.
The technical label of Jieyue Xingchen is multi - modality - the cross - modal generation ability of text, images, videos, and voices.
The uniqueness of this company lies in that both Shanghai state - owned assets and top VCs have placed bets on it. However, its products are currently mainly targeted at enterprise customers, and the C - end brand image has not been established yet.
Fifth: Zhipu's "Second Curve"
Zhipu has already been listed, but it is actually the company most likely to have a "second - curve" IPO among the above list - that is, to spin off a specific business for independent listing.
For example, the code assistant business and the intelligent agent business under Zhipu AI are very likely to be independently spun off in the next two to three years.
Four - Dimensional Evaluation: Who is Most Likely to Become the "Next DeepSeek"
Putting the five candidates into an evaluation framework -
Dimension One: Valuation Echelon
According to public financing information, the latest valuation of Kimi is about 30 billion yuan, Baichuan Intelligence is about 20 billion yuan, the previous - round valuation of Lingyi Wanwu is about $1 billion, and the valuation of Jieyue Xingchen is about 20 billion yuan.
Compared with DeepSeek's valuation of 136.5 billion yuan - these five are currently all in the "second echelon".
Dimension Two: Monetization Ability
Currently, the companies with clear large - scale revenues are mainly Baichuan (medical B - end) and Jieyue (enterprise customers).
Kimi of Kimi is a C - end traffic - driven model, and its monetization model is still being adjusted. The overseas business revenue of Lingyi Wanwu is relatively considerable, but the scale is not significant enough.
Dimension Three: Shareholder Structure
Kimi - Alibaba is the main shareholder.
Baichuan Intelligence - Alibaba, Tencent, Xiaomi, Shenzhen Capital Group, and Kingsoft.
Lingyi Wanwu - Alibaba, Sequoia Capital, and Capital Today.
Jieyue Xingchen - Shanghai state - owned assets, Gao Rong Capital, and GSR Ventures.
From the perspective of shareholder structure, companies with in - depth participation of large Internet companies are more likely to receive support from the capital market in the future.
Dimension Four: Listing Window
Existing cases in the Hong Kong Stock Exchange have shown that Zhipu and MiniMax took the HKEX 18C path (the listing system for specialized and innovative technology enterprises).
This system is relatively friendly, not requiring short - term profitability, only requiring a "clear commercialization prospect".
Corresponding to the five candidates - Baichuan Intelligence and Jieyue Xingchen are most likely to choose the HKEX 18C; according to public reports, Kimi once considered listing with an overseas structure; as an enterprise with a relatively high proportion of overseas business, Lingyi Wanwu also has the possibility of dual - listing.
▸ Kimi: Leading in C - end traffic, with the latest valuation of about 30 billion yuan, and Alibaba as the main investor
▸ Baichuan Intelligence: Medical vertical B - end, with a valuation of about 20 billion yuan, participated by Alibaba/Tencent/Xiaomi/Kingsoft
▸ Lingyi Wanwu: Early start in overseas business, with a valuation of about $1 billion, invested by Sequoia Capital/Alibaba/Capital Today
▸ Jieyue Xingchen: Multi - modality technology, with a valuation of about 20 billion yuan, supported by Shanghai state - owned assets + Gao Rong Capital + GSR Ventures
▸ HKEX 18C system: Both Zhipu and MiniMax have listed through this path (January 2026)
A "Potential Player" Worth Mentioning
In addition to the above five companies, there is a company worth mentioning separately - Silicon Base Mobility.
It is not a "large - model company" in the traditional sense, but an AI computing power service platform - providing cost - effective inference services for open - source models such as DeepSeek.
In the new ecosystem of open - source model commercialization, such a "computing power layer" company plays a key bridging role between models and applications.
The customer base of Silicon Base Mobility is expanding rapidly. If the open - source ecosystem further flourishes after DeepSeek V4, the growth curve of this company may be steeper than that of companies simply doing models.
Companies that don't make models may actually run faster.
Window Period: The Next 18 Months
Based on the HKEX 18C, the fifth set of standards of the STAR Market of the A - share market, and the historical rhythm of NASDAQ, the next wave of IPO window periods for Chinese AI unicorns will most likely fall between -
The fourth quarter of 2026 and the fourth quarter of 2027.
Three key time anchor points -
First - in the fourth quarter of 2026, if DeepSeek V4 can present a clear commercialization rhythm after its release, it may trigger the secondary market to re - price the entire AI sector, which is beneficial to all candidate companies.
Second - in the first half of 2027, after the HKEX 18C rules have been in operation for 18 months, the review channels for the second batch of technology unicorns will become more mature.
Third - in the second half of 2027, after the U.S. presidential election, the market sentiment will stabilize, opening a second door for companies going public in the United States.
This is not the story of the next DeepSeek, but the story of the next Baidu.
In 2005, Baidu listed on NASDAQ.
In 2014, Alibaba listed on the New York Stock Exchange.
In 2018, Meituan - W listed on the Hong Kong Stock Exchange.
Every ten years or so, a group of listed giants with a scale of over 100 billion yuan will emerge in the Chinese Internet industry.
Now in 2026, DeepSeek is the detonation point of the valuation jump.
But what is really worth betting on is which of these candidates can complete the last mile in the next 18 months and become a real "public company".
Information Sources
1. The Information: Tencent and Alibaba are in talks to invest in DeepSeek (April 22, 2026)
2. Securities Times: Just now, a sharp rise, DeepSeek suddenly spreads important news (April 22, 2026)
3. Yicai Global: DeepSeek API upgraded to a 1 - million - token context (April 22, 2026)
4. Cailian Press: Alibaba's U.S. stocks soared by more than 2% before the market opened (April 22, 2026)
5. New Fortune Magazine: With a net profit of 300 million and a valuation of 42 billion, Unitree Technology is sprinting for an IPO (April 2026)
6. 36Kr: Research on the path of Baichuan Intelligence's medical large model (2025 - 2026)
7. Huxiu: The performance of Zhipu and MiniMax in the first month after listing on the HKEX 18C (February 2026)