With Zhipu hitting a valuation of a trillion yuan, is DeepSeek far behind?
Recently, a Chinese foundation model company has refreshed the market value record.
During intraday trading, Zhipu's market value exceeded HK$1 trillion, with a year-to-date increase of over 1,900%. At the close, the company's total market value reached HK$1.1 trillion, approximately half of Alibaba's, nearly twice that of Meituan, and three times that of JD.com.
Less than half a year after its listing, Zhipu has soared from a company with a market value in the HK$50 billion range to join the trillion-HK-dollar market value club on the Hong Kong stock market.
Above the threshold of the "trillion club" on the Hong Kong stock market are industrial giants like Tencent, Alibaba, China Mobile, and Industrial and Commercial Bank of China. All of them have proven user access points, cash flows, or infrastructure attributes.
Zhipu's entry into this league indicates that the public market's valuation of Chinese large model companies is constantly expanding.
However, behind its roller-coaster-like growth rate lies a debate about whether it is a myth or a bubble.
Zhipu's first earnings report after listing revealed its full-year revenue data for 2025: revenues of 724 million yuan, a net loss of 4.718 billion yuan, and R & D expenses of 3.18 billion yuan.
A company with revenues of 700 million yuan last year has already reached a trillion-level market value. Bullish investors see it as a breakthrough in the valuation of China's AI industry, while bearish investors believe that the current stock price has a strong bubble element and has far outpaced the fundamentals.
To some extent, this debate has gone beyond Zhipu and has risen to the valuation of China's large model industry itself.
The latest news shows that DeepSeek's valuation has exceeded $50 billion after a new round of financing, and the capitalization expectations of Jieyue Xingchen and Yuezhi Anmian are heating up, both targeting the Hong Kong stock market.
The valuation curve of Chinese large model companies is being lifted from the primary market to the secondary market. With Zhipu becoming the first large model enterprise with a trillion-dollar market value, this curve is constantly breaking new boundaries upward.
1
The direct catalyst for Zhipu's recent market value surge is GLM - 5.2.
Since last year, the mainstream model upgrades in the industry have focused on coding, agents, long - range tasks, and complex engineering capabilities.
In these scenarios, Zhipu's GLM series can rank among the first echelon, with Anthropic being the world's strongest benchmark.
Since this year, Anthropic has completed $65 billion in financing, with a post - investment valuation of $965 billion. The core reason for its popularity mainly comes from the revenue generated by Claude Code in the developer scenario.
Officially disclosed data shows that Claude Code's annualized revenue has exceeded $2.5 billion, doubling compared to the beginning of the year.
This example has provided a new valuation anchor for global AI companies.
C - end chat products prove the user scale, while code and enterprise agents prove the payment ability. The former creates buzz, and the latter is closer to revenue. The commercialization ability behind revenue is the key factor for the capital market to evaluate large model enterprises.
Zhipu's recent re - evaluation is because the market has placed it in the context of a "domestic Anthropic".
Recently, the interaction between Elon Musk, the CEO of Tesla, and Tang Jie, the founder of Zhipu, has amplified this imagination.
Musk predicted that China may have a model approaching the level of Anthropic Fable in the first quarter of 2027.
Tang Jie responded that it won't take that long. Although this response itself is brief and contains little information, it highlights that domestic models are already in the global narrative of catching up with leading models, and it depends on who can keep up with the pace of the top competitors.
Since this year, in addition to the soaring stock price of Zhipu, there have been continuous capital moves in the domestic model circle.
Zhipu passed a resolution for A - share issuance at the beginning of this month and plans to list on the Science and Technology Innovation Board to raise no more than 15 billion yuan. Among them, 12 billion yuan will be invested in the R & D of general - purpose large foundation models, and 2 billion yuan will be invested in the MaaS one - stop service platform.
Almost at the same time, MiniMax signed a tutoring agreement with CITIC Securities and officially launched the A - share IPO process, and is evaluating the issuance of RMB shares to list on the Science and Technology Innovation Board. The simultaneous progress of the A - share listing paths of two large model companies on the Hong Kong stock market shows that China's AI valuation curve is extending from the flexibility of the Hong Kong stock market to the industrial financing of the A - share market.
Meanwhile, DeepSeek is frantically attracting financing in the primary market.
Recently, DeepSeek completed financing of over $7.4 billion, with a valuation of over $50 billion. This AI giant from Hangzhou has pushed the price of China's large model primary market to a new high with its excellent model efficiency, global developer influence, and open - source ecosystem.
In the secondary market, after Zhipu and MiniMax were listed one after another, the Hong Kong stock market is also turning to embrace the AI industry.
Since this year, 65 companies have been listed on the Hong Kong stock market, raising a total of over HK$176.5 billion. High - tech and high - end manufacturing enterprises contributed nearly HK$129.3 billion, accounting for more than 70%. Semiconductor, electrical equipment, and AI companies have become the main lines of issuance, and funds are using the Hong Kong stock market as a re - pricing market for China's high - tech assets.
In other words, Zhipu's trillion - dollar market value not only comes from the excellent performance of GLM - 5.2 but also from several factors occurring simultaneously: the rising valuation of global AI assets, the accelerating capitalization of Chinese large model companies, the lack of pure large model targets on the Hong Kong stock market, and the re - concentration of high - tech funds.
Therefore, what the market is really buying into is that the valuation boundary of China's AI can continue to expand. However, the wider this boundary expands, the more the debate about the industrial bubble will continue.
2
When most investors see Zhipu's trillion - dollar market value, their intuitive judgment is that there must be a bubble component.
A company with annual revenues of over 700 million yuan has an intraday market value of over HK$1 trillion, which will make any valuation model nervous.
Moreover, Zhipu's stock price has fluctuated significantly before. At the end of May, its intraday market value once exceeded HK$880 billion and then quickly declined. During today's climb to the trillion mark again, the intraday volatility was still severe, and it lost over HK$100 billion in market value in just a dozen minutes after reaching the high point.
However, this is exactly the most special aspect of the AI market. The market admits that the prices are high while continuing to chase the leading targets.
The simultaneous occurrence of bubble warnings and valuation increases forms the background of this round of AI capital cycle.
Meanwhile, the new - share market on the Hong Kong stock market has shown a divergence.
In the first quarter of this year, the break - even rate of IPOs in Hong Kong rose to 44.74%, significantly higher than last year. Traditional industries and companies with insufficient subscriptions are under pressure, while high - tech, AI, and new - energy enterprises are still being chased by funds. Funds are not buying new shares indiscriminately but are concentrating their risks in a few sectors.
Obviously, the AI industry represented by Zhipu belongs to the assets where risks are concentrated. Breaking through the trillion mark relies on the market's scarce expectation of Chinese AI infrastructure companies.
However, capital will not invest indefinitely. In early July, Zhipu and MiniMax will face the first round of restricted - share unlocks after listing.
Zhipu's restricted - share unlock will take place on July 8, with a scale of approximately 25.6816 million shares, accounting for about 5.76% of the total share capital. Based on the recent stock price, the corresponding unlocked market value is about HK$26.9 billion. Before the unlock, Zhipu had about 17.3508 million tradable shares. In other words, even though the proportion of the total share capital is not high, the newly added tradable chips will significantly change the previous low - liquidity structure.
MiniMax is under greater pressure. It will unlock about 107 million shares on July 9, accounting for about 34.25% of the total share capital. Among them, financial investors hold more than one - third of the shares.
Both companies' stock prices have risen significantly compared to their listing prices, so the market naturally worries about the selling pressure in advance.
In fact, not only domestic large models but also the global market is experiencing the same valuation conflict.
Two weeks before Musk and Tang Jie's interaction, SpaceX's market value quickly exceeded $2 trillion after its listing, and its stock price climbed to the range of $210.
This aerospace and satellite - internet company integrated a large amount of resources before its listing to strengthen its position in the narratives of AI infrastructure, xAI collaboration, Starlink network, and future computing power.
Damodaran, a professor at the Stern School of Business at New York University, gave a valuation lower than the market price for this company and reminded that SpaceX's high valuation already includes a large number of future assumptions.
In the past few trading days, SpaceX's stock price has experienced a significant correction and has dropped to the range of $170.
At the beginning of June, Michael Burry, the founder of Scion Asset Management and the prototype investor in the movie "The Big Short", also directed his doubts at Anthropic.
He believes that the R & D cost of the large model behind Claude is too high, and its nearly - trillion - dollar valuation is difficult to support with the current business model alone.
The large cognitive divergence in the industry actually shows that the seemingly opposing views are not contradictory. In the period of rapid industrial growth, the bubble has naturally become a part of the AI valuation.
The key factor determining whether this part of the "bubble" is a long - term premium depends on the commercialization ability of AI enterprises.
3
After breaking through the trillion - dollar market value, the real problem facing Zhipu is the growth rate of its revenues.
In 2025, Zhipu's revenues were 724 million yuan, a year - on - year increase of 131.9%. Among them, the revenues from local deployment were 534 million yuan, a year - on - year increase of 102.3%; the revenues from cloud deployment were 190 million yuan, a year - on - year increase of 292.6%; and the revenues from enterprise - level intelligent agent business were 166 million yuan, a year - on - year increase of 248.8%.
Among them, the ARR of Zhipu's MaaS API platform was about 1.7 billion yuan, a 60 - fold increase in the past 12 months, and the platform's gross profit margin increased to 18.9%. This means that the revenue share of MaaS is constantly increasing.
In the first quarter of 2026, in the face of the strong demand for model calls, Zhipu directly chose to - raise prices.
Since February this year, Zhipu has made a structural adjustment to the price of the GLM Coding Plan package, with an overall increase of at least 30%. The reason is that the user scale and call volume have increased rapidly, and it is necessary to ensure the stability and service quality under high loads.
However, how much revenue growth this part of the revenue corresponds to will be revealed in Zhipu's next earnings report.
Meanwhile, as an independent model company, Zhipu is under great cost pressure.
In 2025, Zhipu had a net loss of 4.718 billion yuan, an adjusted net loss of 3.182 billion yuan, and R & D expenses of 3.18 billion yuan. The R & D investment is more than four times the revenue.
Model companies can burn money, but they must let the capital market see the long - term profit expectation.
Fortunately, the call volume base in the entire Chinese market is constantly rising.
IDC data shows that the call volume of China's enterprise - level MaaS market was 114 trillion Tokens in 2024, soaring to 1,944 trillion Tokens in 2025, and is expected to reach 400,000 trillion Tokens in 2026. Tokens are changing from technical indicators to commercial settlement units.
ByteDance, the leading domestic model company, has seen its daily Token call volume of the Doubao large model exceed 120 trillion, doubling in three months and increasing 1,000 times compared to when it was launched in May 2024. The number of enterprises with a cumulative call volume of over one trillion Tokens has increased from 100 at the end of last year to 140.
Meanwhile, the C - end scenario is also expanding.
QuestMobile data shows that in March 2026, the monthly active user scale of China's AI - native apps reached 440 million. Doubao, Qianwen, and DeepSeek ranked in the top three. AI applications have changed from niche tools to mass products.
However, the user scale does not represent the profit level. The concern about the bubble in the industry is essentially a sense of powerlessness towards the bottomless money - burning of AI and the lack of a profit expectation. In this regard, Anthropic is still the strongest benchmark.
In the second quarter of this year, Anthropic is expected to have revenues of $10.9 billion and achieve an operating profit of about $559 million. If this expectation is fulfilled, it will become a turning - point sample for the commercialization of global AI companies. The market will then believe that large model companies do not necessarily stay in the "burning money for scale" stage in the long term.
This is also the reason why Chinese AI companies are being re - evaluated.
If Anthropic can approach profitability through code, agents, and enterprise APIs, the domestic market will naturally look for corresponding Chinese targets. Zhipu, DeepSeek, MiniMax, Yuezhi Anmian, and several leading AI companies will all be compared on this commercialization curve.
With Zhipu being the first to break through the trillion mark, this has become a common proposition for China's AI industry in the future.
The valuation boundary has been opened, the debate about the bubble will not stop, and market fluctuations and uncertainties still exist.
To consolidate this valuation boundary, large model manufacturers first need to stably convert model capabilities into revenues, and then push revenues towards gross profit and cash flow. At the same time, more domestic manufacturers need to produce effective samples on the way to catch up with Anthropic.
Zhipu's reaching the trillion - dollar market value today is the first high point in the re - evaluation of China's AI industry. Whether this high point can be maintained and whether it can become an industry curve will ultimately be answered by the commercialization achievements of domestic large model manufacturers.
This article is from the WeChat official account "Zimubang", author: Li Zhaofeng. It is published by 36Kr with authorization.