Zhipu and MiniMax's A-share listing bid faces "July Siege"

On June 5th, the global technology market experienced a "Black Friday."

The Asia - Pacific market weakened first. The South Korean KOSPI index closed down 5.54%. SK Hynix fell nearly 10%, and Samsung Electronics dropped more than 6%. The Hong Kong stock market was also under pressure. The Hang Seng Technology Index closed down 1.75%. Large model concept stocks led the decline. Zhipu fell 9.05%, and MiniMax closed down 16.65%.

That evening, US chip stocks faced even more intense selling. NVIDIA fell more than 6%, with its market value evaporating by about $328.4 billion in a single day; Micron Technology fell more than 13%; Marvell Technology fell more than 16%; Intel and AMD both fell more than 11%. The Philadelphia Semiconductor Index closed down 10.3%, marking the largest single - day decline since March 2020.

The market resonance was mainly triggered by two factors: First, the US non - farm payroll data for May exceeded expectations, leading the market to reprice the Federal Reserve's monetary policy path. Second, the technology sector had accumulated excessive gains previously, and its valuation was at a relatively high historical level. Funds chose to take profits at high levels, and the chips that had flooded in earlier began to loosen.

It is worth noting that the correction of external valuations is overlapping with the pressure on the Hong Kong stock market itself. Wind data shows that the scale of Hong Kong stock unlocks in 2026 will reach HK$1.55 trillion, a nearly three - fold increase compared to about HK$600 billion in 2025, setting a new annual record for unlocks.

In this wave of unlocks, Zhipu and MiniMax are particularly worthy of attention. The unlock time for both companies is concentrated in early July. As the unlock window approaches, the two companies almost simultaneously announced plans to return to the A - share market. The combination of external valuation correction and internal unlock pressure is prompting the market to re - examine the pricing logic of these companies.

Intensive Unlocks of AI Star Stocks: Who Will Be at the Center of the Storm?

The correction of external valuations is overlapping with the unlock pressure on the Hong Kong stock market itself. Looking at the time distribution, from the end of June to early July, the following six companies will have concentrated unlocks, forming the first wave of impact in this unlock wave.

Unisound (09678.HK) will be the first to unlock on June 30th. This involves 38 investment institutions, with a total unlock scale of more than 14 billion yuan. As of June 5th, Unisound closed at HK$203.0, with a total market value of about HK$15.16 billion and an unlock market value of about HK$8.12 billion.

Wandian Research believes that the main unlock subjects for Unisound this time are pre - IPO investors and controlling shareholders. Without the centralized restraint of the controlling shareholder, it is easy for the 38 scattered institutional shareholders to form a game where "the first to run has an advantage." This unorganized selling is often more uncertain than concentrated share reductions.

Insilico Medicine (03696.HK) will also unlock on June 30th. A total of about 453 million shares will be unlocked and circulated, accounting for 81.25% of the total share capital, meaning that most of the company's shares will enter the free circulation state. As of June 5th, Insilico Medicine closed at HK$37.98, with a total market value of about HK$21.74 billion and an unlock market value of about HK$17.2 billion.

Wandian Research believes that as a representative of domestic AI pharmaceutical companies, Insilico Medicine is about to face a true "full circulation." The pricing power of the stock price will shift from a small number of chips in the secondary market to the primary market shareholders. The long - term value and valuation logic of the company will become important factors determining the decision of institutional investors to stay or leave, and the final outcome will depend on the institutions' judgment of the long - term value of the AI pharmaceutical track.

Just after the first small peak on June 30th, Biren Technology (06082.HK) will unlock again on July 2nd, involving 23 cornerstone investors. As of June 5th, Biren Technology closed at HK$55.80, with a total market value of HK$136.1 billion and an unlock scale of 6.37 billion yuan in this round.

Wandian Research believes that although the unlock scale seems small, it should be noted that this round of unlocks is mainly for cornerstone investors. Among them, 3W Funds has the highest subscription ratio of 13.74%, and other major institutions include Qiming Funds, Aspex Master Fund, etc. Against the background of floating profits in their positions, the exit intention of cornerstone investors is the key point to observe.

On July 8th, the market will face the key protagonist of this wave of technology unlocks - Zhipu (02513.HK). This time, 11.9859 million shares will be unlocked, involving about 11 cornerstone investors. As of June 5th, Zhipu closed at HK$1297.0, with a total market value of about HK$578.3 billion and an unlock market value of about HK$15.55 billion.

It is worth noting that for Zhipu's real tradable shares, an unlock scale of more than HK$15 billion is no small matter. Currently, Zhipu's real tradable ratio is less than 4%, and the tradable shares are extremely narrow. The stock price is easily influenced by a small amount of funds, and a small amount of buying can drive a huge increase in market value.

Wind data shows that in May 2026, Zhipu rose 83.76% in a single month, with its market value increasing by nearly HK$400 billion. However, the cumulative turnover rate in that month was only 11.89%, and the monthly trading volume was 31.101 billion yuan. A trading volume of tens of billions pushed up a market value of hundreds of billions. As the restricted shares are unlocked and the real tradable shares in the market expand, the valuation supported by scarce chips before will face a direct test, and the pricing basis of the stock price will also be reconstructed accordingly.

Similar to Zhipu is MiniMax (00100.HK). Currently, the company's real tradable shares are less than 6%. After the unlock on July 9th, 107.4166 million shares will be unlocked and circulated, accounting for 34.25% of the total share capital. Among them, financial investors hold more than one - third of the shares, involving 14 cornerstone investors, and the unlock scale is 11.685 billion yuan. As of June 5th, MiniMax closed at HK$553.0, with a total market value of about HK$173.44 billion and an unlock market value of about HK$59.39 billion.

Compared with Zhipu, MiniMax faces more direct unlock pressure. A senior primary market investor told Wandian that financial investors have the strongest motivation to exit and the least restraint. A proportion of more than one - third means there is a certain selling pressure after the unlock. The sudden expansion of the tradable shares, combined with the exit motivation of financial investors, makes MiniMax's unlock impact more severe than Zhipu's in terms of both volume and structure.

Finally, as a representative of embodied intelligence concept stocks, Geek+ - W (02590.HK) will also face a large - scale unlock in July, involving 680 million unlocked shares, accounting for 51.5% of the total share capital. As of June 5th, the company closed at HK$14.99, and the unlock market value exceeded HK$10 billion.

Objectively speaking, unlocks do not equal selling. The unlock natures of the six companies are different. Behind each unlock, there is a different shareholder structure and valuation logic. What is being tested is not just the market liquidity itself. For Zhipu and MiniMax, which have extremely small real tradable shares and huge previous gains, the unlock is precisely the starting point to test the company's value.

From Hong Kong Stocks to A - Shares: The Dispute over Valuation and Route between Zhipu and MiniMax

As the unlock window approaches, Zhipu and MiniMax have successively announced plans to return to the A - share market.

On May 31st, MiniMax announced in the Hong Kong Stock Exchange that the company has hired professional consultants to provide advice on meeting the listing conditions of the Science and Technology Innovation Board and signed a tutoring agreement.

On the evening of June 1st, Zhipu announced in the Hong Kong Stock Exchange that it plans to apply to the Shanghai Stock Exchange for listing on the Science and Technology Innovation Board, issuing no less than 9.0988 million and no more than 38.769 million new A - shares, raising no more than 15 billion yuan for the construction of a general AI base model, a one - stop service platform for large model MaaS, and to supplement working capital.

The almost simultaneous start of the return to the A - share market by MiniMax and Zhipu is not only due to the expected higher valuation arbitrage in the A - share market but also fundamentally driven by the capital demand.

Looking at the investment scale in the global AI competition, one can understand the urgency of financing. At the end of April, Google raised its capital expenditure forecast for 2026 to between $180 billion and $190 billion, doubling that of the full year of 2025, which was $91.45 billion.

Following that, on June 1st, Google's parent company Alphabet announced a plan to raise $80 billion through an equity issuance for the construction of AI computing infrastructure. This will be Google's largest equity financing in history.

When global leading players are investing in the arms race in the order of hundreds of billions of dollars, the financing pressure on large model companies in the Hong Kong stock market that have not yet made a profit is self - evident. Zhipu plans to raise 15 billion yuan, and MiniMax has also started tutoring. It is more appropriate to interpret this as an inevitable choice under the pressure of intense industry competition rather than just valuation arbitrage.

From the perspective of business paths, although MiniMax and Zhipu are in the same large model track, their directions have diverged.

Zhipu is betting on "domestic base + government and enterprise binding." Its first financial report after listing shows that in 2025, the company achieved a total revenue of 724 million yuan, a year - on - year increase of 131.9%; the ARR of its core MaaS platform was about 1.7 billion yuan, a 60 - fold increase within 12 months; the annual comprehensive gross profit margin was 41%, far exceeding the industry average.

Although Zhipu's adjusted net loss in 2025 was 3.18 billion yuan, an increase compared to 2.46 billion yuan in 2024, the certainty of its B2B business, the stickiness of government and enterprise orders, and the layout of domestic computing power adaptation make it easier to be included in the narrative framework of "AI infrastructure."

For Zhipu, labels such as basic model, government and enterprise services, MaaS platform, and domestic substitution are closer to the hard technology logic familiar to A - share investors and are more likely to obtain valuation premiums in the current policy environment.

Compared with Zhipu's focus on the B2B market, MiniMax has chosen a more application - oriented C2C path. The financial report shows that in 2025, the company's total revenue was $79.038 million, a year - on - year increase of 158.9%; the gross profit was $20.079 million, a year - on - year increase of 437.2%, and the gross profit margin increased to 25.4%; the adjusted net loss was $250 million, basically the same as the previous year.

Compared with Zhipu's products, MiniMax's products such as AI companionship, video generation, voice, and multi - modal interaction naturally have stronger communication power and user reachability. While they can quickly gain popularity, the product experience is easy to perceive, and the model's shortcomings are also easy to be magnified. Opportunities and risks coexist.

A recent billing controversy at MiniMax exposed the vulnerability of the C2C path. The company previously announced that starting from June 1st, 2025, it would comprehensively adjust the API service billing model from "pay - per - use" to "pay - per - token."

This change at MiniMax quickly triggered a strong backlash among developers and enterprise users. A large number of users expressed their dissatisfaction through official communities, complaint platforms, and social media.

Users generally reported that after adopting the token - based billing, the token consumption for completing the same scale of tasks was significantly higher than the industry's mainstream level. A developer gave an example that the monthly package quota that could support hundreds of rounds of interaction under the pay - per - use model was exhausted in just a few days under the token - based billing, and it was difficult to trace the specific token calculation details from the background logs. Users questioned the lack of transparency in MiniMax's token definition, word segmentation logic, or billing coefficients.

Facing the public pressure, MiniMax's parent company, Xiyu Technology, issued an apology statement through official channels on the evening of June 2nd, admitting that there were problems of "insufficient communication and inadequate consideration of the transition plan" during this adjustment. It promised to continue to retain the original rights for old users without weekly limits and to give additional free token packages to users with valid payment records and increase the upper limit of the token total in the basic package.

Although this controversy was temporarily resolved with an apology and compensation, it revealed a deeper problem: C2C users are extremely sensitive to pricing, and their switching costs are extremely low. A single billing adjustment can trigger concentrated complaints, indicating that the user relationship is based on a fragile balance between price and experience.

MiniMax founder Yanjunjie once said that in the AI era, traditional products are more like channels, and the real product is still the model itself. However, when the model's capabilities have not yet shown a significant gap compared to competitors, it is difficult to overestimate the price increase space and user loyalty. This is precisely the core reason why MiniMax always has a discount in terms of business model certainty compared to Zhipu.

The path differences between MiniMax and Zhipu are also reflected in the market valuations. In terms of stock price trends, Zhipu's stock price fell from a high of about HK$1993 on May 29th, with a decline of more than 30%, and its total market value remained above HK$570 billion.

In contrast, MiniMax's performance was weaker. Its market value continued to decline after exceeding HK$380 billion in mid - March. As of the close on June 4th, it was about HK$208.1 billion, a decline of nearly 45% from the high point, and its total market value was less than half of Zhipu's. The difference in the correction amplitudes of the two companies reflects the different risk pricing by the market between the certainty of the B2B market and the uncertainty of the C2C market.

Narrative Anchor Change: The Ultimate Test for Zhipu and MiniMax's Return to the A - Share Market

The differentiation in valuation is just a superficial phenomenon. A deeper test lies in: When the AI industry logic shifts from "model priority" to "architecture priority," who can establish a truly irreplaceable ecological niche in the complete architecture of computing power, cost, and business closed - loop?

Looking deeper, whether it is Zhipu or MiniMax, the story they want to tell when returning to the A - share market is essentially not just about simple financing or revaluation, but a systematic reconstruction of the entire AI industry logic in the mainland capital market.

In the past period, large model companies in the Hong Kong stock market were more priced as "new technology consumer goods" or "global platforms," emphasizing the imagination space of user growth, ecological expansion, and cross - market expansion.

However, in the A - share market, especially within the framework of the Science and Technology Innovation Board, the valuation anchor of this narrative logic is more inclined to "industrial chain security" and "strategic infrastructure." There are significant differences in the underlying logic between the two narrative systems of the A - share and Hong Kong stock markets.

Therefore, for Zhipu and MiniMax, the core proposition they really need to answer is: How to transform from "model ability demonstrators" to "builders of independent technology stacks." This transformation of identity is directly reflected in the different choices of the two companies regarding pricing power and cost structure.

Since the beginning of 2026, leading cloud service providers have collectively raised prices, with some increases exceeding 400%. Behind the price increase is the token inflation caused by AI agents, and the low - price competition has essentially ended. At