After the AI-related shares worth HK$1.7 trillion in the Hong Kong stock market were released from the lock-up period, the market value of Zhipu AI and MiniMax has dropped by half.
On July 17, Zhipu closed at HK$1107, down 28.48% for the full day, with its latest total market cap falling to around HK$515.4 billion.
Not long ago, on June 22, Zhipu's market value exceeded HK$1 trillion for the first time. When its stock price peaked at HK$2980, its maximum market cap even surpassed HK$1.3 trillion.
In just 25 days, its market value has evaporated by approximately HK$872 billion.
The situation for MiniMax is also not optimistic: it fell 15.63% on July 17, closing at HK$216 per share, with a market cap of about HK$75.4 billion, down more than 80% from its all-time high of HK$1330.
What is more special is that when market sentiment is extremely fragile, and both retail and institutional investors are asking "Has the Hong Kong stock AI bubble burst?", unconfirmed rumors about Zhipu suddenly spread in the market: as of this month, Zhipu's ARR (Annual Recurring Revenue) has reached 1 billion US dollars. Some anonymous insiders even disclosed that Zhipu's ARR increased 15 times between January and July 2026.
If this data is true, it means the ARR growth rate that Anthropic achieved in 15 months took Zhipu only 5 months.
This looks like a very precise crisis narrative hedging operation. Behind it, Zhipu probably hopes to anchor itself to Anthropic's valuation framework, so that investors will reprice it after the share lock-up expiration as a "Chinese version of Anthropic" rather than a "Hong Kong-listed tech stock".
The "Chinese version of Anthropic" has been the core narrative Zhipu has been promoting since its pre-IPO stage. At the time of its IPO, the market set its base price at HK$116.20.
But the real test starts from the share lock-up expiration.
SPDB International estimates that in 2026, the total market value of unlocked shares in the Hong Kong stock market will be about HK$1.7 trillion, nearly three times higher than that in 2025. There are 6 months in the year where the unlocked share size exceeds HK$100 billion, with September alone reaching as high as HK$530 billion.
Historical data shows that the three months before and after the share lock-up expiration are the period with the most concentrated stock price pressure, and about 58% of companies underperform the Hang Seng Index.
Why does this happen?
Because the IPO sets the initial price, and the share lock-up expiration verifies the real price.
The amplified narrative power brought by listing
Listing amplifies the story, and makes capital pay for imagination.
Let's continue to take Zhipu and MiniMax as examples.
Calculated based on their closing market cap in Hong Kong dollars on July 17 against their 2025 revenue, the two companies have a price-to-sales ratio of 655 times and 122 times respectively, while OpenAI's is 28 times, and Anthropic's is only 20 times.
Although the two companies' price-to-sales ratios have fallen significantly since the share lock-up expiration, they are still about 23 times and 4 times that of OpenAI, and about 33 times and 6 times that of Anthropic respectively.
This certainly does not mean that Zhipu and MiniMax are more valuable than OpenAI, but sometimes the market does not pay for absolute value, but for relative scarcity.
There are very few true large model companies in the Hong Kong stock market, so capital naturally pours all its AI-related imagination into these few players.
During the dot-com bubble, people bet on the internet changing the world; today, people bet on AI changing the world.
Johnny, the initiator of Honghu Summit, whom we once interviewed, noticed that many new stocks rise instead of falling after listing, "There will be a group of people acting like a hype team, creating a set of logic to justify that your current valuation is reasonable."
The Nobel laureate in Economics Robert Shiller's Narrative Economics tells us: People often buy stories rather than cash flows. Chasing rising prices and following investment trends are all driven by narratives.
The benefits that narratives bring to enterprises are obvious, such as improved financing capabilities, faster access to resources, and faster-than-expected development speed.
But stories can never replace real performance. Because if reality fails to catch up with the story, capital will not keep paying unconditionally.
The IPO is a magnifying glass, as well as a risk warning. In the short term, the carnival brought by narratives will push stock prices higher; but in the end, the narrative must be backed up by fundamental performance.
Share lock-up expiration changes the rules of the game
Once the share lock-up expires, the rules of the game change completely.
Before the expiration, the market kept discussing how high Zhipu and MiniMax's stock prices could go? How much value would Moons AI have after listing?
Now the market is gradually getting its answers: On July 8, Zhipu opened at HK$1563, and now it is at HK$1107; on July 9, MiniMax opened at HK$359.8, and now it is at HK$216...
Next, the market will face massive share supply.
A trillion-level unlocked share size is not easy for the Hong Kong market to digest. Many companies have small floating share capital, and market liquidity is concentrated in a small number of leading companies.
Johnny shared his judgment on this: "Investors will definitely rush to sell. The real problem these companies face at that time is not just the decline in market value, but more likely the depletion of liquidity. Investors' main goal may become to exit at the peak as soon as possible, choosing to hold cash and wait and see. Moreover, after some tech innovation stocks are included in the index, if their prices fall after the lock-up expiration, it will inevitably drag down the related index, and this index decline will trigger a chain reaction that affects other stocks."
The share lock-up expiration means a sharp increase in floating share capital, and it also means the market starts to rediscover the real price.
Data from the Hong Kong Stock Exchange shows that the average daily turnover of the Hong Kong stock market in the first four months of 2026 is HK$271.1 billion. Using this number for a rough calculation, the HK$1.7 trillion unlocked share capital is equivalent to about 6.3 trading days of total turnover; if calculated based on the HK$530 billion monthly unlocked share size, the September unlock is equivalent to about 2 trading days of total turnover.
Of course, the share lock-up expiration does not mean all shares will be sold, but this comparison can reflect the market's absorption pressure.
Historical data is more intuitive.
SPDB International surveyed 766 newly listed Hong Kong stocks with regular share lock-up arrangements between 2018 and 2025. They found that the three months before and after the share lock-up expiration are the period with the most concentrated stock price pressure, and about 58% of the companies underperform the Hang Seng Index.
The market may not remember every company's story, but it remembers what happens when supply increases.
The situation faced by AI companies is even more special, as they have several characteristics at the same time: small floating share capital + sharp price increase + huge paper profits for early investors + not fully verified commercialization, and the most important point - extremely high valuation.
Zhipu and MiniMax are currently traded at hundreds of times the price-to-sales ratio while still in a loss-making state. This contrast itself is a risk signal, and it is also the most unique feature of the AI industry.
If the IPO determines at what price an AI company can enter the capital market.
Then the share lock-up expiration determines at what price the capital market is willing to continue to trust it.
An AI version of the Minsky Moment?
The most interesting part of the capital market is that risks often do not arise from pessimism, but from excessive optimism.
When everyone believes in the story, it is actually the safest period, because there will always be buying demand.
The truly dangerous moment is when someone starts to doubt the story.
Economist Hyman Minsky summarized this process into a famous theory: from Hedge Finance, to Speculative Finance, and then to Ponzi Finance where you just need someone else to take over the assets.
Eventually, when a certain event occurs, the market begins to re-examine valuations, and the Minsky Moment arrives.
Will the wave of share lock-up expiration become the Minsky Moment for Hong Kong's AI stocks?
Perhaps the market has already opened a real-life "textbook".
On its listing day, Zhipu's market value was about HK$54 billion. Calculated based on its market value at that time against its historical revenue, its price-to-sales ratio was about 75 times.
This valuation was already very high, but it was still within a range that "can be explained by narrative": the Chinese version of OpenAI, Tsinghua University background, leading GLM technology.
A 13% increase on the first day proved that investors believed in its narrative, but the valuation anchor was still close to "fundamental performance".
In the following 5 months, Zhipu's stock price surged from HK$116 all the way to HK$2980, with a peak market cap of HK$1.33 trillion, and its P/S ratio skyrocketed to 1570 times.
If you deposit HK$1.33 trillion in Bank of China, according to the fixed deposit interest rate in June, the annualized interest will be about 16 billion Hong Kong dollars, while Zhipu's 2025 revenue of 7.24 billion yuan is very tiny compared to that interest income.
This means the only reason investors buy it is because they believe someone will take over their shares at a higher price in the future — retail investors from the A-share market after its IPO, overseas funds, or the next person who believes in the "Chinese version of Anthropic" story.
On the share lock-up expiration day, nearly 70% of cornerstone investors stated they would hold Zhipu's shares for the long term. The next day, the company announced a HK$31.4 billion top-up placement. Afterwards, investment banks kept raising their target price to HK$2400, and the company started the process of dual A+H listing. On July 17, there was even an "unofficial" positive rumor about its ARR reaching 1 billion US dollars. Zhipu's fundamental performance has not changed at all, and GLM-5.2 is still the top model...
But the 28% plunge shows that the market has begun to re-examine the company's price.
Once doubt sets in, buying demand may not disappear completely, but faith starts to be discounted. Buyers no longer fully believe the story, sellers no longer expect a rebound, the valuation anchor will shift from the initial imagination to discounted cash flow, and the valuation will return to a reasonable range.
Capital has become rational and picky.
The boom of "having AI concepts" may have passed with the share lock-up expiration, and "whether it is worth this price" has become the top concern for everyone now.
Not all high valuations will collapse. Anthropic provides a reference for passing this stress test: its ARR surged from about 9 billion US dollars at the end of 2025 to 44 billion US dollars in May 2026, with its revenue growing 5 times in just a few months. Its valuation increased from 380 billion US dollars to 965 billion US dollars, which is geometric-level growth on an annualized basis.
For such companies, high valuations can at least be backed up by fundamental performance. But what really matters is never the valuation itself, but whether the growth can catch up with the narrative.
The next company to enter the examination room will most likely be Moons AI.
In the past six months, Moons AI's valuation skyrocketed from 4.3 billion US dollars to 20 billion US dollars, nearly 5 times higher. Its total financing has exceeded 37.6 billion yuan, making it the large model startup with the largest cumulative financing in China. Its ARR reached 200 million US dollars in April this year; calculated based on its valuation at that time, its price-to-sales ratio was about 100 times.
The extremely high valuation in the secondary market provides a psychological safety margin for its pre-IPO financing. In other words, it is also using data to support its story.
Whether from the perspective of public attention or financing capabilities, it is becoming the next consensus in the capital market.
The capital market never lacks stories.
One story ends, and another starts immediately.
The wheel keeps turning, and life goes on forever.
But when Moons AI really sits at the gaming table, the market will still ask the same question: Can your narrative support your shares after the lock-up expiration?
This article is from WeChat official account "UnDefined", author Cheng Kele, published with authorization from 36Kr.