HomeArticle

MiniMax's market capitalization has fallen by nearly 90% from its peak.

Tech商业2026-11-17 18:25
MiniMax has lost 90% of its market capitalization.

On November 16, MiniMax's share price once again hit a new low since its listing. It fell by as much as 15% during intraday trading, bringing its market capitalization down to HK$44 billion. Compared with its peak of over HK$410 billion in March, the company has lost nearly 90% of its value.

For a company listed for less than a year, such a performance curve is striking. But what makes MiniMax special is that its decline did not start from a single underperformance event, nor did it correspond to any specific isolated incident. More accurately, it has undergone a shift in its valuation system — shifting from being a "scarce listed AI target" to a "company that needs to prove its profitability". This transition took less than eight months.

July 9, the Watershed

The root cause of the share price decline can be traced back to July 9. On that day, approximately 153 million shares were lifted from the lock-up period, accounting for 48.9% of the total shares of this Chinese AI developer. With nearly half of the shares gaining tradable rights, the market priced in the expectation of selling pressure in advance, regardless of whether the shareholders actually intended to sell. Although most pre-IPO investors and cornerstone investors promised to hold their shares, the selling pressure materialized rapidly: the share price fell by 17.98% on the day, and accumulated a decline of nearly 40% in the following three trading days.

The lifting of the lock-up became a watershed because it changed the supply structure of the stock. In the early stage of listing, the floating share capital was small, and a small amount of buying orders could push up the price; when nearly half of the shares were unlocked, the pricing power shifted from scarcity to real supply and demand. Since then, the share price has kept falling, hitting a new low on November 16. The gap between the commitment to hold shares and the actual selling pressure also shows that in the face of the lifting of restrictions on nearly half of the shares, willingness statements can hardly offset the market's repricing of liquidity.

Similar scenarios are not uncommon among new stocks on the Hong Kong Stock Exchange. The cornerstone investor commitments repeatedly emphasized at the time of listing often only cover a small part of the total share capital and have a limited term; what really determines the medium-term share price is the absolute size of tradable shares after the expiration of the lock-up period. The proportion of shares unlocked in MiniMax this time is close to half, which has exceeded the normal range of most companies. The market gave a direct response with a nearly 40% decline in three trading days — in the face of changes in the supply structure, the fundamentals have almost no say in the short term.

Four Years From Establishment to Listing, Eight Months of Retracement

MiniMax was listed in January 2026 at HK$165 per share, only about four years after its establishment in 2022, making it one of the fastest-listed companies in the AI sector. On the first day of listing, its share price soared by 109%, and market enthusiasm reached its peak; in March, it once hit a high of HK$1,330, rising more than seven times over the issue price, and then began to fall continuously.

From HK$410 billion to HK$440 billion — calculated from the high point in March, the market value has evaporated by nearly 90% in less than eight months. It is worth noting that the current share price has fallen below the issue price of HK$165. That is to say, investors who won the new share subscription lottery and have held the shares until now have not only given up all the gains, but are also in a floating loss state; buyers who chased high on the first day of listing have suffered even more considerable book losses.

The significance of this set of figures is not to compare who lost more, but to mark the beginning and end of a round of market sentiment. The 109% surge on the first day of listing and the sevenfold increase in March were priced based on "AI scarcity"; the current share price of HK$131 is priced based on "uncertainty". The same company has been valued twice with two completely different logics within eight months.

Going public just four years after its founding is itself a product of this AI boom. Companies need to obtain capital market ammunition as early as possible to support computing power investment, while investors hope to bet on the next platform-level opportunity, and the two sides reached a highly consistent optimistic expectation in early 2026. This expectation was magnified in the early stage of listing, and then corrected one by one by the lock-up expiration, financial reports and the overall sentiment decline of the whole sector. The rapid retracement precisely shows how much of the initial pricing came from imagination about the future rather than calculation of the current situation.

Revenue Up 283%, Loss Up 111%

If the share price reflects expectations, the financial report reflects reality. In the first half of 2026, MiniMax's revenue increased by 283.1% year-on-year to 116.6 million US dollars, exceeding the full-year revenue of 79 million US dollars in 2025. Looking at this figure alone, the growth is almost astonishing.

But on the other side, the adjusted net loss expanded by 111.2% year-on-year to 293 million US dollars — the problem of "growing revenue but struggling to make profits" still exists. With a revenue growth rate of 283.1% and a loss growth rate of 111.2%, the two figures together show that the absolute amount of loss is still expanding, but the growth rate is lower than that of revenue. In other words, the scale effect has not really emerged.

This is the most typical financial profile of current AI companies: the revenue curve rises steeply, and the cost curve is almost equally steep. Computing power, model training, and talent are all rigid expenditures. Before revenue outpaces costs, expanding losses is almost an inevitable result. The market was previously willing to pay for this profile because it believed that growth would eventually cross the critical point; when valuation returns to reality, patience becomes something that needs to be re-justified.

It is worth noting that with revenue of 116.6 million US dollars and a loss of 293 million US dollars — the loss is more than twice the revenue. This ratio means that for every dollar of revenue the company earns, it incurs more than two dollars in adjusted net loss. This does not mean that the business itself is not viable, as both the open platform and native products have clear paying users, but it shows that the current investment intensity far exceeds the speed of revenue growth. For primary market investors, this can be interpreted as "seizing the market window"; for secondary market investors who review reports on a quarterly basis, this is the real pressure that needs to be questioned quarter by quarter.

Overseas Expansion and Revenue Structure

The composition of revenue provides another perspective for observation. MiniMax's revenue is divided into two major segments: revenue from open platform and enterprise-level AI services (API calls and token subscriptions) reached 73.93 million US dollars, accounting for 63.4%; revenue from AI native products (such as Conch AI and MiniMax Code) reached 42.64 million US dollars.

The former is an infrastructure-style business that wins by call volume; the latter is a user-oriented product that wins by experience and low substitutability. At present, APIs and subscriptions contribute nearly two-thirds of the revenue, indicating that MiniMax's main customers are still enterprises and developers, not ordinary consumers. The advantage of this positioning is more stable revenue, while the disadvantage is that the growth ceiling depends on the technical selection of customers, rather than the spontaneous spread of the mass market.

According to other disclosures, as of June 30, the products cover more than 230 countries and regions, and 60.8% of the revenue comes from regions outside mainland China — its degree of internationalization is relatively high among large model companies, but high growth has not yet translated into profitability. The fact that overseas revenue accounts for more than half means, to some extent, higher compliance and localization costs, which are usually reflected in the financial statements in the form of sales and R&D expenses.

Not Only MiniMax

MiniMax is not an isolated case. The share prices of other large model companies on the Hong Kong Stock Exchange, including Zhipu, have also fallen simultaneously, and investors are re-evaluating the value of the entire industry.

After the parameter competition and financing boom, the market's evaluation criteria are changing: compared with model scale and financing amount, more importance is now attached to a reliable profit path. Parameters used to be the easiest story to tell — the larger the number, the more leading it sounded. But when model capabilities gradually converge, the corresponding relationship between parameter scale and commercial returns no longer holds, and the financing amount has changed from a plus point to a cost that needs to be explained.

A deeper change is that large models are increasingly regarded as infrastructure that requires long-term investment, rather than a channel for quick cashing out. The characteristics of infrastructure are huge upfront investment, long return cycle, predictable cash flow but limited explosive growth. Using the valuation logic of growth stocks to price infrastructure itself will lead to drastic corrections when the growth narrative slows down. For AI companies that are still burning large amounts of capital, this means a repricing of their valuation logic.

This revaluation does not mean denying the commercial value of AI. The call volume of open platforms is rising, native products have gained users, and overseas revenue also accounts for 60% — these are all real progress. What has changed is only the reference frame for pricing: in the past, the market asked "how big can this company grow", and now there is an additional question "when can it support itself". Only when both questions can be answered can the valuation be supported; if only the first one can be answered, the share price can only be propped up by sentiment, and sentiment is never stable. For the whole sector, this round of decline is more like a forced coming-of-age ceremony.

Conclusion

Back to the figure of nearly 90%. It is eye-catching enough, but it easily hides the real change: what has been lost is not the value of a certain company, but the applicability of a set of valuation methods. Attributing every fall in share price to the lock-up expiration, or interpreting every rebound as a reversal, will miss the real message conveyed by this curve — the market is learning to price AI in another way, and the cost of learning is usually paid by the first batch of listed companies.

For MiniMax, the 283.1% revenue growth is real, and the 293 million US dollar loss is also real. What the capital market is doing is putting these two things back on the same scale to re-weigh. This process will not end soon, nor will it only happen to one company.

This article is from the WeChat official account "Tech Business", written by Tech Business, and published with authorization by 36Kr.