MiniMax's Hong Kong shares fall below IPO price, with high revenue growth and share supply pressure coexisting

On October 22, MiniMax's Hong Kong shares closed at HK$160.7, falling below the IPO issue price.

During intraday trading, the stock once touched HK$158.9, with an intraday maximum decline of more than 7%. In March this year, MiniMax's share price once hit a historical high of HK$1,330, when the company's market value briefly exceeded HK$400 billion; now its market value is approximately HK$56.3 billion.

In contrast to the share price trend, MiniMax's revenue is growing rapidly. In the disclosed 2026 interim results, the company achieved operating revenue of US$116.6 million in the first half of the year, a year-on-year increase of 283.1%, already exceeding the full-year 2025 revenue scale of US$79 million.

In addition, MiniMax also released M3.1 Flash this month. M3.1 Flash ranked first among domestic models on the AI agent evaluation benchmark ALE, with online inference speed increased by nearly 3 times and the comprehensive usage cost of Agentic tasks reduced by 70%.

But MiniMax not only fell below the issue price; its share price also recorded a maximum decline of 88% from its historical high. What exactly happened?

01. Why did it fall below the issue price?

In January this year, MiniMax listed in Hong Kong, becoming, together with Zhipu, one of the first two large-model companies in China to go public. At the initial stage of listing, MiniMax's total share capital was approximately 314 million shares, with approximately 17.07 million freely tradable shares, accounting for about 5.4% of total share capital.

On July 9, MiniMax welcomed its first large-scale share lock-up expiry after listing, involving approximately 153 million shares, about half of total share capital. On the day of the lock-up expiry, MiniMax's share price fell 17.98% to close at HK$297.4, and its market value fell below HK$100 billion.

In addition to the lock-up expiry, MiniMax also expanded its total share capital through a new share placement. On July 10, MiniMax announced a new round of financing plan, intending to raise funds through placing new shares and issuing convertible bonds. On July 14, the company completed the placement of 35.6 million new Class A ordinary shares at a placement price of HK$268 per share, obtaining net proceeds of approximately HK$9.444 billion. The placed shares accounted for approximately 10.19% of the company's issued share capital after enlargement.

On July 16, MiniMax completed the issuance of convertible bonds with an aggregate principal amount of HK$6.5 billion, obtaining net proceeds of approximately HK$6.433 billion. The bonds will mature in 2027, with an initial conversion price of HK$335. The two financings raised a total of approximately HK$15.877 billion in net proceeds, nearly three times the company's net IPO proceeds in January.

In addition, MiniMax also retained room for further issuance of new shares. At the extraordinary general meeting held on September 15, the company approved a general mandate granting the board the authority to issue additional Class A ordinary shares, with an issuance cap of 20% of the total number of issued shares on the date the resolution was passed. According to the monthly return on changes in share capital disclosed on October 2, in September it added approximately 3.2449 million Class A ordinary shares due to the equity incentive plan; as of the end of September, total share capital had increased to approximately 352 million shares.

This means that half a year after listing, MiniMax not only expanded its total share capital through new share placement, but also faces a potential increase in share supply from future conversion of convertible bonds. In addition, the new share issuance mandate obtained in September also leaves room for the company to further expand its share capital.

Senior artificial intelligence expert Guo Tao believes that MiniMax's fall below the issue price is the result of a combination of capital-market disturbances and corrections to fundamental expectations. In addition to the increase in share supply, he believes that with the continuous increase in global AI investment targets, Hong Kong stock funds' risk appetite for high-loss growth stocks has somewhat declined.

02. Can commercial revenue and AGI be balanced?

MiniMax launched M3.1 Flash on October 19. The new model's online inference speed is nearly 3 times that of the previous generation, the comprehensive usage cost of Agent tasks is reduced by 70%, the API input price is US$0.15 per million tokens, and the output price is US$0.60 per million tokens. This is MiniMax's latest move to continuously reduce model usage costs.

The business logic of model price cuts is clear: by lowering token prices, attract more enterprises and developers to call the model, and drive platform revenue growth through growth in call volume.

MiniMax founder Yan Junjie said in the August results disclosure that as of July 2026, the token consumption scale on the MiniMax platform had reached 20 times that of January of that year. On September 4, Citi released a research report stating that in August MiniMax's ARR had exceeded US$800 million, a significant increase from US$150 million in February. Among this, enterprise customers contributed about 80% of ARR.

MiniMax's 2026 interim report shows that the company's "open platform and other AI enterprise services" revenue reached US$73.9 million, a year-on-year increase of 703.1%, and its proportion of total revenue rose from 30.3% in the same period of 2025 to 63.4%, becoming the company's largest source of revenue.

In its interim report, the company explained that the rapid growth in enterprise service revenue mainly came from an increase in the number of paying individual users and enterprise customers, higher API call volume, and the rapid adoption of Token Plan.

But in Guo Tao's view, the long-term profitability of the low-price API volume model still faces tests. The B-end API track has entered red-ocean competition, and customers won through low prices have low migration costs, making it difficult to maintain long-term payment.

Guo Tao believes that MiniMax cannot rely solely on low-priced APIs to expand its customer base, but should further extend into industry-customized solutions, private deployments, and high-value-added multimodal solutions, using deeper business integration to improve customer stickiness and per-customer value.

At the same time, however, MiniMax's investment in model R&D and computing power has continued to increase.

In the first half of 2026, MiniMax's R&D expenses reached $296.9 million, up 138.8% year-on-year, already exceeding its full-year 2025 R&D spending of approximately $252.8 million. Over the same period, the company achieved gross profit of $20.81 million, but R&D expenses were more than 14 times gross profit.

In its prospectus, MiniMax emphasized obtaining computing resources by leasing hardware and purchasing cloud services to reduce the burden of directly holding large amounts of hardware assets. But in the first half of 2026, MiniMax's cash expenditure on purchases of property, plant and equipment reached $187.6 million, compared with only $383,000 in the same period of 2025. At the same time, MiniMax began emphasizing the construction of an independently controllable core computing power cluster.

But building independent computing power takes time, and MiniMax still needs to purchase cloud computing services.

On August 26, the company announced an increase in the upper limits of its cloud computing service procurement transactions related to Alibaba Cloud. Among them, the annual cap for 2026 was raised from $115 million to $300 million; for 2027, from $125 million to $400 million; and for 2028, from $135 million to $500 million. The combined procurement transaction cap for the next three years reaches $1.2 billion, a marked increase from the previous $375 million.

Guo Tao believes that MiniMax may be forming a hybrid supply model combining proprietary computing power with public cloud resources: the former handles long-term, stable computing demand, while the latter is used to cope with sudden inference traffic and elastic resource demand.

He pointed out that whether this model can reduce costs depends not only on procurement scale, but also on operator optimization, inference frameworks, and hardware utilization. Especially as models continue to upgrade, the computing power consumption per task may also increase, and growth in call volume may not automatically bring down unit costs.

At the same time, MiniMax is still loss-making. Financial reports show that in the first half of 2026, MiniMax's net loss was $358 million, narrowing by 11% from $402.2 million in the same period last year. Adjusted net loss increased from $138.7 million in the same period last year to $293 million, expanding by 111.2% year-on-year.

But for MiniMax, continued investment in computing power and R&D is also a necessary condition for achieving its long-term technological goals. Since its founding in early 2022, Yan Junjie has established achieving AGI as MiniMax's long-term strategy and ultimate goal.

Now, with its Hong Kong stock breaking below its issue price, MiniMax also faces a more practical question: how to find a balance between long-term technological investment and commercial returns.

(The content and views in this article are for reference only and do not constitute investment advice. Investment involves risks, and caution is required when entering the market.)

This article comes from the WeChat public account "Fourth Wave," author: Wang Chen, published by 36Kr with authorization.