Why did MiniMax significantly retreat from its historical highs? An in-depth analysis of the stock price correction logic and prospects

In June 2026, the AI large model sector is undergoing a fierce valuation recalibration. Once regarded as one of the "AI Twin Stars" in Hong Kong stocks, MiniMax has experienced an especially sharp decline amid this industry-wide correction.

According to data from the Gate stock trading page, as of June 15, 2026, MiniMax's current quote is $432, up 1.7%. In comparison, its all-time high in March this year was $1,330. What market consensus is reflected in the valuation range after a significant retreat from the high? Can the company's fundamentals support a rebound in stock price?

How pricing strategy fluctuations reflect the competitive landscape

A research report released by JPMorgan in mid-June shows that MiniMax's rating was downgraded from "Overweight" to "Neutral," and its target price was sharply cut from HKD 1,100 to HKD 400, a decrease of about 64%. The key basis for the report is not the slowdown in revenue growth—in fact, the bank raised its revenue forecasts for MiniMax for 2026-2027 by 34%-74%—but points to a deeper issue: pricing power.

The direct reason triggering this rating adjustment was a pricing move by MiniMax. On June 1, this year, the company announced its flagship model M3, with an initial price approximately double that of its predecessor, M2.7. However, just about a week later, MiniMax announced a permanent 50% price cut for M3, bringing it back to a level close to M2.7. JPMorgan's analysis suggests that, in an environment where AI demand still exceeds inference supply, no developer is willing to voluntarily give up premium pricing. The rapid price decline from the high suggests that the actual improvement in M3's intelligence level did not fully match the initial pricing.

Contrasting this view is the performance of Zhipu AI. The company's API prices have doubled this year, while sales volume has continued to grow. This difference points to a core proposition: as industry scenarios converge toward API, code generation, agents, and enterprise workflows, model capability leadership becomes increasingly critical, and pricing power will shift more from product coverage breadth to capability itself.

How unlocking pressure affects market re-pricing of valuations

Market expectations for the unlocking of restricted shares is another key factor suppressing MiniMax's stock price. According to public information, MiniMax will experience its first large-scale unlocking of restricted shares on July 9, 2026, involving about 146 million shares. At that time, approximately 46.44% of shares will transition from lock-up to tradable, causing supply to increase nearly tenfold compared to the current free float.

Among investors holding such shares, over one-third are financial investors. These investors have clear exit cycles and have already realized substantial gains early on, making profit-taking during the unlocking window a strong motivation.

CICC's analysis further highlights the structural differences in unlocking pressure: although both AI large model companies will face unlocking, Zhipu's unlocked shareholding ratio is about 11.6%, mainly held by state-owned investors, with higher stability; whereas MiniMax's unlocked shares constitute a higher proportion of Hong Kong stock capital, with a larger share held by financial investors, leading to a more direct impact on supply.

Additionally, the scarcity of listed global large model targets is gradually diminishing. OpenAI and Anthropic have secretly filed IPO applications, and Starry Star is expected to submit a Hong Kong listing prospectus soon. Moon Shadow recently launched a new round of financing. As more comparable targets enter the capital markets, the valuation premium of existing AI stocks faces systemic compression pressure.

Is the product pricing adjustment turmoil a short-term event or a structural risk?

On June 1, this year, MiniMax announced the launch of the M3 model and simultaneously shifted its billing method from per-use to per-token. The minimum package price increased from 29 yuan to 49 yuan, without prior notice to users. Some developers found that, under similar tasks, actual token consumption far exceeded expectations, with costs increasing by up to 257%. One user calculated that, previously, consuming 3 to 5 billion tokens per month cost only 49 yuan; after the change, the same usage now costs about 175 yuan.

This controversy quickly escalated into a trust crisis. The company issued an apology late that evening, admitting that "this adjustment was not sufficiently communicated in advance," and introduced compensation measures such as retaining unlimited quotas for old users.

Technically, per-token billing is standard industry practice, adopted by global peers like Anthropic and OpenAI. But the key issue lies in execution: changing billing rules suddenly without sufficient buffer time or prior explanation, especially in a developer market highly sensitive to price, caused significant short-term shocks.

A deeper issue is that this price hike reflects a certain structural tension in the company's business model. Over 70% of revenue comes from overseas C-end products (mainly virtual companionship app Talkie), with a gross margin of only 4.7%. Given the high investment in model R&D, profitability in this segment is limited. As computing costs continue to rise and the original low-price strategy becomes unsustainable, price increases become almost inevitable. How to raise prices and maintain user trust will directly influence market perceptions of the company's platform transformation capability.

Does industry capital flow still support AI large model company valuations?

From a macro perspective, the overall capital landscape of the AI sector has not fundamentally reversed. As of June 2, 2026, southbound funds have net bought approximately HKD 280.8 billion year-to-date, mainly flowing into two sectors: high-dividend defensive stocks and AI technology growth sectors. Globally, major cloud providers like Microsoft and Google have planned over $700 billion in AI capital expenditure in 2026, with no change in pace. The World Semiconductor Trade Statistics (WSTS) forecasts that the global semiconductor market will increase by about 90% year-over-year in 2026, surpassing $1.5 trillion.

However, the direction of capital is also changing. As industry scenarios become clearer, investment logic is shifting from hype to performance certainty. Investors are avoiding crowded sectors and focusing more on valuations that are reasonable and growth paths that are clear. This shift increases valuation compression risks for companies lacking pricing power.

Additionally, Bank of America Securities initiated coverage of MiniMax on June 15, giving a "Buy" rating and a target price of HKD 500. The bank believes that as the performance gap between leading models narrows, sustainable differentiation will be driven more by computing power, product distribution, and commercialization execution, rather than marginal outperformance based solely on benchmark tests. MiniMax has certain advantages in these areas. The bank projects that MiniMax's revenue will grow from $79 million in 2025 to $2.8 billion in 2028.

Can fundamentals support a long-term upward shift in valuation?

The key to whether MiniMax's valuation can rebound lies in the pace and direction of fundamental evolution. In 2025, the company's total revenue was $146M, up 158.9% year-over-year, with over 70% coming from international markets. Gross profit surged from $3.738 million in 2024 to $15k, with gross margin rising from 12.2% to 25.4%. By the end of 2025, the company had served over 236 million users across more than 200 countries and regions, with 214k enterprise clients and developers.

In 2026, growth continues. As of February, the company's ARR (annual recurring revenue) exceeded $150 million; daily token consumption of the M2 series text models increased over six times compared to December 2025, with token consumption from Coding Plan increasing over tenfold. Cost structure optimization shows "reducing marketing, increasing R&D"—sales and distribution expenses decreased by 40.3% YoY, while R&D expenses increased by 33.8%, though at a much lower rate than revenue growth.

However, in terms of profitability, the company reported an annual loss of $1.87 billion, with about $1.6 billion attributable to "financial liabilities fair value losses" accounting factors. Adjusted net loss remains at $250 million. The profitability inflection point still requires time and ongoing model efficiency improvements.

Looking at upcoming catalysts, Yuanta Securities summarized that the launch of the Hailuo new model on June 26, the expiration of the IPO lock-up on July 8, potential inclusion in the Hong Kong Stock Connect on August 6, and the possible release of the full M3 version in the second half of the year are key time points. These events will influence market sentiment and valuation trends in a phased manner.