The stock market journey of MiniMax has been a rollercoaster over the past six months, with its valuation plummeting from a peak of over 400 billion Hong Kong dollars to below 100 billion.

As of the close on July 21st, shares of MiniMax were trading at 222.4 Hong Kong dollars, giving the company a total market capitalization of approximately 776.7 billion Hong Kong dollars. The company debuted on the Hong Kong Stock Exchange half a year earlier, on January 9th, with its share price surging 109% on the first day. The stock reached an all-time high of 1,330 Hong Kong dollars on March 18th, pushing its market cap to around 410 billion Hong Kong dollars and briefly surpassing that of Baidu.

This peak was short-lived, however, as the share price subsequently reversed its upward trajectory. By the July 21st close, the company's total market value had fallen below the one trillion Hong Kong dollar mark, representing a decline of over 300 billion Hong Kong dollars.

In the Hong Kong market, MiniMax and KNOWLEDGE ATLAS are often referred to as the "twin stars" of large language models. Interestingly, while MiniMax's stock began to decline, KNOWLEDGE ATLAS entered a period of significant growth. In June, KNOWLEDGE ATLAS's share price reached 2,980 Hong Kong dollars, giving it a market capitalization exceeding one trillion Hong Kong dollars. It has also retreated significantly since then. As of July 21st, KNOWLEDGE ATLAS had a market cap of about 567.6 billion Hong Kong dollars, more than seven times that of MiniMax.

Short-Term Pressures and Weakening Scarcity

The recent weakness in MiniMax's share price is directly linked to factors such as the expiry of share lock-up periods and a reduction in its perceived scarcity.

Both MiniMax and KNOWLEDGE ATLAS faced share lock-up expiries on July 8th and 9th. The proportion of shares becoming eligible for sale was significantly higher for MiniMax, with approximately 46.44% of its total shares being released, compared to about 5.76% for KNOWLEDGE ATLAS. The substantial increase in available shares for MiniMax led to selling pressure and stock price volatility, with the stock falling 17% on the day the lock-up expired.

While the stock price reflects short-term sentiment, the deeper, long-term challenge for MiniMax is whether it can successfully establish a closed loop involving its model capabilities, product offerings, and commercial monetization.

From AI Star to Reality Check

Having seen over 300 billion Hong Kong dollars in market value evaporate from its peak above 400 billion to its current sub-one trillion level, the company is facing challenges in the capital markets.

Objectively speaking, MiniMax is not currently in a crisis but is experiencing short-term stock price pressure. The signal from the secondary market is clear: the company is undergoing a process of valuation "de-glamorization," transitioning from being a highly touted AI large model star to a more typical listed technology company.

The immediate catalyst was the significant selling pressure following the lock-up expiry. On July 9th, approximately 145.7 million shares of MiniMax were released for trading, representing 46.44% of its total share capital. Currently, MiniMax is not included in the Southbound Stock Connect scheme, limiting inflows from mainland Chinese investors. This lack of stable capital inflow made it difficult to effectively counteract the selling pressure, meaning even modest sell-offs could cause price swings. On that day, its stock price plunged 17.98%, with trading volume surging to over 6.8 billion Hong Kong dollars.

The stock continued to fall sharply for the next two days. In response to the declining share price and weakening market confidence, MiniMax acted swiftly to stabilize the situation.

On July 10th, MiniMax announced a placement of 35.6 million new Class A shares and a concurrent issuance of zero-coupon guaranteed convertible bonds due in 2027. These transactions raised a total of approximately 16.04 billion Hong Kong dollars (net proceeds around 15.96 billion). The company stated it would allocate 80% of the raised capital towards building AI infrastructure and the iterative research and development of its large models.

Notably, on the same day, founder Yan Junjie issued an internal letter announcing he would no longer receive a salary and would instead allocate 5% of his personal shareholding in the company for team incentives and open-source community support.

These two measures by MiniMax did not appear to stabilize the stock price for long, as the overall downtrend continued.

If the short-term rise in caution was the catalyst for the recent drop, the peak in the stock price around mid-to-late March was the result of a shift in the underlying investment thesis.

At its IPO in January, the market awarded MiniMax a high valuation premium due to the scarcity of pure-play AI model companies, its impressive portfolio of consumer-facing overseas products like Talkie, Starry, and Conch AI, and the hot narrative around multimodal large models.

However, the AI industry evolves rapidly. With developments like the reported shutdown of OpenAI's Sora project and commercial hurdles for video models like Runway, the multimodal narrative faced greater scrutiny. Subsequently, the AI trend shifted towards "Agents" and "AI Coding."

KNOWLEDGE ATLAS, with its strengths in business-to-business services, demonstrated a faster pace of commercial deployment. Its model, GLM-5.2, ranks among the top globally among publicly available models, even surpassing flagship products from OpenAI and Anthropic. KNOWLEDGE ATLAS has become a strong contender as an enterprise AI infrastructure provider. In contrast, MiniMax focuses on full-modality coverage. After launching its flagship model MiniMax-M3, the company unexpectedly cut its API prices by 50% following a quiet price hike, reflecting a degree of "passivity" in its pricing power and commercial strategy. Some analysts subsequently lowered their target price for MiniMax, citing concerns that aggressive price cuts would severely erode its profit potential.

Financially, the consumer segment is MiniMax's mainstay, contributing 67% of its revenue in 2025. Media reports indicate that monthly active users for its Talkie and Starry products declined by 60% quarter-on-quarter in Q4 2025, signaling a drop in user metrics for its consumer business.

Furthermore, internet giants like ByteDance and Alibaba have also made significant investments in multimodal AI, intensifying competition in a field where products are becoming increasingly similar. This puts greater pressure on MiniMax in terms of user retention and conversion to paid services.

For a company like MiniMax, which has been public for less than a year, the initial valuation was largely predicated on the scarcity of pure-play large model investment opportunities. Now, with news that DeepSeek is preparing for an IPO and the potential for more AI firms to go public, the pool of investable large model companies is expanding. This increases competitive pressure and is beginning to erode the scarcity premium that MiniMax once enjoyed.

The Aspiration to Become an AI Platform Company

From its inception, MiniMax chose to pursue a multimodal technology path, aiming to build a full-modal model capable of simultaneously understanding text, images, audio, and video.

In its early startup phase, pursuing multiple technological fronts in parallel was considered a "risky" strategy due to increased technical complexity and higher R&D costs. Before its models were fully proven and commercialization was achieved, the company faced skepticism.

However, the successful commercialization of its AI products and expansion into overseas markets laid the foundation for its eventual public listing.

In 2025 and prior, MiniMax was primarily an AI large model developer. Entering 2026, the company set a new goal: to become an AI platform company.

MiniMax defines an "AI platform company" as one that can define and push the boundaries of intelligence. Each time a new intelligence boundary is broken, it creates numerous new scenarios, attracts new customers and users, forms new ecosystems, and generates fresh commercial opportunities. The company has proposed a formula: Platform Value = Intelligence Density * Token Throughput.

If the internet era emphasized the flow economy, the AI era emphasizes the token economy. This implies that as the capabilities in both intelligence density and token throughput increase, the value of the platform becomes more pronounced.

MiniMax aims to demonstrate its platform value through technological iteration, a dual-engine strategy targeting both consumer and enterprise markets, and building a hardware and software ecosystem. In terms of business model, MiniMax positions itself as a global player with a "C+B" model. The consumer side targets individuals with products like Conch AI and Starry, monetized through paid subscriptions and advertising. The enterprise side serves businesses and developers through API services with usage-based billing.

In summary, MiniMax aims to serve both consumer and enterprise markets. Its commercial logic is to use its models to create products, commercialize those products globally to acquire data and users, and then use that data to fuel further model iteration, creating a virtuous cycle between these three elements.

The critical question is whether this virtuous cycle can truly be closed.

On the technology front, MiniMax showcased its full-modal models and products, including M3 and H3, at the 2026 World Artificial Intelligence Conference, with its new-generation multimodal generative model H3 attracting significant attention.

Additionally, MiniMax recently announced the update to MiniMax Code 2.0 for desktop, focusing on improvements in conversation startup speed, stability for long-running tasks, and context continuity for tool calls. The new version also optimized features like file preview and chart interaction and plans to further expand financial research capabilities.

Regarding its hardware and software ecosystem, in the robotics field, MiniMax is collaborating with Zhiyuan Robot, providing full-process AI text-to-speech support. Its M3 model also powers the Vbot super-powered robot dog "Big Head BoBo." In consumer electronics, Rokid's AI glasses have integrated MiniMax's voice capabilities. AI toys for children like Lingji Tianci's Jollybubu and the KATA Friends companion robot utilize MiniMax's voice model to support multi-character, multi-language, and low-latency dialogues.

In its transition from a model developer to a platform company, MiniMax cites its industry-leading model iteration speed and intelligent R&D capabilities, its integrated model-and-product development approach, and its broad global ecosystem of developers and users as key competitive advantages.

However, MiniMax still relies heavily on its consumer business. For a user, switching from Conch AI to a competitor like Kimi is as simple as tapping a different app icon, with virtually no switching or data migration costs. This highlights the difficulty in building a strong moat for consumer AI applications. The company needs to demonstrate that its products and advantages can cultivate strong user loyalty, retain customers, and generate data traffic substantial enough to genuinely support its model development.

Simultaneously, MiniMax remains in a phase of "high growth and high investment." As of the end of 2025, the company reported revenue of 555.5 million yuan, a year-on-year increase of 158.95%. During the same period, its R&D expenses reached 1.777 billion yuan, and it reported a net loss attributable to shareholders of 13.16 billion yuan.

In the past, the secondary market focused on metrics like model capability improvements and user growth when evaluating large model companies. In the current environment, surrounded by AI unicorns and with internet giants doubling down on AI, MiniMax's journey towards becoming an AI platform company will also require a comprehensive assessment of the sustainability of its business model, the strength of its core competitive advantages, and its ability to improve profitability.

Whether MiniMax can ascend to the next level and reclaim a market valuation in the hundreds of billions remains a key point of observation for the market.