By Yang Jianyong

The realm of information technology has advanced at a breathtaking pace, with each technological leap giving rise to remarkable companies. Since OpenAI introduced ChatGPT, it has not only ushered artificial intelligence into the era of generative AI but also emerged as the world's most dazzling AI large-model unicorn, with its valuation projected to soar beyond $1 trillion.

China has also witnessed the emergence of numerous large AI model companies, with DeepSeek standing out as a prime example. Thanks to its low-cost, open-source strategy, and exceptional performance, DeepSeek has facilitated widespread adoption across diverse industries, expediting the commercialization of AI and the digital transformation of numerous sectors.

From a $410 Billion Peak to $72.1 Billion: MiniMax's Valuation Reassessment

Meanwhile, MiniMax and Zhipu stand as the only two large-model companies listed on the Hong Kong stock market, drawing significant market attention and pursuit. In particular, the popularity of OpenClaw earlier this year catapulted MiniMax to unprecedented heights, reaching a peak price of HK$1,330, marking a cumulative increase of 706% from its issue price of HK$165.

With its market value surging sevenfold, MiniMax once surpassed the HK$410 billion milestone, establishing itself as a prominent AI large-model company. However, this meteoric rise also sparked concerns about a potential bubble, given that such a colossal market capitalization far outstripped its revenue capacity.

Once the market's darling, MiniMax has since lost some of its luster, with its current market value standing at a mere HK$72.1 billion—a staggering HK$340 billion plunge from its peak, representing an 83% correction. This decline stems from a combination of share lock-up expiration pressures and a market shift from euphoria to bubble-squeezing in valuing large-model companies, prompting a reevaluation of their commercialization and profitability prospects.

Currently, MiniMax's revenue predominantly hinges on AI-native products and open platforms. During the commercialization journey, heightened user willingness to pay for products and the continuous promotion of offerings like Hailuo A have fueled robust revenue growth. By 2025, revenue had reached $79 million, marking a year-on-year increase of 158.9%.

With the rapid evolution of AI large-model technology, the commercialization of large models is gaining momentum. Over 70% of the revenue originates from international markets, serving more than 236 million users across over 200 countries and regions, as well as 214,000 enterprise clients and developers from over 100 countries and regions. This underscores the vast potential for commercializing AI large-model technology.

The AI Computing Power Race Drives Up Costs, Leaving Large Model Companies Mired in Losses

It's crucial to note that the competition underlying AI large models is essentially a battle for computing power in AI infrastructure—a quintessential technology- and capital-intensive industry heavily reliant on AI infrastructure, necessitating substantial computational resources for model training.

OpenAI alone is anticipated to invest $600 billion in computing power by 2030. Meanwhile, the AI expenditures of Microsoft, Amazon, Meta, and Google are projected to reach $5.3 trillion by the end of 2030 to bolster their competitiveness in the AI market.

For large-model startups lacking the financial might of tech giants to construct extensive AI infrastructure, they resort to procuring computing power services from cloud providers like Alibaba and Tencent. According to MiniMax's previously disclosed related-party transactions, the annual service caps with Alibaba Cloud for 2026, 2027, and 2028 are set at $115 million, $125 million, and $135 million, respectively.

From research and development to AI computing power, costs for large-model companies have skyrocketed. MINIMAX accumulated losses of $2.679 billion (approximately RMB 18.1 billion) from 2022 to 2025. In 2025 alone, losses exceeded RMB 12.8 billion, with adjusted losses also reaching RMB 1.7 billion.

Despite the buzz surrounding AI large models, companies selling them have yet to turn a profit. The primary culprit is the hefty AI computing power investment, making short-term profitability elusive and undoubtedly contributing to the formation of an AI bubble. A rational perspective on the long-term development potential of AI large models is imperative.

Notably, in response to significant market fluctuations, the founder of MINIMAX recently issued an internal letter addressing recent market volatility, stating, "The market will fluctuate, and there will be external noise, but the direction of progress remains unchanged."

In Conclusion

The generative AI wave is sweeping across the globe, with various sectors actively embracing the intelligent transformation ushered in by generative AI, gradually unlocking market demand potential. However, in the large-model arena, ideals are abundant, but reality is unforgiving, with application deployment being the linchpin. Only by eschewing hype and focusing on robust deployment can companies navigate through cycles and ultimately achieve long-term value through technology-driven differentiation and commercial sustainability.

Yang Jianyong, a Forbes China contributor, expresses views that are solely his own. He is dedicated to in-depth analysis of cutting-edge technologies such as AI large models, artificial intelligence, the Internet of Things, cloud computing, and smart home appliances.