Summary of Key Points
In MiniMax’s first half-year financial report since going public, the most notable aspects are not the revenue of $117 million (which exceeds the annual target for 2025), but rather a significant shift in revenue structure (B2B accounting for 63.4% compared to B2C), continued high research and development (R&D) expenses (2.5 times the revenue), an improvement in gross margin (from 12.1% to 17.9%), and a majority of revenue coming from global markets. The most critical point, however, is that the company has not yet turned a profit. MiniMax has transitioned from selling C2C products (such as “Hailuo AI”) to providing model capabilities to businesses and developers. However, to prove that this business model is sustainable, they still need to overcome one final hurdle.
Detailed Analysis
1. The Shift from Selling Products to Selling Model Capabilities
MiniMax initially gained popularity with its C2C product “Hailuo AI” (for text, voice, and video generation), which accounted for nearly 70% of revenue last year. In this year’s half-year report, however, B2B revenue (from open platforms and enterprise services) amounted to $73.93 million, accounting for 63.4%, while C2C revenue dropped to $42.64 million, accounting for 36.6%.
- C2C focuses on attracting users to use the product: Users purchase specific functions (e.g., image generation), which are either one-time purchases or subscription-based, with limited potential for growth.
- B2B focuses on providing model capabilities: Businesses purchase the ability to integrate these models into their systems (e.g., for customer service or code generation), and developers are charged each time the model is used or tokens (the unit of text processed by the model) are consumed. This model-based approach generates continuous revenue with higher potential for growth.
In simple terms, MiniMax has shifted from being an “AI product company” to a “model capability provider.”
2. Spending More Than You Earn: Why Is There a Loss?
The most striking figure in the financial report is the R&D expenditure of $297 million (2.55 times the revenue), resulting in a net loss of $29.3 million (a 111% increase year-over-year). Why is there still a loss despite significant sales?
- The cost of large models differs from traditional internet models: For traditional apps, additional users generally reduce costs (e.g., adding a user to WeChat is almost cost-free). However, large models require more computing power for training and inference, leading to higher costs with each new customer and API call. As revenue increases, so do expenses.
- The focus is now on whether customer growth can outpace cost increases: MiniMax has shown that there is demand, but the challenge is to reduce costs per dollar earned.
3. Improving Gross Margin: A Sign of Profit Potential
The good news is that the gross margin has increased from $3.69 million to $20.81 million, rising from 12.1% to 17.9%. This indicates that revenue growth has outpaced cost increases. Although the margin is still lower than that of mature software companies (e.g., Microsoft’s margin can exceed 60%), the direction is positive, suggesting that MiniMax is improving efficiency—e.g., by using less GPU for model training, faster inference, and more efficient token usage.
This is a crucial step from demonstrating market demand to achieving economies of scale. In the past, having a strong model and users was enough; now, the goal is to deliver the same level of intelligence at a lower cost. For example, if it used to cost $1 to generate an article, it might now cost only 50 cents—this is the foundation for long-term profitability.
4. Over 60% of Revenue Comes from Overseas
MiniMax’s globalization is a highlight, with $70.83 million in overseas revenue (60.8%) from more than 230 countries. This is rare among domestic large-model companies.
- AI globalization is easier than traditional internet businesses: Traditional companies need to establish channels, local teams, and supply chains (e.g., logistics for e-commerce). AI companies can simply provide their models through APIs, allowing global distribution with minimal additional costs.
- Challenges include: Exchange rate fluctuations, compliance with international regulations (e.g., EU AI laws), and differences in local market demands. However, MiniMax’s revenue structure indicates a viable path for “Chinese technology + global market,” offering greater potential than focusing solely on the domestic market.
5. The Final Hurdle: Sustainability
MiniMax has three key advantages: a strong model, a demand for its products, and willing customers. However, it still needs to turn a profit. Market expectations have changed since the company’s initial public offering; investors now focus on whether the business model is profitable, not just the potential of the AI industry.
- The core question is: How much profit can be retained for each dollar earned? With a current gross margin of 17.9%, for every $100 in revenue, 82 cents are spent on costs, indicating a long way to go before profitability. However, if revenue growth continues to outpace cost increases, the gross margin will eventually rise, leading to profitability.
This represents a critical challenge for the entire large-model industry: the transition from having products sold to generating sufficient profit through efficiency improvements, cost control, and a mature business model.
In Summary
MiniMax has transformed from an AI company that sold C2C products to a provider of model capabilities, demonstrating market demand. To be truly successful, it needs to reduce its high R&D expenses and increase profit margins. This is not only a challenge for MiniMax but also a common issue for the entire large-model industry.