Summary of Key Points
MiniMax’s revenue in the first half of the year increased by 283% year-on-year, with B-side revenue accounting for a staggering 80% of total revenue. However, its market value evaporated by 76% within just 5 months (from HK$43 billion to HK$10.98 billion). The reasons for this are as follows:
- The revenue growth appears impressive, but it is actually due to a “false prosperity” (low base and methodological issues).
- Although profits have seemingly narrowed, the actual operating losses have doubled.
- The company’s models lack competitiveness (outdated text models and difficulties in monetizing multiple modalities).
- Investors have lost confidence in the company’s strategy due to its inability to leverage all modalities effectively.
1. High Revenue Growth? Beware of the “False Prosperity” Trap
MiniMax’s B-side revenue increased by 703% year-on-year, rising from 30% to 80% of total revenue. However, upon closer inspection:
- Low Base Growth: Last year, the B-side revenue was only $9.2 million; this year it reached $73.9 million. The high growth rate is due to the low starting point (it’s like earning $100 last year and $800 this year, a 700% increase, but the total amount is still small).
- Misleading Calculation of ARR (Annual Recurring Revenue): The industry typically calculates ARR as “average monthly revenue × 12,” but MiniMax uses “weekly revenue × 52.” If there is a surge in traffic during a particular week, this figure can be inflated (for example, earning $1 million in one week results in $52 million for the year, but only $30 million if calculated monthly).
- Token Consumption Increase: Token consumption in July was 20 times that of January, not due to a sudden surge in demand. Instead, it was because the M3 model switched from a pay-per-use model to a token-based model, and prices were reduced by 50%, prompting users to use more tokens (similar to using more data after a mobile data plan discount).
These factors make the revenue growth look good, but the quality of the growth is not substantial.
2. Profit Data Shows a Twisted Picture: Narrowing on the Surface, but Actual Losses Are Worse
There is a stark contrast between the two figures for losses in the financial report:
- Surface: Net Losses Narrowed by 11%: From HK$402 million last year to HK$358 million this year, which seems like an improvement.
- Reality: Adjusted Losses Increased by 111%: Last year, the core operating loss was HK$139 million; this year, it is HK$293 million (doubling).
The reason for this is the one-time impairment of financial liabilities amounting to HK$254 million last year, which reduced the overall loss.
Even worse:
- Excessive Spending on R&D: R&D expenses in the first half of the year were HK$297 million (2.5 times the revenue), meaning for every dollar earned, HK$2.5 was invested in research and development (like running a milk tea shop where you earn $5 from each cup but spend $12.5 on ingredients and equipment).
- Gross Margin Halved: The gross margin dropped from 33.5% in the second half of last year to 17.8% in the first half of this year. This is due to price wars in the B-side market (competitors lowering prices to attract customers), and MiniMax had to follow suit, as well as the need to compensate users with tokens after the price adjustment for the M3 model.
3. Model Competitiveness: Falling Behind in Text Modeling, Difficulty in Video Monetization
MiniMax’s model advantages are fading:
- Text Model Lagging: The parameters of its flagship M3 model are only HK$42.8 billion (compared to Zhipu GLM-5.2’s HK$74.4 billion and DeepSeek V4’s HK$1.6 trillion), indicating weaker capabilities for handling complex tasks like coding. Coding is a key area for large model commercialization, as companies are willing to pay for advanced coding capabilities. Additionally, the M3 model is not being updated as quickly as Zhipu’s GLM5.3.
- H3 Model’s Limited Success: The H3 video model, although well-received, has struggled with monetization. Although its open-source version was downloaded 24 million times in three weeks, it is difficult to generate revenue. ByteDance’s Seedance 2.5 was launched around the same time, and with ByteDance’s large platform, MiniMax faces significant competition (it’s like having a great filter but no social media platform to promote it, so users won’t generate much revenue).
4. Why Did Investors “Vote with Their Feet”? Strategic Uncertainty Is the Key Reason
The sharp decline in MiniMax’s market value reflects investors’ lack of confidence in its strategy:
- Distributed Resources with a Multi-Modal Approach: The company initially focused on both text and video modalities, but as coding became the core area for commercialization, the text model failed to keep up, and the video model did not gain a significant market share.
- Misguided Pricing: In March, MiniMax suddenly changed the pricing model for the Coding Plan from unlimited to a token-based system, causing user dissatisfaction. Later, prices were reduced by 50%, which not only failed to generate additional revenue but also weakened the company’s pricing power and reduced the gross margin.
Investors seek certainty: either the text model needs to be improved to a top level or the video model must be successfully monetized. Since MiniMax has not achieved success in either area, investors have lost confidence, leading to a significant drop in its market value.
5. Overcoming Challenges to Regain Market Confidence
To regain market confidence, MiniMax must address the following issues:
1. Improving Coding Capabilities: Can M3.1 enhance coding capabilities? Can M3 Pro reach a trillion parameters?
2. Breaking Through in Video Monetization: Open-source success is not enough; a stable monetization strategy is needed (e.g., collaborating with companies for video generation services).
3. Stabilizing the Gross Margin: By optimizing the model architecture and reducing computing costs, the gross margin should be restored to a reasonable level (at least stopping the halving).
4. Focusing on a Clear Strategy: The company needs to decide on a focal area—either improving the text model or making the video model a hit product—and stick to it.
If these issues are not resolved, MiniMax may indeed fall out of the top tier of companies in this industry.
(The entire analysis is presented in plain language to make it understandable to non-experts, revealing both the company’s successes and underlying concerns.)