--- title: "\"Under the stimulus of 'lobster': In the first two months of the year, MiniMax's annual recurring revenue (ARR) grew by 50%, and the usage of the M2 model Token increased sixfold" type: "News" locale: "en" url: "https://longbridge.com/en/news/278299286.md" description: "Morgan Stanley stated that driven by the \"Lobster Effect\" (OpenClaw ecosystem), the commercialization of MiniMax has accelerated beyond expectations. Its ARR skyrocketed from USD 100 million to USD 150 million in just two months, an increase of over 50%; the token usage of the M2 model surged sixfold from December to February, while the inference cost per token simultaneously dropped by over 50%. The upcoming M3 model aims for top-tier global capabilities, with expectations for even higher gross margins" datetime: "2026-03-09T01:18:34.000Z" locales: - [zh-CN](https://longbridge.com/zh-CN/news/278299286.md) - [en](https://longbridge.com/en/news/278299286.md) - [zh-HK](https://longbridge.com/zh-HK/news/278299286.md) --- > Supported Languages: [简体中文](https://longbridge.com/zh-CN/news/278299286.md) | [繁體中文](https://longbridge.com/zh-HK/news/278299286.md) # "Under the stimulus of 'lobster': In the first two months of the year, MiniMax's annual recurring revenue (ARR) grew by 50%, and the usage of the M2 model Token increased sixfold Author: Dong Jing Source: Hard AI At the beginning of 2026, a lobster named "OpenClaw" became a sensation in the global developer community. In just over a month, driven by the "lobster effect," the large model company MiniMax is rapidly transitioning from the technology validation phase to the scale monetization phase due to its deep integration with the OpenClaw ecosystem. On March 9th, according to Hard AI, Morgan Stanley stated in its latest research report that the commercialization momentum of the Chinese AI unicorn MiniMax far exceeds expectations—**the annual recurring revenue (ARR) skyrocketed from $100 million to $150 million in just two months, an increase of over 50%**. Meanwhile, the usage of M2 model tokens surged **sixfold** in February 2026 compared to December 2025, while the inference cost per token significantly decreased by over 50%. The firm maintains an **Overweight** rating with a target price of **HKD 990**, indicating about a 23% upside from the current stock price. ## ARR Soars 50% in Two Months: Commercialization Enters Accelerated Path MiniMax's commercialization data has caught the market's attention. According to Morgan Stanley's report, the company's ARR rose rapidly from **$100 million in December 2025** to **$150 million in February 2026**, with an increase of over 50% in just two months. Behind this growth is the resonance explosion of multiple business lines within the open platform: > - **Coding Plan** contributed a significant proportion of the open platform's revenue and has achieved **positive gross margins**, while deepening MiniMax's binding relationship with global developers, promoting the widespread adoption of models in multimodal scenarios; > - **OpenClaw, Coding Plan, and other cloud APIs** are the core drivers of the strong growth in token usage from January to February, with these applications heavily utilizing the M2.1 and updated M2.5 models; > - The **ARPU (average revenue per user)** of AI-native products still has multiple levers yet to be unlocked, and the company's current priority strategy is to expand the user base to build momentum for future monetization. It is worth noting that the company has suspended updates for the Hailuo video generation product after v2.3, concentrating all resources on the development of the next-generation architecture (end-to-end output, multimodal input, long video generation), with the v3 series expected to significantly enhance its market competitiveness. ## M2 Token Usage Grows Sixfold: Scale Effects Drive Cost Plummeting The explosive growth in token usage not only confirms strong demand but also directly drives cost optimization on the supply side. The report reveals that **the usage of M2 model tokens increased sixfold in February 2026 compared to December 2025**, while the inference cost per token significantly decreased. Morgan Stanley pointed out that this cost reduction mainly comes from two major drivers: > 1. **The optimization space is far from exhausted**: Theoretically, M2 can still reduce the inference cost per Token by half; the current efficiency improvements come from algorithm optimization rather than hardware price reductions. > 2. **Scale effects are evident**: As the scale of Tokens expands, the load balancing of computing power and the utilization rate continue to improve, further diluting marginal costs. On the computing power procurement side, MiniMax maintains strong bargaining power with computing power suppliers due to its large and rapidly growing demand, ensuring favorable pricing. Additionally, a diversified revenue structure helps hedge against price war risks in specific scenarios. The research report indicates that management holds a clear optimistic attitude towards long-term gross margin improvement, believing that the enhancement of pricing power brought by model capability upgrades, continuous optimization, and scale expansion will jointly support the long-term upward trend of gross margins. ## M3 Model: A Generational Leap from "Cost-First" to "Global Top Tier" MiniMax's judgment on its long-term ceiling is always anchored in the level of model intelligence. The research report shows that the company is about to launch the **M3 series**, aiming directly at **truly global top-tier capabilities**. This stands in stark contrast to the positioning of M2—M2 was optimized around **cost-effectiveness and speed** under resource constraints at the time; whereas M3 represents a **leap in model capabilities** achieved after accumulating sufficient talent, computing power, and data, rather than a gradual catch-up. Management emphasizes that the design of each generation of models—including functional positioning, market scale, and profit margin structure—is planned in advance. **M3 has been clearly set to achieve higher gross margins than existing model deliveries**. According to the research report, in terms of speed, MiniMax believes that inference speed remains a key differentiating competitive advantage, and the company will continue to seek the optimal balance between cost and speed. Furthermore, the optimization results of large language models (LLM) can also feed back into the efficiency improvements of multimodal models, creating a technological synergy effect. ## Redefining Platform Value: Developer Ecosystem as the Core Moat MiniMax's definition of a "platform company" is fundamentally different from traditional internet platforms. Morgan Stanley states that the company clearly indicates that its platform value **does not lie in controlling internet traffic or acting as a consumer entry point**, but in **driving breakthroughs in intelligent boundaries**. In MiniMax's logical framework, the platform value truly emerges when new intelligent capabilities give rise to entirely new product categories. The research report points out that currently, the company's programming products and agent pipelines are showing strong growth momentum, confirming the effectiveness of this logic. In building user stickiness, MiniMax emphasizes that **rapid and continuous model iteration is fundamental**: developers tend to choose platforms with fast model advancement and reliable iterations. Morgan Stanley believes this means that MiniMax's moat is essentially a competition of "model evolution speed"—whoever can maintain technological leadership will be able to lock in the developer ecosystem ### Related Stocks - [MINIMAX-WP (00100.HK)](https://longbridge.com/en/quote/00100.HK.md) ## Related News & Research - [MiniMax’s ARR tops USD 150 million as it pivots toward an AI platform model](https://longbridge.com/en/news/277591914.md) - [MiniMax Group, Inc (0100): New Buy Recommendation for This Technology Giant](https://longbridge.com/en/news/277711815.md) - [I went to ClawCon, where OpenClaw obsessives ate free lobster tails and debated about AI](https://longbridge.com/en/news/278067857.md) - [China's Minimax reports strong revenue growth, charts broader AI ambitions](https://longbridge.com/en/news/277484992.md) - [Bank of Canada tests tokenized bonds on blockchain in Project Samara](https://longbridge.com/en/news/278099263.md)