Another case of AI devouring software? Google's Nano Banana goes viral, Meitu’s stock price plunges
```
The explosive popularity of Google's new AI model Nano Banana has triggered deep market concerns over whether "AI will devour application software." However, Morgan Stanley analysts believe these concerns are excessive.
On September 2, a research report from Morgan Stanley pointed out that Meitu's growth trajectory has not been affected by the AI model Nano Banana, and its true value lies in providing "last mile" solutions beyond the reach of foundational AI models.
The report emphasizes that Meitu has built a solid moat with its deep accumulation in vertical sectors, proprietary data, and focus on core paid features. In addition, the company remains open and neutral toward all AI models and has seamlessly integrated new technologies like Nano Banana into its products.
Morgan Stanley reiterates its "Overweight" rating on Meitu, with a target price of HKD 15.70, believing its long-term value remains solid. Revenues are expected to reach 4.021 billion RMB in 2025, growing to 5.12 billion in 2026, and further to 6.771 billion RMB in 2027.
Market Panic Triggered by Nano Banana
Recently, Google's new image AI model "Nano Banana" quickly gained popularity on Chinese social media.
The model was released on August 26, and with its powerful ability to convert photos into 3D model images through prompts, it set off a storm of discussion on August 30.
This event immediately triggered investor concerns—continually improving foundational AI models might cross over and "erode" the core business of application software like Meitu.
Under this panic sentiment, Meitu's stock price plunged 14% on the trading day when the news fermented, while the Hang Seng Index rose 2% during the same period.
It is worth noting that this is not an isolated case.
Since the beginning of this year, whenever heavyweight AI models such as GPT-4o or Google Veo3 have released updates, Meitu's stock price has experienced similar volatility.
This reflects the market's widespread anxiety about the survival model of application-layer software in the AI era.
"Last Mile" Value Is Hard to Replace
Morgan Stanley analysts emphasize that in the AI era, the value of application software lies in providing "last mile" services to optimize outcomes, which is precisely what general-purpose AI models cannot fully achieve.
The "last mile" in the visual industry is particularly complex, due to two reasons:
Fragmentation of scenarios and diversity of demand: Users have vastly different needs, ranging from personal photo beautification, ad image generation, to poster design—each scenario has its own standards and workflows.Subjectivity of personal preferences: Aesthetics and visual preferences are highly personalized, which are difficult to fully capture via standardized prompts and cannot be perfectly accomplished by general-purpose AI models.
It is precisely these two characteristics that make it difficult for general-purpose AI models to provide a perfect end-to-end product to meet all user needs.
Therefore, application software that focuses on specific scenarios and optimizes user experience not only retains its value, but it actually becomes more prominent.
Meitu's Advantages: Scenarios, Data, and Core Paid Features
Morgan Stanley believes that through its long-term accumulation, Meitu has established significant advantages in maximizing "last mile" value.
First, the company is focused in its business and has amassed deep scenario insights. Meitu focuses on non-professional casual and non-professional productivity scenarios, further refining the latter to e-commerce design and other fields.
This focus enables it to deeply understand user pain points in specific scenarios and to develop highly optimized workflows that enhance user experience.
Next, massive vertical data builds a barrier. Through partnerships such as with Alibaba and by operating its own products long-term, Meitu has accumulated vast, high-quality vertical data, especially in portrait beautification and e-commerce design.
Meitu uses this proprietary data for "post-training" on third-party foundational models, creating small models with results far exceeding general models, and outcomes that more closely match user needs.
Finally, a clear business model: users pay for "core features." Meitu's features can be divided into "fun features" (like AI emoji generation) and "core features" (like AI face retouching).
The former may attract users to try the product, but the latter is what fundamentally drives paid subscriptions. General-purpose AI models can imitate the former (such as Nano Banana's 3D models), but it's hard to penetrate into the high-quality “core features” Meitu offers.
In addition, Meitu adopts an AI model neutrality strategy, using AI model container technology to rapidly integrate the capabilities of different AI models into its products.
Prior to the widespread dissemination of 3D model images, Meitu had already integrated Google Nano Banana into its overseas products, including Meitu Xiuxiu, BeautyPlus, and RoboNeo.
This article is from WeChat Official Account "Hard AI". For more frontier AI news, please go here.
Risk Warning and DisclaimerThe market involves risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of any particular user. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Investments made according to this article are at your own risk. ```