MEITU recently released its 2025 performance report, which should serve as strong evidence of the company's deepening AI imaging business. After years of dedicated effort and empowerment through AI technology, MEITU has achieved sustained rapid growth. In 2025, total revenue increased by 28.8% year-over-year to 3.86 billion yuan, while adjusted net profit attributable to parent company owners rose 64.7% to 965 million yuan. The subscription business continued its steady climb, with paying subscribers reaching a record high of 16.91 million, representing a subscription penetration rate of 6.1%. Productivity tools are also beginning to emerge as a new engine for the company's growth.
In stark contrast, as the wave of large AI models sweeps across the tech industry, debates about "large models吞噬 applications" have intensified. MEITU's stock price has experienced a significant pullback from last year's highs. In response to the volatility, MEITU not only initiated a 300 million Hong Kong dollar share repurchase plan but also announced that starting in 2026, it will provide quarterly updates on core business operational metrics.
However, skepticism in the capital market persists. Against the backdrop of rapid iteration in large models, can MEITU, a company focused on vertical imaging applications, maintain its high growth rate? What is MEITU's positioning for itself in future industry competition?
In an exclusive dialogue, MEITU's Chief Financial Officer Yan Jinliang stated directly, "The company's stock price is undervalued, and the buyback is our most direct statement." He further presented a core viewpoint: "General large models and applications occupy different ecological niches, and personalization is difficult to achieve, which leaves significant opportunities for AI applications."
He elaborated that demand in the visual field is highly subjective, and a single model cannot meet the personalized needs of all users. Furthermore, large model companies, constrained by their business models, are unlikely to delve deeply into all vertical scenarios for精细化布局.
Based on this assessment, Yan Jinliang dissected MEITU's core moat and future growth logic from multiple dimensions including user demand, organizational culture, and global development during the conversation. The seasoned manager, with extensive experience in finance and corporate strategy, did not回避 controversy but directly addressed each market concern.
When asked about the impact of the "large models吞噬 applications" discussion, Yan stated that while the company is aware of the industry debate, analysis has反而 strengthened their resolve to explore new opportunities in vertical fields. MEITU currently employs two core innovation mechanisms: an internal innovation studio system and a product challenge competition mechanism. The company judges that large and general models cannot outperform AI applications in vertical domains, so the current focus is on seizing vertical opportunities more rapidly. AI technology has significantly shortened the R&D cycle for vertical products and reduced labor costs. The core task is to quickly capture opportunities in vertical fields, embrace innovation, and gain a first-mover advantage.
Regarding the stock price and buyback, Yan Jinliang affirmed that the stock is undervalued, which prompted the repurchase plan. He suggested the primary reason for the price decline is likely the market's普遍 belief that large models will吞噬 applications. Under this logic, strong company performance alone may not prevent stock price pressure. However, after careful analysis, MEITU believes it is difficult for large models to吞噬 all apps, especially in the imaging field, which differs significantly from other areas. The diversity of demand means no single image generation model can fully satisfy all a user's needs across different scenarios.
The core characteristic of large models is generality, making it challenging for them to achieve sufficient depth in vertical fields. Even if large models attempt to enter verticals, their general dialog mode cannot precisely match the specific needs of each vertical user group. Additionally, data for many verticals is not easily accessible to large model companies. For instance, before launching its AI teeth straightening feature, MEITU specifically purchased or photographed data of teeth before and after straightening. Such deep vertical investments are judged incompatible with the business models of large model companies. Furthermore, visual demands are highly subjective, requiring manual intervention and fine-tuning during generation and use, with high demands for controllability. If everyone used large models for simple image generation, the results would be uniform. Users like e-commerce sellers and live streamers need personalized AI images to enhance competitiveness, creating opportunities for vertical AI application companies like MEITU. Therefore, from the perspectives of user demand, large model business considerations, and model training, the probability of large models吞噬 all vertical imaging applications is very low.
Concerning MEITU's model strategy, Yan explained the decision logic is simple: the sole criterion is "which model delivers better results." MEITU does not develop its own general-purpose models, but this does not mean it is not布局 vertical models or optimizing and training based on external models. This is closely tied to their business model, which focuses on the vertical field of productivity tools,锁定 specific user needs without requiring a general model for all scenarios. He addressed concerns about rising costs from using a "model container" approach, stating that MEITU plans to gradually replace third-party models with self-developed ones. The strategy involves initially using third-party models for basic user needs. Once high-frequency usage is identified for a specific need, the tech team assesses if the third-party model delivers optimal results or has shortcomings. If there's room for improvement, MEITU will tackle the need point through self-development, subsequently eliminating reliance on the third-party model. An example is the popular AI snow scene feature in MeituXiuxiu, which initially used third-party model support but was later replaced by MEITU's own fine-tuned model, resulting in better effects and significantly lower costs. The core logic is to精细调优 high-frequency user scenarios, achieving better results than third-party models while reducing costs.
Addressing concerns about over-reliance on imaging product subscriptions, Yan Jinliang stated this is not a major concern. The primary monetization methods for imaging products are advertising, subscriptions, and one-time purchases. Advertising can disrupt the user experience and potentially conflict with imaging products. The subscription model is undoubtedly the most stable optimal solution among the three. By clearly focusing on subscriptions and深耕 vertical fields, concerns about a singular revenue structure are mitigated.
On the sustainability of growth, particularly with MAU appearing to plateau domestically, Yan clarified that high MAU growth is not the current primary focus. The company sets an annual MAU growth target of only 3% to 5%, a point emphasized for years. The domestic lifestyle scene market is relatively mature, with expected seasonal MAU fluctuations. The growth focus for the lifestyle scene is actually on extending value: using AI features to deepen付费点 and convert ordinary users into subscribers. The future动能 for revenue growth核心 comes from the交替 of the second and third growth curves, with the third curve being productivity scene subscriptions plus Token consumption. This交替 is expected to continue through 2026. Subscription growth in the lifestyle scene will inevitably slow as the base enlarges and it matures. However, the productivity business, though still small and not yet in large-scale development, holds immense growth potential sufficient to offset the lifestyle scene's slowdown. Overall revenue growth in 2026 is expected to remain high, similar to 2025. Crucially, the ARPU elasticity of productivity tools is far greater than that of the lifestyle scene. For example, the highest subscription tier for Meitu Design Studio was 30 yuan/month, but after the launch of Agent, a tier接近 100 yuan/month was added. A new product under testing could achieve an ARPU of over 50 USD, equivalent to over 300 RMB, compared to the lowest MeituXiuxiu monthly fee of 15 yuan. Furthermore, productivity products target the global market, where user ARPU potential is higher than in China, offering greater growth space. Some high-frequency users of Meitu Design Studio have monthly Token consumption equivalent to nearly 20,000 yuan, indicating the potential energy of the productivity business, which could see growth not by multiples but potentially by an order of magnitude.
Looking five years ahead, Yan expects revenue from the productivity scene to be several times greater than that from the lifestyle scene, which will彻底改变 the current revenue structure. He envisions MEITU becoming an AI application company focused on various vertical fields within imaging design. The lifestyle scene itself can be broken down into multiple verticals. MEITU aims to be the leading company in all imaging-related vertical fields globally, helping users significantly reduce costs and increase efficiency, with MEITU capturing a reasonable share of this value.
Regarding capturing user needs, Yan highlighted the advantage of MEITU's user base. While not massive compared to domestic internet giants, the 280 million users are focused on单一产品, making their usage behavior a rich source of inspiration. The birth of apps like "Kaipai" and "Wink" stemmed from observing high-frequency usage of specific features within existing products.
Beyond the user base and vertical focus, Yan identified a unique moat in MEITU's organizational structure, where the design team holds a very high status. This决定了 the core competitiveness in imaging effects. In many companies, technical staff often decide if an AI product launches, potentially overriding designer concerns about user satisfaction. At MEITU, the entire company culture revolves around effect quality. Designers' demands for效果 influence both products and technical R&D, and product managers also prioritize效果. This focus on效果, accumulated over time, forms a moat difficult for AI to replicate and maintains MEITU's industry advantage in imaging effects, a culture whose importance is magnified in the AI era.
On vertical opportunities, Yan mentioned several areas, such as products for offline restaurants and features for global professional photographers to enhance editing efficiency, with more details potentially revealed at an upcoming imaging festival in June.
Discussing global differences, Yan noted that付费 habits between domestic and international users are converging. Domestically, users initially resisted paying for apps but gradually accepted subscriptions and are now adapting to Token consumption. Internationally, acceptance of app payments has always been higher, now moving towards Token models. In terms of payment decisions, domestic users prefer pay-per-use, even if the average single payment is higher than a monthly fee, while international users lean towards annual packages for savings. However, this difference is expected to narrow, especially under Token models, ultimately leading to homogenization, as users pay directly for results. The key metric for AI success is not Token consumption volume but delivery capability – the ability to help users produce more expected results.
Finally, on global positioning, Yan stated that MEITU is already a globally positioned company, not merely an "outbound" company, which implies a China-centric view of the world. MEITU aims to "stand globally to view the globe." Localization is core in the visual field due to significant aesthetic differences worldwide. MEITU will not simply复制 domestically successful models overseas but will深入了解 local user needs in key countries/regions to build products符合 local culture. Regarding data compliance, MEITU has been preparing for years, having launched its first outbound product over a decade ago, and has mature experience, adjusting measures promptly according to global AI regulations.
Comments