Can AI help Keep (03650) emerge from its darkest hour?

Recently, the concept of sports and fitness has received frequent attention. During the two sessions, relevant officials from the National Health Commission stated that they will implement a 3-year action plan for "weight management year" to promote healthy lifestyle habits. This plan is in line with the implementation plan for the "weight management year" activity formulated by the National Health Commission and 16 other departments in June last year, aiming to establish a supportive environment for weight management by 2024 through about three years of effort. "Weight management" has been widely discussed, and some netizens have directly called out to Keep on social platforms, hoping to make better use of Keep's platform for scientific fitness courses, smart fitness equipment, and professional membership services to more efficiently achieve their weight loss goals in the context of policy promotion. These policies and phenomena seem to be favorable for the sports and health industry. However, according to Keep's latest 2024 annual report, the domestic leading fitness platform's losses have not decreased but actually increased the financial report shows that Keep's net loss for the whole year increased by 59% compared to the previous year. Why is Keep not making money despite the increasing number of fitness enthusiasts? The answer may lie in the details of the financial report. In the second half of 2024, Keep's research and development investment increased significantly, with AI-related expenses accounting for the majority. Keep also stated in the financial report that the main reason for the increased losses was the continuous investment in new strategic business and technological innovation plans to strengthen long-term competitiveness, leading to increased related expenses. Clearly, Keep is trading short-term losses for long-term barriers in AI fitness. From intelligent personal training to AI-generated courses, Keep is attempting to use technology to reshape the user experience. With the ongoing attention to fitness policies and the continuous expansion of the population participating in sports, these macro backgrounds may indeed bring a surge of users to Keep similar to that during the pandemic, presenting another real opportunity for wealth. At the same time, whether this "Normandy landing" strategic investment can catch this wealth and truly lead Keep to break through its growth dilemma still remains to be verified by the market. 01 The "slow business" of fitness verticals vs. the "fast content" of traffic platforms. Let's first address the first issue. This perennial question, whenever a new "hot chicken" emerges on a traffic platform, Keep is criticized. It started with Pamela, and then followed by Liu Xinghong. Undoubtedly, with the rise of platforms like Xiaohongshu, Douyin, and Bilibili, there is a concern that their massive UGC content might erode the audience for fitness verticals, posing the biggest threat to Keep. After all, when Liu Xinghong's Douyin live workouts gained 40 million followers in 12 days for free, that's real data, while Keep, after ten years of hard work, only managed to accumulate a monthly active user base that barely matched it. But are the users in live rooms truly flowing in one direction from live rooms to Keep's core users? A little thought will reveal that they are not. Apart from the rise of training bloggers on Bilibili, there is a significant amount of "Keep course follow-up check-ins" on the platform, which is the "second flowering" of Keep's original users on Bilibili. On Xiaohongshu, there is a myriad of "fitness equipment reviews," but the recommended fitness gear is often from Keep. When fitness KOLs on Douyin analyze their workout results, they often use Keep data as endorsement. Clearly, phenomena like Liu Xinghongs explosive growth in traffic and Keeps sports content are not competitive, but rather constitute different dimensions of the fitness market the former brings mass participation heat, while the latter provides systematic and sustainable training programs. In short, the difference lies in Keep's positioning, which offers a "solution" rather than "content consumption." The professional systematized content built on accurate user insights is Keep's moat. This company, established for nearly a decade, has maintained its irreplaceability in what seems like a crowded field, thanks to over 40,000 recorded courses, 3 billion user behavior data points, and a precise breakdown of the pain points of "newbie users." Since its launch in 2015, Keep's course content has covered over 200 sub-scenes from fat reduction and shaping to recovery and stretching, with over 90% of the content being self-developed. Unlike the relatively disorganized and fragmented "popular fitness follow-along videos" on social platforms, Keep's course design follows the "FITT principle" (frequency, intensity, time, type), forming a stair-step advanced path. For example, the "3-minute fragmented training" for beginners and the "marathon preparation plan" for advanced users have been long refined by the sports science team and hundreds of professional coaches. The systematic content results in significant user stickiness. According to official data, Keep users open the app an average of 4.2 times per month, far exceeding the industry average for fitness verticals. A fitness industry expert commented, "The popular fitness content on Douyin is like a fast food, while Keep is more like a private kitchen users know they can find a solution tailored to their body here." The deeper moat lies in Keep's understanding of its core users. By analyzing billions of user training records, such as completion rates and dropout rates, Keep can accurately identify user pain points. Based on this, Keep embeds dynamic difficulty adjustment features in its courses and provides psychological motivation content tailored to different stages. In contrast, social platform algorithms tend to recommend content with high completion rates rather than meeting real user needs. For example, popular fitness videos on Xiaohongshu are often visually stimulating like "7-day slim waist," but Keep's "21-day scientific fat burning plan" includes heart rate monitoring, dietary advice, and other long-term services. This difference results in Keep's user retention rate far exceeding platforms that rely on traffic distribution. An analyst from a consumer industry agency pointed out, "The endgame of the fitness industry won't be a war of traffic but a depth of service. Keep has shown over the past decade that when users seek certain results, the value of systematic content + data loop far exceeds fragmented entertainment." In the widely regarded as anti-human fitness industry, slow may indeed be the fastest way. 02 All in AI, Keep's "Normandy" moment"Como ests?" "How are you?"This time, Keep wants to solve with AI. Breaking down the financial report, it is not difficult to find that in the first half of 2024, Keep's losses narrowed significantly by 28% year-on-year, continuing the trend of continuous optimization in financial performance over the past few years. However, in the second half of the year, unexpected losses expanded. Behind this "anomalous" data, it should be a decisive move by Keep: increasing investment in AI research and development. In the eyes of investors, this may mean that all negative news has been released, but Keep's ambition is obviously greater - it does not want to just be a fitness app, but aims to become a vertical AI application in the fitness field, like Meitu and Duolingo that have already reaped the benefits of AI. Many institutions have compared Keep with the above two and have pointed out that in both image editing and small scene teaching, AI has helped platforms achieve efficiency improvement and user experience optimization. The same logic applies to Keep. Especially with Duolingo, which is also a self-discipline check-in app and is used for "teaching and learning" scenarios, Duolingo's stock price has risen more than 400% over the past two years, which undoubtedly points Keep in the right direction and brings confidence. However, the guidance of peers is not a decisive factor, and Keep's AI layout is not a mere imitation, but is based on a deep understanding of the pain points in the fitness industry. The problems that Keep faces in content production, as mentioned earlier, such as insufficient coverage of categories and limited content quantity, can all be solved through AI. In the annual performance conference call this year, Keep CEO Wang Ning explained that in the ten years since its launch, Keep's core competitiveness lies in sports content services. However, under this model, both user scale and commercial scale have ceilings. The arrival of the AI era can help Keep break through these ceilings. "AI will significantly and rapidly increase the efficiency of Keep's content production, allowing the platform to quickly add more sports categories in one go, such as golf and tennis courses that were previously limited due to factors such as manpower costs. Following this, Keep can quickly attract more users for various course categories, thereby expanding the user base." With this logic, the monetization problem that Keep has always faced also has a solution. Wang Ning pointed out that currently, most of Keep's members target Keep course content, with an ARPU value of around 20 yuan. However, with the arrival of AI Coaches, from interacting with users, providing emotional value and plan design, has evolved from a simple video to providing a complete sports fitness service. This means that Keep has shifted from a core content selling mode to a core service delivery mode, with the potential for ARPU value to increase by more than 10 times. In addition, AI can also solve the two major bottlenecks that traditional fitness content has always faced: low personalization and weak interactivity. By dynamically generating training plans, providing real-time movement correction, and even combining user physiological data to provide nutritional advice, Keep can upgrade from "standardized content" to "personalized digital personal training for everyone." Of course, this transformation requires a huge initial investment. Data shows that in the second half of 2024, Keep's R&D expenses increased by 18% year-on-year. At the same time, the company actively cleared inefficient courses and redundant business lines accumulated in the early stages, leading to an enlarged loss on the books. This strategy is similar to the early "long-term loss for growth" logic of Amazon - cash flow can be repaired before a technical turning point, but once the strategic window is closed, it is difficult to restart. It can be said that this decision has truly brought Keep to its own 'Normandy moment'. Once successfully landed, the horn of counterattack will be sounded immediately. The capital market's reaction to Keep's financial fluctuations is subtle. Although losses expanded in the second half of the year, the stock price did not drop significantly. Analysts generally believe that this is because Keep's investment direction is in line with expectations: AI model training, AI Coach creation - these are all "one-time investments" that can significantly improve future gross margins. More importantly, Keep's cash flow remains healthy. By the end of 2024, its cash reserves were sufficient, and the subscription revenue from members in 2024 was on a year-on-year growth trend, proving that the core business's ability to generate blood remains strong. One institutional investor commented, "The current losses are a deliberate choice, just like Tesla's early burning money to build super factories - short-term pains are exchanged for long-term pricing power." In conclusion: Keep's "all in AI" is essentially a prediction of the industry's endgame: the core competitiveness of future fitness services is no longer the quantity of content, but the depth of AI-driven services. It must be acknowledged that at this point, Keep has to face the dilemma of long-term benefits of technological investment and short-term cost pressures. Although kinetic.ai has already launched, user feedback data has not been clear yet, and whether users are willing to pay for AI remains to be further verified. Under pressure, above the dark clouds, whether Keep can truly emerge from this "darkest hour" and successfully land on Normandy depends on when it achieves its profit turnaround goal and fulfills its performance commitments.