In 2024, the competitive logic of the domestic AI industry has been altered by a major large model enterprise. The sudden rise of Kimi has provided an answer to the industry’s anxiety over creating hit products – through investing in traffic to drive user growth. As everyone joins this marketing game, it has brought a brief prosperity to the domestic large model market. Starting from March, the traffic of leading domestic AI products has seen a significant increase. In March, among the top 10 AI applications in terms of access volume in China, 9 had a growth rate of over 40%. Source:钛媒体.
The Rise of Kimi and the Shift in Competitive Logic
The changes originated from the sudden rise of Kimi. Starting from February this year, the access volume of Kimi began to surge, doubling in January to 3.05 million in February. In March, Kimi’s traffic surged to 12.61 million, a growth of 321.58%. Since then, Kimi’s traffic has remained above 20 million. At that time, the sudden rise of Kimi had significant implications for the domestic large model industry. On the one hand, although everyone was focused on OpenAI, due to the differences in market environments, there was no clear path for localization. On the other hand, since 2023, most AI products have been fleeting, with few hit products. The industry generally exists anxiety over hit products. In this situation, Kimi found a clear growth path for the industry – investing in traffic.
The Tipping Point towards Large Enterprises
Large model products are being harvested by large enterprises. From the data, the proportion of large enterprises in leading AI products has increased significantly. According to AI product rankings, among the top 10 AI applications in terms of access volume in China in March, half came from start-ups, including Kimi, Mita AI Search, Zhipu Qingyan, AiPPT, and Gaoding AI. However, by July, among the top 10 AI applications in terms of access volume, only two were from start-ups, Kimi and Mita AI Search, while the rest were from internet giants, including Baidu, Alibaba, ByteDance, and 360. This is not surprising. Under the logic of traffic competition, large enterprises have a natural pool of users and ample marketing budgets, which are more capable of winning this user competition war.
The Importance of Payment Rate as the North Star Metric
In product development, there is a concept called the North Star metric, which is the most core indicator of product quality. From PC internet to mobile internet, the North Star metric has always been changing. The North Star metric is often related to the technical characteristics and business model of the product form. Taking mobile internet as an example, the core of mobile internet is scenario extension. With the platform migration from non-smartphones to smartphones, the number of internet users has expanded from 200 million to 1.4 billion, giving rise to countless new scenarios such as food delivery, ride-hailing, and short videos. In the era of mobile internet, the user growth dividend in new scenarios, combined with factors such as platform network effects, has become the most core indicator affecting the competitive results. From the business model perspective, advertising is the most mainstream business model in mobile internet. The potential of advertising business is highly related to user activity, such as retention, usage time, and user scale, etc. However, in the AI era, the North Star metric based on user activity has become invalid. Unlike mobile internet, AI does not create new scenarios or so-called network effects. Its value lies more in efficiency improvement, including a significant reduction in content costs, new experiences in old scenarios, and replacing repetitive professional work, etc. In short, AI products are more like productivity tools. This can also be seen from user retention. Currently, the retention rate of AI products is much lower than that of mobile internet products. According to a recent set of data on the Internet, the retention rate of large model products after 30 days is less than 1%. That is to say, among the 100 users attracted by large model products, the most 30 days later will only be one person left. Although this data has not been confirmed, Sequoia Capital has also mentioned this point. According to Sequoia Capital’s data, the retention rate of mobile internet products is as high as 85%, and the willingness of users to continue using generative AI products and the ratio of opening them every day are only 56% and 41%, respectively. In addition to the low retention brought by the tool attributes, the business model of AI products is also different from the 羊毛出在猪身上，狗来买单 model of mobile internet. Up to now, the mainstream business model of AI products is subscription. Among the top 40 AI products with global influence, 70% of companies adopt the subscription model in traditional software, 10 companies (26%) adopt a mixed pricing model combined with subscription fees and usage, and Fin (Intercom), EvenUp, Chargeflow (OpenView portfolio company), and 11x.ai are trying to pay based on results, that is, customers only need to pay for successful results. Whether it is a subscription model or a result-based payment model, it shows that in the logic of efficiency improvement, AI products have huge commercialization space for users. Therefore, indicators such as payment rate and repurchase rate that reflect payment conversion are more reflective of the value of AI products than user scale. From this perspective, without running through the payment logic, pursuing user scale alone has limited significance. If the competition of AI products is just simple traffic competition, then domestic large model products will eventually be harvested by large enterprises.
Conclusion
The rise of Kimi has brought a new growth model to the AI industry, but the industry must also pay attention to the sustainable development of this model. In the face of fierce competition and increasing costs, AI products need to focus on improving their core competitiveness and exploring new business models to ensure long-term development.
Views: 0