Entering the "Hundred-Million-Level Club", AI to C approaching the eve of the Super App.

2025, Everything Focuses on C

In the era with the Internet as a precedent, when the large model technology swept in, the "Super App of the AI Era" once became a story like "the boy who cried wolf" in 2023.

However, in 2024, various signs indicate that the Super App has reached the eve of its emergence.

First, let's look at the cost of infrastructure. Just as the decline in the cost of communication and mobile devices promoted the explosive growth of the mobile Internet, in 2024, the model price war broke out, and the cost of AI has dropped by nearly 90% - the improvement of model capabilities and the reduction of model costs have brought AI to C applications to the critical point of explosion.

Next, let's look at market demand. As the AI industry has gradually moved from the explosive period of customer acquisition driven by technology to the calm period of customer acquisition relying on product power, many AI applications that have landed in scenarios and met specific rigid demands have verified PMF (Product-Market Fit) - and it is the C-end AI applications that bloom in many fields such as productivity, companionship, and entertainment that make the general public truly feel the power of AI.

At both the scale and commercialization levels, the C-end has shown a quite high ceiling. For example, MiniMax among the "AI Six Tigers" was previously reported to have an annual revenue of more than 70 million US dollars, and most of its income comes from the overseas C-end application Talkie; as of the end of December 2024, the monthly active users of only the AI function of Baidu Wenku have exceeded 90 million, and the paying users have exceeded 40 million.

The rapidly increasing number of users and annual revenue are still continuously verifying the market potential of C-end AI applications. According to the data of the AI Product List, among the top 10 products on the global AI product list in October 2024, 8 are C-end products, and the traffic is still increasing - among them, Microsoft's Copilot increased by 87.57%, and the AI search application Perplexity increased by 25.53%.

In the verification year of AI PMF, many manufacturers have chosen to "focus on C" and seize the broad C-end market. For example, Zhipu, which had a rather distinct To B color at its establishment, shifted its business focus to the C-end application "Zhipu Qingyan" in 2024; another AI unicorn, Dark Side of the Moon, in September 2024, reduced and merged its overseas products Ohai and Noisee into its leading C-end product Kimi - and the consequent is the targeted investment of resources. App Growing data shows that the delivery volume of Kimi soared in October 2024, approaching the total delivery volume of its entire Q3 in 2024.

The development process of the mobile Internet proves that in the early stage of the implementation of new technologies, productivity tools often become the dominant ones, and later they will also develop into large-scale rigid demand scenarios.

The AI industry is no exception. The consensus of many entrepreneurs is that the leap in AI generation capabilities and generation efficiency directly solves the productivity problem, such as improving the efficiency of users' information acquisition, summarization, and input in scenarios such as writing, searching, and design. The "AI Productivity Tools Summer Development Report" released by Yuehu Data shows that the compound growth rate of the monthly active user scale of AI productivity tools in the first half of 2024 reached 9.8%, and the number of users increased by 60%.

Among them, Quark and Doubao, who are from technology giants, have already formed a new pattern. According to the statistics of Quantum-bit Think Tank on the cumulative downloads from January to October 2024, Quark and Doubao are the only two players in the 166 major domestic Apps that have entered the "Billion-level Club". Among them, Quark, which ranks first, has a cumulative download volume of more than 200 million, laying an important foundation for its position in Alibaba's AI to C product layout.

Cumulative downloads of domestic AI Apps from January to October 2024, Image source: Quantum-bit Think Tank 2024 Annual AI Top Ten Trends Report

In addition to attracting new users, user activity is also an important indicator to evaluate product power. Analysys data shows that according to the monthly active user performance in November 2024, the domestic mobile AI applications have formed a pattern where Quark (71.02 million) is the first, and Jianying and Doubao (62.22 million) are in the first echelon.

It is not difficult to find that these AI products that have preemptively verified PMF have put a lot of thought into their product positioning. For example, Kimi, which firmly holds the top position among start-up companies, has the label of "long text"; Doubao's label is the "big tree" behind it: ByteDance, which controls the important traffic entrance of Douyin.

Recently, Quark upgraded its brand slogan - "AI All-round Assistant for 200 Million People", choosing a different development direction from Kimi and Doubao. Quark's advantages in AI capabilities and the number of users have enabled Quark to smoothly start a new story in the AI era. Quark has become an easy-to-use and all-round "AI product power" like a Swiss Army knife. By exerting efforts in multiple productivity scenarios such as AI search, AI writing, AI PPT, and AI summary, Quark not only innovates the product experience but also firmly grasps the young user group.

In the "Top Ten Trends of AI Industry Development in 2025" report released by Analysys, Quark is the only AI application where the number of "Post-00s" users accounts for more than 50% - this group of future consumption main forces is an important cornerstone for AI products to seize the commanding heights of the C-end.

It can be seen that although their previous development processes are different, the outstanding performances of Quark and Doubao in the market have made the market see the possibility of the emergence of Super APP. Winning at the starting line, the support of the large group behind is crucial. But in the final analysis, the cliff-like leading advantage is inseparable from the two teams' understanding and cognition of users and AI, as well as their continuous exploration and promotion of Super App.

Over 100 Million, the Watershed of AI to C

The progress bar of the AI industry has passed the large-scale model training in 2023 and the search for PMF in 2024. At the beginning of 2025, the main proposition of the AI track has quickly switched to expanding the user scale and finding a business model - among them, having more than 100 million users will become the ticket to enter the next stage.

"Over 100 million" means that under the existing 1.2 billion mobile Internet user scale, the product has a user penetration rate of more than 10%, which is an important milestone from establishing a mindset to large-scale customer acquisition. And at the commercialization level, "over 100 million" also means the verification of PMF.

However, after reaching "over 100 million", AI To C enters the next stage. In mature industries, the importance of the first-mover advantage is self-evident. Securing a position of "over 100 million" is equivalent to seizing a piece of the cake in the market first. But in the AI track, which is still in the stage of making the cake bigger, it is more important to accurately understand the needs of the remaining 1 billion users.

Now, both Quark and Doubao have become members of the "Billion-level Club". In order to seize the next "1 billion", Alibaba and ByteDance both regard C-end AI applications as a new growth space in terms of investment.

On September 12, 2023, the first all-staff letter issued by Wu Yongming, CEO of Alibaba Group after taking office, announced that the business priority would be reshaped based on the two strategic focuses of "User First, AI-Driven". In May this year, Alibaba's management once again emphasized at the Q4 2023 earnings conference that by investing resources in researching AGI and developing large models, it can help business products innovate and transform in the AI field. Coincidentally, ByteDance is also quite aggressive in the layout of C-end products. It not only established a new AI product line Flow, but also the investment in "Doubao" is the largest among similar conversation products.

Along with the simultaneous increase in strategic importance, the best ladder composed of talents, products, etc. has become the backing of Quark and Doubao. For example, both Quark and Doubao are personally supervised by the group's top leader, and young managers with technical and product backgrounds are dispatched for management.

In order to win the battle of AI To C, Wu Jia, President of Alibaba's Intelligent Information Business Group, will lead Quark, Tongyi App, etc. to explore Alibaba's AI to C business.

In addition, an important factor for a C-end AI product to achieve rapid growth is: product power. Quark's pragmatic, focused and unique product logic and concept have become a clear stream in the domestic Internet market. Specifically, Quark always takes serving the actual needs of users as the starting point, uses "AI technology" to amplify the user value of productivity tools, and finally creates the "Super Search Box" of the Quark mobile terminal and the new Quark PC terminal and other products with excellent user reputation.

Quark has created a "simple, efficient, and intelligent" product form, which also makes Quark a model for many similar products in different periods. The emergence of the AI large model has accelerated Quark's ability and efficiency to serve the needs of ordinary people, and Quark's AI gene has a larger stage to show.

Some industry experts believe that the large demand for funds, computing power, talents, and data also confirms from another level that AI to C is a competition participated by large companies. Zhang Lu, the founding partner of Fusion Fund, once judged in an interview that currently, 90% of the opportunities in consumer-grade AI are in the hands of large technology companies because they have advantages in the "ABC" aspects: A is algorithm, B is big data, and C is computing power.

And the team's cognition of the product and AI is the hidden barrier of AI applications. If we disassemble Quark's development path, we will find that Quark has never limited the product value of itself to the traditional search engine category. This "not being defined" approach requires the Quark product team to have considerable innovation courage and precise insight into user needs.

From innovating popular scenarios such as AI search, AI writing, and AI PPT, to upgrading tools such as AI network disk and AI scanning, Quark's new story in the AI era is not only to become an "AI all-round assistant for 200 million people", but to infiltrate Quark AI capabilities into every function to allow users to obtain an "ultimate experience". If the old users of Quark enjoyed the improvement of information acquisition efficiency brought by the minimalist search in the era of the mobile Internet information explosion. Then in the AI era, what can be enjoyed from "Quark" is the full-process efficiency service from information acquisition to information production - the significantly perceptible improvement in the user experience is also an important reason why Quark quickly obtained the ticket to the finals in the AI era.

And Quark's phased success in the AI era will in turn bring another possibility for Alibaba AI to C. Quark's comprehensive cultivation in the fields of AI search, AI writing, and AI summary, as well as the high-granularity AI vertical landing experience, can also provide diverse ideas for Alibaba's exploration of AI to C.

Applications, Solving Users' Practical Problems

When the AI industry enters the scale-up year of the C-end, it is difficult to ignore a phenomenon that many AI to C products are facing the "inverted U-shaped" curse of growth.

For example, Lensa, an AI photo generation application that once topped the download charts in the Apple Store in many countries such as the United States, Canada, and the United Kingdom, with a maximum daily download volume of more than 15 million, saw its download volume drop below 2 million just over a month later; Jasper.ai, a star unicorn in the AI copywriting generation field, saw its user scale rapidly decline from 10 million in February 2023 to 7 million three months later, and the stock valuation for employees also decreased by 20%.

The reason why the "inverted U-shaped" curse is difficult to escape is that either the product hits the individual pain points of users, but the product function is too single-point and not deep enough, resulting in a thin barrier and low user stickiness; or because it overly emphasizes "intelligence", while integrating AI functions, the product power does not keep up, resulting in the product not being easy to use and not truly solving the real pain points of users.

When the pursuit of AI Super App becomes a curse on enterprises, the strategy presented by Quark returns to the essence of polishing the product power itself.

How to understand Quark's product power? "Generally speaking, compared with Doubao and Kimi, users will more easily accept Quark's transformation from a search engine to an AI all-round assistant." Kong Xin, a senior product expert, mentioned in an interview with the media, "Because the former is to let users directly understand the AI assistant, which is quite difficult for ordinary users."

In July 2024, Quark launched the "Super Search Box" on the mobile terminal, integrating one-stop AI services of "search, creation, and summary". It is not uncommon to launch an AI search product based on Internet information. The valuable point is to make the user demand scenarios detailed and in-depth.

For example, in the field of education, Quark upgrades the function of "AI Question Searching". When users encounter learning difficulties, they can take a photo to search for various difficult and new questions, and support AI explanations; another example is the "Take ID Photos" function of Quark Scanner King, which allows users to complete the shooting, retouching and sharing intelligently without going to a professional photo studio, and make various ID photos anytime and anywhere.

Currently, the AI service ecosystem of the mobile terminal has entered the fast lane of development. However, Quark focuses on the lag of AI services on the PC terminal - the current single AI plugin and Copilot cannot provide users with comprehensive and cross-system AI services. While continuously upgrading the mobile terminal, Quark also released a new PC terminal last year, seamlessly integrating AI technology into the user's computer usage process.

Quark PC terminal interface

Jumping out of the traditional plugin mode, Quark PC terminal allows users to invoke Quark's AI functions at any time through interactive methods such as desktop AI assistants, word selection, screenshots, and right-click in computer scenarios, to achieve operations such as intelligent search, efficient writing, and information summary. This makes every computer enhanced by Quark AI an AI PC, and allows AI interaction to seamlessly integrate into the user's daily operations.

Not only crossing from the mobile terminal to the PC terminal, "multi-terminal integration" is Quark's unique product concept. Nowadays, most of the users' content production and consumption behaviors are fragmented and scattered in different device terminals such as PC and App, as well as different applications such as browsers and network disks. This also makes the user's content experience unable to form a complete ecosystem.

Quark is one of the first batch of AI applications in the industry to launch a "multi-terminal integration" product experience, allowing users to realize AI search, AI writing and creation, AI summary, as well as information collection, generation and sharing in scenarios such as network disks and scanning through AI-generated content and AI tools on clients such as App, iPad and PC.

Of course, for most enterprises, cross-terminal is a "tough and dirty job". If native development is carried out, the R & D cost will increase exponentially for each additional terminal, and the subsequent maintenance difficulty will also be greatly increased.

But focusing on doing "difficult but correct" things is the rare quality of the AI era presented by