Does ecological failure at the starting line mean that even strong technology cannot be monetized? Original Author: Wang Siyuan, Editor: Sai Ke

"We hope to focus on creating a product and push it to the extreme." Yang Zhilin, the founder of Dark Side of the Moon, has always had high expectations for Kimi.

In March of this year, the large model Kimi suddenly exploded in popularity, making Dark Side of the Moon a favorite of capital, reaching a valuation of $3 billion in just over a year since its establishment.

However, like other large model startups, after two months of "fever," Kimi began to cool down rapidly. But as Dark Side of the Moon firmly adheres to the C-end application route, it did not miss this unique commercialization opportunity.

Not long ago, Kimi launched a paid option called "Fuel Kimi," with amounts ranging from 5.2 yuan to 399 yuan. This gameplay, similar to "tipping," has led outsiders to speculate whether Kimi has already begun exploring commercialization.

Frankly speaking, Kimi's monthly active users, both on the web and app versions, have achieved a qualitative leap in a short period of time. But in the short term, the threshold for large models in C-end applications is indeed lower, and the return is faster. However, industry competition is also more intense, so for Kimi, this path to success is steeper.

Part.1

ToC is a Steady Card

After the Hundred Model War, the debate over the ToB and ToC routes for large models' domestic commercialization has been brewing in the AI circle for over a year.

One view is that B-end applications are currently relatively clear, meaning they already have a certain market demand and application scenarios, making it easy to achieve widespread application in multiple scenarios. In contrast, the C-end market requires a highly attractive hit app to lead the market and achieve rapid growth.

Another view is that the B-end market currently faces the problem of insufficient profit margins. Due to fierce market competition, many companies are engaging in price wars, greatly compressing the profits of large models. At the same time, domestic companies have a relatively low willingness to pay for software, further aggravating the difficulties of the B-end market. Compared to the B-end market, the C-end market can attract a large number of users more quickly, thereby achieving profitability.

From a capital perspective, the potential user base of the ToC market is huge, and the capital market has higher expectations for the future of ToC projects. China's rapidly growing internet industry is the best example. Companies that succeed in the ToC market based on internet companies usually obtain higher valuations and more investment opportunities.

"AI obviously conforms more to the growth laws of the consumer internet in the ToC market and is naturally more favored by capital," said an investor focusing on AI ToB businesses.

The reason why Dark Side of the Moon was able to obtain $1 billion in financing in a short period of time and achieve a valuation of $3 billion is precisely because its ToC strategy aligns with the capital market's expectations for high growth and high returns.

From a product and market perspective, ToB also means a long implementation cycle, with input not proportional to expected returns.

Many customers in the ToB market, especially domestic customers, have high requirements for customized services, with diverse needs that are difficult to standardize products. The development cycle for products and services is long. Although AI companies have raised considerable financing, even star startup teams like Dark Side of the Moon have limited personnel size and funding costs, making it difficult to divert attention to developing products for a large number of different scenarios.

Li Kaifu, CEO of Zero One AI and Chairman of Sinovation Ventures, also stated that most traditional companies are reluctant to adopt disruptive technologies for business innovation at this stage. More seriously, most Chinese companies do not recognize the value of software and are unwilling to pay for it.

Simply put, the barriers in the ToB market almost only come from technology. Not only does one need to have good technology, but one also needs to have a deep understanding of the industry and the customer's technology. In the ToC market, the success of basic products can quickly accumulate user reputation and form a brand effect. Then, relying on products and brands, more segmented fields can be opened up.

Therefore, whether from a capital or market perspective, ToC is an extremely stable card for startups like Dark Side of the Moon. However, choosing the right path does not mean that Dark Side of the Moon can quickly form a core advantage.

Part.2

Difficult to Create Super Apps

A significant feature of Dark Side of the Moon's C-end commercialization is to create "super models" + "super apps," a dual-drive approach combining technology and products.

Initially, Dark Side of the Moon was known for its large model Kimi, which has excellent long-text capabilities. Based on Kimi, related applications such as Kimi Chat were launched. Through an immediate feedback mechanism where C-end users directly provide feedback on the product's experience and functions, Kimi Chat was able to quickly iterate and optimize its product.

Data shows that from December 2023 to February 2024, Kimi's monthly active users were 508,300, 1,128,500, and 2,984,600, respectively. Especially in February 2024, the number of users was almost six times that of December 2023.

Data from both the web and app sides also shows a growth trend. On the web side, the daily active user count on March 14th was 345,700, a significant increase compared to 127,900 on March 9th. On the app side, the daily active user count on March 16th was 50,000, while it was less than 20,000 on March 2nd.

But the reality is that Kimi is still far from being a true super app. The industry consensus on super apps is that they must meet three conditions, including high application reliability, the ability to fully understand user intent, and a DAU between 30 million and 300 million.

Obviously, Kimi still has a gap in these three aspects. If we look back at the mobile internet era or earlier technological revolutions, each new technology emergence requires a certain cycle, including technological improvement, cost reduction, and hardware support. When these conditions are met, C-end hit apps that can solve real user needs will emerge.

However, for tool-type applications like large models, embedding them in other application ecosystems in the form of intelligent assistants is an important direction. Baidu, Tencent, iFLYTEK, ByteDance, and 360 have all released AI app assistants, indicating that major companies are betting on C-end applications. With their advantages in ecosystems and users, large companies will be easily accepted by users in the C-end market.

QuestMobile, a domestic business intelligence data service provider, released its 2024 China Mobile Internet Spring Report. The report showed that as of March 2024, the user base of the AIGC App industry has exceeded 73.8 million, representing an eight-fold increase year-on-year. Among them, Doubao under Douyin and Wenxin Yiyan under Baidu both have monthly active users in the tens of millions.

It's not that Kimi doesn't have the potential to become a super app, but rather that large companies with super apps and ecosystems have already won at the starting line.

Part.3

Many Opportunities but Limited Time

Dark Side of the Moon's focus on the C-end market is also aimed at accelerating the pace of commercialization.

Yang Zhilin, the founder of Dark Side of the Moon, is known as a representative figure of the technological faith school in the domestic technology industry. However, this does not mean that Yang Zhilin intends to "generate electricity for love" forever. Just like Dark Side of the Moon, which is not short of funds, it cannot ignore commercialization forever. Kimi's gray-scale testing of tipping functionality is considered a commercialization attempt.

Industry insiders have analyzed that large models themselves are very costly. If companies cannot find a direction in technology for a long time, their future development space will be greatly limited, affecting investors' patience, which forces companies to quickly obtain commercial benefits.

However, as Kimi has not yet become a true hit app for the general public, Dark Side of the Moon's tipping model has also sparked discussions about whether the exploration of large model commercialization is premature.

In fact, not only Dark Side of the Moon, but also all large model companies, cannot find a relatively perfect business model for C-end applications. Taking OpenAI as an example, its annual recurring revenue has grown to $3.4 billion in the past six months.

However, from its business development perspective, what drives revenue growth is the subscription fees for the ChatGPT series and the fees paid through API calls to the model. It is estimated that the total revenue from ChatGPT Plus subscriptions in May may be around $60 million, with the annualized revenue from ChatGPT Plus membership fees exceeding $720 million, accounting for only one-fifth of the total revenue of $3.4 billion.

That is to say, OpenAI's main source of income is likely to come from APIs, and most of those who pay for APIs are B-end enterprise customers. This is not good news for Dark Side of the Moon, which is firmly committed to the C-end market.

It cannot be denied that large models are still in their early stages of development, and true super apps have not yet emerged. There are opportunities for any player. However, the fact that Dark Side of the Moon must accept is the high cost of training, inference, and customer acquisition for large models, which is a must-answer question for all large model companies, especially those targeting the C-end market.

It has been previously reported that investors have revealed that the customer acquisition cost for Kimi users reaches 12-13 yuan, with an average daily download volume of 17,805 in the past month, resulting in a daily customer acquisition cost of at least 200,000 yuan. Coupled with the costs of purchasing computing power, research and development, training, maintaining daily operations, manpower, and material resources, relying solely on early-stage financing may be difficult to sustain long-term operations.

The high computing power and customer acquisition costs make large models seem less "sustainable." Monetization is better done earlier rather than later. However, the current situation is that large model products have not yet become necessities in people's work and life. Once money is involved, the situation may well reverse.

"The commercialization model that Kimi is exploring actually targets users' needs for efficiency and stability," said Wang Peng, a deputy researcher at the Beijing Academy of Social Sciences. He believes that a paid priority model may become a differentiated competitive strategy for Kimi, but it is also necessary to pay attention to the impact of an increase in paid users on free users, such as a decline in service quality.

In the current business environment, Dark Side of the Moon has undoubtedly achieved certain results with its exploration of the C-end market using its large model Kimi. With continuous technological progress and the gradual maturing of the market, Kimi, as a flagship product, has demonstrated strong potential and appeal. However, commercial success does not happen overnight. It requires time to cultivate the market, optimize products, and build ecosystems.

In addition, facing fierce market competition and high cost pressures, Dark Side of the Moon must recognize that although there are many opportunities, time is not abundant. Competitors emerge endlessly, user needs change rapidly, and technological iterations are astonishing. In this process, Dark Side of the Moon needs to continuously learn and adapt, keep up with the pace of large companies, realize its potential as soon as possible, and explore more possibilities.

/ THE END /