Capital market sentiment continues to wane, but the big model craze is still raging.

Overseas markets, marked by GPT4.0 and Sora, show the big model's breakthrough in multi-modality and creative video capabilities; the domestic market also had amazing performance. Kimi, an AI application known for processing long texts, quickly became popular, and major manufacturers rushed to death.

Obviously, the wind of the big model is still blowing; it's just that the wind direction is gradually changing.

On the one hand, GM's large model release “ran aground”, and the number of large vertical models “soared”. On the other hand, the big model has been “rolled” into the deep-water application zone, seeking matching scenarios to accelerate application implementation, and has become a breakthrough in commercialization. In particular, it has been agreed that major domestic manufacturers rely on technological and ecological advantages to use large models to restructure all products to promote the dual improvement of user experience and business efficiency.

As one of the major manufacturers active at the cutting edge of major models in China, 360 has now also explored the commercialization path of “big model+knowledgebase+agent”, and earned the “first pot of gold” for AI models in the first half of last year.

Judging from the company's latest financial report for 2023, with the support of the “360 Smart Brain” general model, the “AI content” of the product continues to increase, and the revenue model and structure are constantly being optimized. Take the Internet value-added business as an example. With the full power of the big model, annual user payment performance increased 50% year-on-year, material production increased 200% compared to AI, and corresponding revenue increased to about 1.1 billion yuan. In addition, revenue from all Internet products increased 65.58% year over year to 138 million yuan.

It can be seen that big models have spawned new opportunities for full application implementation, which are also being projected in the financial reports of big manufacturers like 360. Does this further indicate whether 360 is expected to be highly successful?

Losses were drastically reduced in 2023, and the big model monetized for the first time

Perhaps the answer is hidden in the company's latest financial report.

Overall, 360 shows signs of reversing the situation -- losses were drastically reduced in 2023. Although revenue growth is still under pressure, the pace has clearly slowed. According to financial reports, the company achieved annual revenue of 9.055 billion yuan (RMB, same unit) and net profit to mother of 492 million yuan, a sharp decrease of 77.65% year-on-year losses, which is in line with expectations.

In addition, the company's net operating cash flow also increased significantly, rising 65.84% year-on-year in 2023 to 924 million yuan, with a compound growth rate of 136.26% in the past two years. Generally speaking, the larger the cash flow, the faster the flow, the more stable the financial base.

By the end of last year, the company's monetary capital was over 26 billion yuan, an increase of about 3.3 billion yuan over the end of the previous year.

As you can imagine, the more abundant your cash reserves are, the more motivated you are to innovate. 360 has always insisted on strong R&D investment. The R&D cost in 2023 was 3.104 billion yuan, and the R&D rate reached 34.28%, all of which are at an all-time high.

To a certain extent, these improvements in KPIs are also a commercial leap forward in the 360 model.

Product AI continues to be recognized by the market, and it is expected to increase overall traffic and share in the future

If 2023 is the first year of large-scale models being implemented, then 2024 will be a golden period for the explosion of application products. 360 is a typical example. Last year, the company reshaped national Internet products based on the self-developed general model “360 Smart Brain”, released AI products such as “360AI Search”, “360AI Browser”, and “360AI Digital Human Platform” one after another, and received strong market feedback.

According to the “Global AI Product Growth Dark Horse List” published by AIWatch.AI, the “360AI Browser” and “360AI Search” products are continuously at the top of the global AI product high growth rate list. According to the list, the number of visits to “360AI Search” increased by 1677% month-on-month in March; in April, the product was even more popular and went out of business, forcing the government to respond quickly.

In addition to this, in office application scenarios, 360's AI products have also achieved impressive results -- the company's “360 Soda Office” also ranked in the TOP10 list, and its number of visits surpassed WPS in March. Furthermore, the recently launched “360 AI Office” (which includes hundreds of AI applications) received close to 20 million visits on the first day.

As can be seen from this, 360's large model ecosystem with the “360 Smart Brain” as the base was basically formed, and in the “AI running” storm set off by major technology companies, it had a good start, establishing a certain leading edge in the first half.

According to the company's plan, in 2024, it will continue to expand new products on the basis of consolidating superior products, using the office scene as an opening point. The company expects that as these AI applications gradually mature, the company's overall product market share and PC traffic are also expected to increase further.

How imaginative is it to monetize big models?

Currently, the commercialization path of 360's big model is quite smooth. It goes without saying that there are several key reasons or advantages behind this:

The first is ecological status. 360 competes with a series of national-level products developed in the fields of search, security management, browsers, and smart hardware, integrating entrances, channels, and content. In particular, it has a certain voice in the PC and security fields.

According to the company's financial report, by the end of last year, 360's PC browser and navigation DAU was close to 200 million, and the MAU was 411 million, with a market share of over 86.85%, ranking at the top of the domestic browser market. In addition, the company's PC security products have a DAU of over 210 million and a MAU of 470 million, continuing to rank first in the security market.

Since the application ecosystem is more prosperous, the number of users increases, and the data source is continuous, it is possible to ensure that large models have sufficient fuel for continuous iteration. Today's “360 Smart Brain” has jumped to version 4.0, and after a lapse of less than three months, it was iterated on version 4.

Second, there is a natural “military training ground”, especially in various fields such as search, browsers, and digital offices, all of which have long been in a leading position in the domestic market.

Coupled with years of technology accumulation and engineering capabilities of major manufacturers, opening up the link from evolution to monetization of big models should be a matter of course.

If future performance growth depends more on the monetization ability of big AI models, then what is the key to successfully starting to monetize big models?

Currently, Big Model has four mainstream business models, namely 1) developing GPT applications and paying according to the membership subscription system, such as ChatGPT Plus published by OpenAI; 2) selling large model API interfaces and charging companies or developers according to the number of calls, such as OpenAI, Google, etc.; 3) selling large models and customized development services to export large model industry solutions to traditional enterprises. For example, some startups and cloud vendors; 4) Use big models to transform existing businesses to improve the competitiveness of products and solutions to obtain commercial returns.

Judging from 360's past layout and current positioning, there is room for all of the above models. Currently, the first and fourth types should be the first to be cashed out. For example, “360AI Office”, which was only recently launched, uses a membership subscription model and plans to achieve 1 “small goal” in the first year.

At the end of the day, the basic capabilities and commercialization capabilities of the big model are the two key keys that determine the final implementation performance of the big model. At present, 360's performance in both areas is obvious to all.

Among them, “360 Smart Brain” has passed the functional evaluation of the Trusted AIGC Big Language Model of the Chinese Academy of Information and Communications Technology. Comprehensive capabilities rank first in the country, and use this as a foundation. The company has also launched large-scale vertical models such as the Big Tax Model and the Big Medical Model.

In the direction of commercialization, it was not only limited to the software application layer; 360 also pioneered the use of an agent model recognized by the industry. For example, in the second half of last year, 360 released the 360 Intelligent Marketing Cloud, which is a leading example of integrating “big model+knowledge base+agent”. Under the 360 Intelligent Marketing Cloud, products such as AI digital people, AI digital employees, and 360 smart painting have been launched one after another, covering multiple business processes such as customer acquisition, production efficiency improvement, and multi-modal interaction.

Major model makers at home and abroad are turning to AI Agent -- this is regarded by the industry as the ultimate product of AGI, and also found a key to breaking the 2B game.

Most agents currently on the market, including GPTS launched by OpenAI, actually only build a Chatbot based on a specific knowledge base or professional data. These agents are mainly used for question-and-answer interactions, such as obtaining industry information, reports, etc. This shows that the development of this route is still in its early stages, and there are still some problems that need to be overcome in the future. Especially for enterprises, if AI agents such as GPTS are only used for knowledge questions and answers, then their role will be very limited and will not be able to penetrate deep into the enterprise's business processes.

So what is the key to agent implementation?

According to industry analysts, model ability, industry experience, scene fit, etc. will currently affect the performance of agents. Among them, model ability is the core, industry experience is the key, and the carrier is the guarantee. It is worth mentioning that the domestic software industry has created the ability of domestic manufacturers to customize and personalize products, which is conducive to the further implementation of agents.

In fact, at the above levels, 360 is also continuing to work hard.

Among them, “open source+cooperation” are two major paths for 360 to further expand model capabilities. For example, after announcing strategic cooperation and joint development of a major model with Smart Spectrum AI and YunZhisheng in May of last year, in April of this year, it also open-sourced its newly developed big model, 360 Smart Brain 7b, on GitHub, which currently has 7 billion parameters.

According to BofA GLOBAL RESEARCH, the performance optimization speed of the open source model is currently faster than the closed source model. It is inevitable that the performance of the closed source model will be matched by the open source model, so looking at scenarios in the medium to long term is better than making a model.

Taking the collaboration between 360 and Yunzhisheng as an example, 360 can not only obtain a large voice model, enhance 360's intelligent brain's ability to achieve voice interaction, but also consider Yunzhisheng's leading position in the field of smart living and smart medical care in China. It can also obtain data on medical and home scenarios, improve understanding and understanding of more racetracks, and help expand the boundaries of commercial applications. Whether C-side users will pay for AI agents in the future is yet to be verified, but the B-side is regarded as the ultimate foothold.

Overall, the 360 layout is ambitious, and the future is promising.

Epilogue

At present, the company has obtained some results, tasted some sweetness, and achieved initial self-certification. Of course, this is just the beginning. With the deepening of its layout at all levels, fundamentals are expected to improve further. Considering that 360's stock price has also been drastically adjusted along with the market, 2023 is likely to become an inflection point for it to regain its upward trend.