In the spring of 2025, the AI track experienced a "supernova explosion", with DeepSeek surpassing ChatGPT to become the fastest-growing AI application. Immediately, the developer of ChatGPT, OpenAI, announced in colors that within less than 3 months, ChatGPT had added 100 million weekly active users, and the total weekly active users had reached 400 million.

Undoubtedly, the existing market pattern of AI assistants has been rewritten. As the flag bearer of open-source AI, Meta has finally stopped being a bystander. Recently, there have been reports that Meta will launch an independent AI chatbot application based on the Llama series open-source models in the second quarter of this year to meet users' demand for AI services.

Why does Meta want to launch an independent AI assistant application? It is said that the direct reason is that Meta founder Zuckerberg has set a KPI. He stated in the earnings call at the end of January that "I expect this year to be a year of highly intelligent and personalized AI assistants covering over 1 billion people". Currently, Meta AI serves about 700 million active users per month, still some distance from the 1 billion user scale.

Compared to ChatGPT's 400 million weekly active users and DeepSeek's 60 million monthly active users, Meta AI's 700 million monthly active users actually have a lot of water. Because Meta's algorithm is that Meta AI is integrated into Facebook, Instagram, WhatsApp and Facebook Messenger, and there are 700 million users using the AI services in these Apps every month.

So the question is, are Meta AI's 700 million users the natural traffic of Facebook and Instagram, or is Meta's AI business playing a role? Meta's biggest problem right now is that although they are shouting loudly about their AI slogans, the public's perception of their AI business is extremely vague.

Although Meta's Llama series models are the first open-source large models, they have never had the propagation value of ChatGPT and DeepSeek, belonging to the typical "little secret known only in the circle". Even today, the performance of the Llama series models can be ranked in the first echelon, but they just "can't break out". The reason for this phenomenon is very simple, because "Meta AI" lacks brand awareness.

When mentioning OpenAI, people can think of ChatGPT; when mentioning Google, they can associate with Gemini; when mentioning Microsoft, they will relate to Microsoft Copilot. Although Meta has a huge presence in the AI track, the outside world knows nothing about Meta's AI brand. Although Meta AI does exist, what it is, a large number of ordinary users may not be able to say.

Moreover, an independent AI chatbot application also means that a paid subscription service will be provided. Although Zuckerberg did not disclose the official name, release time and subscription price of this AI assistant, the outside world unanimously believes that Meta will follow the example of OpenAI, Gemini and Anthropic to provide both free and paid versions, with the paid version offering enhanced AI experiences and more advanced features.

In addition, Meta is also urgently in need of demonstrating the commercialization capabilities of its AI business to the outside world. From the fourth quarter of 2024 and full-year financial report released at the end of January, the improvement in advertising placement efficiency driven by AI technology has driven Meta's digital advertising business to achieve revenue of $46.78 billion in the quarter, an increase of more than 20%, but the Reality Labs division directly related to the AI business is still in a loss.

At the same time, Meta is continuing to increase its investment in AI infrastructure. Recently, there have been media reports that asset management giant Apollo is in talks with Meta to lead an investment of about $35 billion to support the latter's construction of a series of data centers in the US, which is also believed to be part of Meta's plan to invest $200 billion to build a giant AI data center project.

Although the rumors of the AI bubble bursting are rampant, an undeniable reality challenge is that if these tech giants cannot achieve tangible performance growth and recoup the huge investments made in AI infrastructure within the next three to five years through large models, the AI industry will really encounter its own "Internet bubble". So how can these AI manufacturers achieve performance growth? Obviously, they need to make users willing to pay for AI.

This is why OpenAI has been working hard to attract users to subscribe to ChatGPT Plus, as well as selling their API services to enterprise customers, because paid subscriptions and APIs, which represent MaaS, are almost the only two business models that have been successfully implemented in the current AI industry.

So for Meta to build an independent AI chatbot application and launch the corresponding paid services is, in a sense, a display of posture, proving that their large-scale investment in AI infrastructure is not just a brute force approach.

[The images in this article are from the Internet]

This article is from the WeChat public account "Three Easy Life" (ID: IT-3eLife), author: Sanyi Bacterium, authorized by 36Kr for release.