Meta's AI advantage does not lie in the model, but in advertising — Tencent and ByteDance are also following the same path.
Meta's position in the new round of AI competition is completely different from what it was a year ago.
In October 2026, the latest Muse AI agent released by Meta recorded an initial user retention rate of 73%, nearly twice the figure of ChatGPT in the same period. The two flagship models launched alongside, Muse Wonder and Muse Wonder Full, have rekindled investor interest. Wall Street has given positive feedback on Meta's AI progress.
But what we want to discuss is not the models themselves. Meta's real advantage lies hidden in its advertising business. The logic behind this advantage is equally important for understanding the competitive landscape of Chinese AI companies.
Meta's Advantage: Advertising as the Fuel for AI
Analysts at Bank of America, led by Justin Post, wrote in an October 15 report that Meta's higher-performance models will complement and strengthen its advertising business. More powerful models can improve the effects of content recommendation and ad targeting, raise the return on investment for advertisers, and further drive the growth of ad spending.
This logic is not complicated, but the key is that Meta already has a mature business model to accommodate the improvement of AI capabilities. Anthropic relies on enterprise clients, ChatGPT relies on the subscription model and an advertising platform that is still in its early stage, while Meta already owns one of the most profitable digital advertising machines in the world. In the second quarter of 2026, Meta's advertising revenue reached 59.363 billion US dollars, accounting for 97.6% of its total revenue, representing a year-on-year increase of 27%.
The cash flow from its advertising business provides the funding source for Meta's investment in AI infrastructure. Meta's capital expenditure in 2026 is expected to reach 115 billion to 135 billion US dollars. Mark Zuckerberg has explicitly stated that this round of large-scale AI investment is mainly continuously funded by the advertising business.
At the product level, Meta has embedded AI capabilities into advertising tools. Its Muse Image model is integrated into the Advantage+ advertising platform, which can understand complex creative demands, automatically adjust styles and generate variants based on advertisers' existing materials, and automate the ad material production process that used to rely on manual work. As of July 2026, 9 million small and micro enterprises have used at least one AI advertising creation tool, and the annualized revenue scale of the Advantage+ end-to-end delivery tool has exceeded 75 billion US dollars.
Meta's logic can be summed up in one sentence: use the money from the advertising business to support AI, and then use AI to improve the efficiency of the advertising business, forming a positive cycle. OpenAI and Anthropic do not have this cycle, because they do not have an advertising business. Google has an advertising business, but its search advertising revenue is being impacted by AI search, so it needs to complete the AI transformation while holding onto its basic market.
Among Chinese companies, who has the same structure? If we screen by the three criteria of "owning a mature advertising business + using AI to feed back the advertising business + having consumer-facing AI products", two companies in China are worth discussing in depth.
Tencent: Yuanbao in the WeChat Ecosystem
Tencent's financial report for the first quarter of 2026 shows that the revenue of its marketing service sector increased by 20% year-on-year to 38.2 billion yuan. Tencent explained in the financial report that this is mainly due to the upgraded AI-driven advertising recommendation model, which expanded the closed-loop marketing capability within the WeChat ecosystem, thus driving the improvement of advertising performance and the growth of ad unit price.
This description is highly similar to Meta's narrative: AI improves the capabilities of recommendation and targeting, driving the growth of advertising efficiency.
Tencent's AI assistant Yuanbao has been connected to WeChat, appearing in the user's chat list in the form of a contact. Users can directly call Yuanbao for conversations within WeChat, and Yuanbao can also parse official account articles and files. Martin Lau, President of Tencent, proposed at the earnings call in May 2026 that in the AI era, enterprises should focus more on identifying high-value implementation scenarios instead of blindly pursuing the scale of daily active users.
The similarity between Tencent and Meta lies in that both of them have a social ecosystem covering billions of users, a mature advertising monetization system, and are embedding AI assistants into their core social products. But there is a key difference between Tencent and Meta. Almost all of Meta's advertising revenue comes from its own platforms, and the improvement of AI capabilities can be transmitted to the advertising business almost without loss. Although Tencent's advertising revenue is also growing, the ad load rate within the WeChat ecosystem has always been relatively restrained, and the advertising monetization space brought by the improvement of AI capabilities is limited by product strategies.
ByteDance: The Closed Loop of Doubao and Douyin E-commerce
The logic of ByteDance's AI layout may be more similar to Meta's.
Byte started with recommendation algorithms, and built a "model-scenario-distribution" closed loop relying on the massive video data and distribution scenarios accumulated through its content ecosystem including Doubao, Jianying and Douyin. Its Seedance 2.5 video model has been connected to TikTok Symphony, and is open to paying advertisers. It can not only charge through subscriptions for creation tools and model calls, but also expand the value of existing businesses by improving the production efficiency of advertising materials.
The integration of Doubao and Douyin E-commerce is a key step in Byte's AI commercialization. In October 2025, Doubao was initially connected to Douyin E-commerce, and users can directly jump to the product card to complete payment and order placement by talking to Doubao. Douyin Mall started to return sales data to the Doubao side and take a commission on transactions generated therefrom.
This model is very close to the direction that Meta is advancing. Meta Enterprise Agent has been launched globally, with 1 million enterprises using it every week to connect with customers and complete transactions on WhatsApp and Messenger. The logic of both sides is: the AI assistant becomes the transaction entry, and the platform obtains revenue from the transactions.
But there are also differences between Byte and Meta. Advertising revenue is the core source of revenue for Meta, and the improvement of advertising efficiency by AI is directly reflected in revenue growth. In Byte's revenue structure, e-commerce and live streaming account for a higher proportion. Although advertising revenue is important, it is not the whole. This means that Byte's AI monetization path is more diversified than Meta's, but the scale effect of each path may not be as concentrated as Meta's.
A Notable Difference That Needs Attention
The paths of Meta, Tencent and Byte are similar, but a key difference needs to be pointed out.
Meta's AI products and advertising business are two links in the same system. Users interact with AI on Facebook or Instagram, AI recommends content and ads on the same platform, and advertisers place ads on the same platform. The entire chain is a closed loop.
The situations of Tencent and Byte are more complex. Tencent's advertising revenue mainly comes from social products such as WeChat and QQ. Although Yuanbao has been connected to WeChat, the collaboration between AI-driven advertising recommendation and AI assistants is still in the early stage. Although Byte's Doubao and Douyin have been connected, the conversion efficiency between Doubao's user scale and Douyin's advertising revenue remains to be further verified.
The AlphaWise China Advertiser Survey Report released by Morgan Stanley in September 2026 points out that AI is changing the way advertising budgets are allocated, but at least before 2030, it is more like a redistribution of the stock market rather than a new incremental cycle for advertising. 72% of advertisers said their total advertising budget has not changed, and 74% believe that cross-platform advertising spending is further concentrating. This means that the improvement of AI to the advertising business is more reflected in efficiency optimization and share competition in the short term, rather than creating brand new increments.
Advertising Is the Fuel for AI, But Not the Only Fuel
Back to Meta's advantages. Its real moat does not lie in model capabilities, but in the monetization path. While Anthropic and OpenAI are still exploring how to obtain revenue from enterprise clients and subscribing users, Meta already has a proven, large-scale revenue source. The improvement of AI capabilities can be directly converted into the improvement of advertising efficiency, and the improvement of advertising efficiency can be directly converted into revenue growth.
Once this cycle is successfully operated, the capital pressure of AI investment will be greatly alleviated. Meta's advertising revenue increased by 27% year-on-year in the second quarter of 2026, while Anthropic was still losing money in the same period. This is not because Meta's model is stronger than Anthropic's, but because Meta has something that Anthropic does not have: a money printing machine that is already running.
In China, Tencent and Byte have similar structures. They both use the cash flow from their advertising business to support AI investment, and also use AI capabilities to improve advertising efficiency. But they face different constraints respectively: Tencent is limited by the restrained advertising strategy of the WeChat ecosystem, and Byte faces the balance of revenue structure between e-commerce and advertising. Meta's model can be referenced, but cannot be simply replicated.
Mark Zuckerberg said this is a huge bet. But the money he used to place the bet is not entirely his own. This is perhaps Meta's most underrated advantage in the AI competition.
This article is from the WeChat official account "Tech Business", authored by Tech Business, and published with authorization from 36Kr.