Tencent also cannot achieve the "impossible trinity": traditional business growth, AI business breakthroughs, and positive net cash flow

11/19 2026 525

Tencent has just released its Q3 2026 financial report: operating revenue grew 8% year-over-year, below expectations; Non-GAAP net profit fell 6% year-over-year, basically in line with expectations; capital expenditure was 65.1 billion yuan, far above expectations; free cash flow was negative 69.1 billion yuan, the worst record since Tencent's listing.

Tencent's core business revenue situation is actually not bad: domestic game revenue grew 11% year-over-year, only partially offset by a 9% decline in international game revenue; advertising revenue surged 19%, fully demonstrating the strong commercialization capability of the WeChat Channels ecosystem. However, the market had originally hoped Tencent could achieve revenue growth of around 10%, but it only managed 8%. Where did it fall short?

Part of it comes from competition: the rise of AI short dramas and AI music continues to create pressure on Tencent Video and Tencent Music; Tencent's non-game social network revenue has basically stagnated, and paid content users continue to decline. Another part comes from the slow monetization of AI businesses: WorkBuddy is very popular, but from what I understand, its revenue is almost negligible and will be unlikely to have any positive effect on Tencent's financial reports within the next few quarters.

In addition, pouring resources into AI businesses will obviously mean traditional businesses receive fewer resources. This quarter, Tencent's operating cash flow was only 2.6 billion yuan, severely affected by prepayments for computing power procurement. Please note that Tencent's (and any internet giant's) AI computing power investment includes two major parts: first, purchasing hardware equipment and building data centers, which is counted as capital expenditure (Capex); second, purchasing computing power services from third-party service providers, which is counted in operating cash flow. Tencent has been very aggressive in both areas, so it is conceivable that businesses unrelated to AI probably cannot get much new resources.

"New AI products," including Hunyuan, Yuanbao, WorkBuddy, CodeBuddy, and Xiaowei, constitute a huge financial burden, dragging down Q3 Non-GAAP net profit by 15.2 billion yuan. I guess WorkBuddy accounts for the largest share. According to the financial report, "(WorkBuddy's) more than 10 million daily users collectively consume over 100 trillion Tokens," and at present the cost of these Tokens is clearly borne mostly by Tencent rather than users. The gray-scale testing scope of Xiaowei is also continuously expanding, and in the coming quarters Tencent's Token consumption will probably become higher and higher. In addition to the expense of burning Tokens, AI-related marketing spending is also large—without overwhelming advertising, WorkBuddy's users could not have grown so quickly.

Can we expect "new AI products" to quickly begin contributing revenue in the short term? No. I think that even by fiscal year 2027, if these products combined can contribute 1-3% of Tencent's revenue, that would already be cause for gratitude.

Can we expect the losses of "new AI products" to narrow quickly? Unlikely, because the battle for AI office is not over yet, Xiaowei is still scaling up, and Tencent also wants to challenge in the AI video field, so there will be more and more places to burn Tokens, and the hope of earning it back is slim.

Can we expect Tencent to voluntarily reduce investment in "new AI products"? Not a chance. Firm investment in AI is the consistent strategic policy of Tencent, Alibaba, and ByteDance; whoever retreats means strategic failure. A few days ago, Pony's internal speech made the meaning very clear: at all costs, even by shrinking traditional businesses, the AI battle must be fought to the end.

What the capital market wants is an "impossible trinity": traditional businesses should maintain healthy growth, preferably with both revenue growth and margin improvement; AI businesses should achieve breakthroughs, not only in users and functions but preferably also in revenue; and free cash flow should preferably remain positive, otherwise dividends and buybacks will be affected.

In the current economic environment, the probability of satisfying the above "impossible trinity" is close to zero. Some may think ByteDance can satisfy it—but ByteDance's capital expenditure is far higher than Tencent's or Alibaba's, and at certain stages may even exceed the sum of the two, so the pressure on its free cash flow must also be great; we just cannot see its financial reports.

U.S. tech giants once achieved the "impossible trinity" for a period of time, with Google, Microsoft, and Meta doing especially well. However, that was achieved under conditions of strong expansion of U.S. domestic consumer demand and relatively loose monetary conditions, and U.S. enterprise users have a very high willingness to pay for AI. Even so, after entering 2026, the "impossible trinity" could no longer be maintained; both Google and Meta have experienced situations in which free cash flow was drained dry; last quarter, Google's core search business growth fell short of expectations, triggering high alert on Wall Street.

Domestic internet giants are almost all "consumer internet giants," whose foundation lies in consumer-facing businesses based on free services, with value-added services paid for by users and advertising as the main monetization methods. In the AI era, the foundation of this business model is being shaken. A friend of mine said: "For AI businesses, free users are meaningless, purely a burden." This statement is somewhat arbitrary, but not entirely unreasonable. After all, ByteDance, long known for making its C-end user services free, was the first among its products to launch a paid membership plan for Doubao. Globally, only a very few AI Chatbots such as ChatGPT have launched ad-supported versions, and the results have not been very good.

Relying on AI for e-commerce shopping guides looks beautiful, but it requires strong domestic consumer demand. This year's Double 11, industry-wide GMV growth was only in the low single digits; how can AI shopping guides create incremental growth? Through cannibalization?
Charging money from enterprise users' pockets also looks beautiful, but China's To B business has always been hard, especially since it is very difficult to collect money from small and medium-sized enterprises, and a large portion of it has gone into the pockets of Claude Code & Codex.

Relying on AI to reduce costs and improve efficiency internally has actually been happening all along, and quite a bit of money has already been saved. Besides, what the capital market wants is not only cost control but also revenue growth. History has repeatedly proven that relying on cost reduction and efficiency improvement cannot conjure revenue growth.

So the above deadlock can only continue. The capital market wants this, wants that, and wants more; internet giants can at most satisfy two aspects of demand, or even only one aspect; and there is something worse—being unable to satisfy any aspect of demand. I will not name names.

Interestingly: unprofitable AI businesses are seriously dragging down the valuations of internet giants, while even less profitable AI startups are enjoying valuations in the hundreds of billions. There is no evidence proving that the latter's long-term revenue prospects can exceed the former's. This is not a paradox but the norm of the capital market, and it has appeared countless times. This phenomenon objectively exists; there is no need for us to judge whether it is right or wrong, but only to respect the logic behind it and try to judge how that logic will change in the future.

Solemnly declare: the copyright of this article belongs to the original author. The reprinted article is only for the purpose of spreading more information. If the author's information is marked incorrectly, please contact us immediately to modify or delete it. Thank you.