11/20 2026
489

LAIKA
November 19, 2026
What exactly is the market punishing Tencent for? There is no consensus anchor.
Editor's Note
Tencent is using current profits and cash flow to buy a ticket to the AI era. No one opposes the direction, but the cost is real, and the market is now repricing that ticket. Believers in the old framework cannot understand the new story and retreat; believers in the new framework are still waiting for revenue to materialize and dare not take heavy positions. In between lies a pricing vacuum, where the stock has no anchor and anyone can knock it down.
Tencent fell 6.55% today (November 19) to HK$348.40, hitting a new low since August 2024. And the decline did not start today. From its November 5 high, the stock has fallen 19% in half a month—and 41.32% year to date.
Just yesterday, Tencent released its Q3 financial report. Judging by the report alone, the company has not collapsed. So why the drop?
Revenue came in at 207.3 billion yuan, up 8% year-on-year, with gross margin holding steady at 56%. The core business has not collapsed.
But operating profit was 59.1 billion yuan, down 7% year-on-year—the first quarterly operating profit decline since the first quarter of 2023. Non-IFRS net profit attributable to shareholders was 66.5 billion yuan, down 6% year-on-year.
Net profit attributable to shareholders of 63.6 billion yuan looks like a "slight increase of 0.8%," but it is actually propped up by investment gains (10.06 billion yuan vs. 2.82 billion yuan last year) and lower income tax—not by operational improvement.
But when you turn to the AI page, the bill is a bit startling.
Third-quarter capital expenditure was 81.4 billion yuan (on an accrual basis; 65 billion yuan on a cash basis), up 527% year-on-year; free cash flow showed a net outflow of 69.1 billion yuan, negative for two consecutive quarters; more critically, the company swung from net cash of 58.2 billion yuan in the previous quarter to net debt of 13.7 billion yuan. This is a turning point in its financial structure. In addition, buybacks in the quarter were only HK$7 billion, down 65% from last year's quarterly average of HK$20 billion—so that support is also weakening.
But thinking one layer deeper, what Tencent is doing now is actually reasonable. The confusion is this: for a company that earns more than 60 billion yuan in net profit every quarter, isn't it natural to use profits for dividends and buybacks to reward shareholders? That is what the market thinks. Tencent's buybacks in the quarter shrank 65% year-on-year, so the message is: "If you don't buy back, I'll vote with my feet."
01
About the Future
But if the future is used to define the present, the question changes.
On the first level, dividends are essentially a concession. Buffett said management should pay dividends only when it cannot find a better use for the money than returning it to shareholders. The subtext of large-scale dividends is: we admit we cannot find a place with higher returns. Obviously, under the current AI wave, there is no such tepid environment at all. If Tencent really distributed all its profits as the market wants, it would be publicly announcing that it has no future. For a technology company to say that, there is a price to pay.
On the second level, many investors previously thought of Tencent as a cash cow, but this metaphor itself is toxic. What the market loves is milking: stable, predictable, a fixed amount every year. But companies that are milked grow old; Kodak and Nokia were once cash cows. Tencent management said it even more bluntly in a recent speech: not investing to embrace AI means being disrupted; this is a matter of life and death. Tencent does not want to be a cash cow; once the milk is gone, there is nothing left.
On the third level, when buying the same ticket, whether the market believes it depends on whether there is evidence that "the ship has already sailed." Microsoft also spends heavily on AI, and the market reaction has been mild because Azure's AI revenue has already materialized; on Tencent's side, with the numbers above (81.4 billion yuan in capital expenditure and a 69.1 billion yuan cash flow gap), what the market sees is a huge bill, not the corresponding revenue. This is the real subtext of today's 7% drop: it is not opposition to investing in AI, but a lack of belief that the investment has made a splash.
So what Tencent is answering now is not a multiple-choice question of "how should the money be allocated," but a survival question of "can we live to the next era." Tencent must think about it, and has already submitted its answer through actions: use current profits and cash flow to buy a ticket to the AI era. But after the ticket was handed over, the market has been slow to take it.
02
No Consensus Anchor
The author is stuck on one point: profits are rising, the core business has not collapsed, so why is the stock still down 40% year to date? And it can be said that almost all technology companies have performed this way this year.
Looking at those sets of numbers over and over, the answer is not in the numbers. Only after figuring out one thing did it become clear: Tencent is still the same Tencent, but the market's temperature has changed—and of course, that is thanks to AI.
Over the past decade, investors measured Tencent with a ruler called discounted cash flow.
How much free cash flow a company can steadily squeeze out each year, discounted back, plus a certainty premium, is its price. Buybacks, dividends, and the high-margin core business are all markings on this ruler. It rewards the visible present; the steadier and more predictable it is, the higher the valuation. Tencent was once the top student in this system.
Now this ruler no longer works. Tencent is using the money originally intended for buybacks to buy computing power, build data centers, and train models. In the eyes of the old ruler, this is a negative: it does not generate current cash flow and only adds depreciation and uncertainty. But on another ruler, this money is an option premium—using today's certainty to buy an admission ticket to the future.
Unfortunately, option pricing does not ask how much you earn today; it only asks whether that growth curve has a chance of materializing.
The gap when switching rulers is the pricing vacuum. Those holding the old ruler see single-digit profit growth and negative free cash flow and choose to leave; those holding the new ruler are still waiting for evidence that the ship has really sailed and dare not catch the falling knife. Buyers and sellers each hold a ruler, and neither can convince the other. A large part of the 41% decline is not a collapse in fundamentals, but a mismatch of rulers.
To end this tug-of-war, only one narrative can work: AI revenue truly grows independently and is seen separately. Once it can be listed separately in financial reports and valued separately by the market, the new ruler will finally have markings.
However, AI is a long-cycle investment of five to ten years, while listed companies must report every quarter. Using quarterly report cards to evaluate a protracted war is itself a mismatch. So this 41% is less about what Tencent did wrong and more about the toll it paid for entering a valuation no-man's land. And the first marking on the new map is already hidden in this financial report.
Tencent specifically highlighted in the report: WorkBuddy has more than 10 million daily active users and average daily Token consumption of 100 trillion—already the most-used AI product in China. The company is actively steering the AI narrative from "burning money" toward "having users and revenue," which is the key point to verify in the coming quarters.
CICC gave a more specific timetable in its commentary: management has made clear that cash-basis capital expenditure is at its peak, may fluctuate around this level in 4Q26 and 2027, and will decline from 2028. As prepayments peak, inference efficiency improves exponentially year by year, and domestic chip supply fills the gap, CICC judges that the investment trend is expected to improve from 2028, and the earnings elasticity and free cash flow recovery are worth watching.
Below is a condensed summary of last night's earnings call:
Tencent Q3 Earnings Call ● Condensed Minutes
I. AI: From "Burning Money" to "Revenue Validation"
Hunyuan Models
Hunyuan 4 ranked first among Chinese models on Agent Arena at launch, without targeted optimization; added visual and computer operation capabilities
The lightweight Flash version is priced at 1/20 of Hunyuan, with usage far exceeding expectations; among models released this year on OpenRouter, it has the highest daily token consumption in China
Multimodal development has been unified under the Hunyuan team; 3D generation is industry-leading, image generation is improving, and video generation is under development
WorkBuddy (the fastest-growing piece)
China's No. 1 AI application by token usage, with DAU exceeding 10 million and daily consumption exceeding 100 trillion tokens
Annualized revenue over the past 7 days exceeded US$500 million, more than twice that of Tencent Meeting; Q3 revenue grew more than 5 times quarter-on-quarter
Enterprise edition launched in June; 60% of China's top 100 enterprises are already customers; some existing users pay up to 40,000 yuan per month for tokens
Tiered monetization: free customer acquisition → basic payment → heavy add-on purchases; Flash models are used for customer acquisition, while complex tasks lead to higher-priced models
WeChat Xiaowei
Tested in June, interaction DAU 30 million, touchpoints expanded to more than 30 (Moments, Official Accounts, Channels, etc.)
Based on a customized WeLM small model, with extremely low cost; even if opened to 1.4 billion users, quarterly cost would be lower than Yuanbao at the beginning of the year
Positioned as a lightweight assistant (not an all-purpose Agent), prioritizing advertising/transaction monetization; capabilities expected to reach Meta Muse level in 12-18 months
Management expects it to bring a 10%-20% increment to WeChat's annual revenue base of tens of billions to hundreds of billions, clearly visible within 6-12 months
Cloud Services
Q3 growth 35% (Q2 was 21%, and 2025 was in the teens), driven by WorkBuddy, MaaS, and GPU rental
MaaS revenue grew more than 3 times quarter-on-quarter, GPU rental grew more than 30% quarter-on-quarter
Chip restrictions have no substantial impact: the main business is efficient model inference, which domestic chips can support; Hunyuan 4 Flash costs 95% less than Hunyuan 4
II. Core Business: Games and Advertising Stable, Finance Dragged by Consumption
Value-added services 98 billion yuan (+3%): domestic games +11% (driven by Delta Force, Valorant Mobile, and Honor of Kings), international -9% (high base); Delta Force peak DAU exceeded 58 million in October
Marketing services 43 billion yuan (+19%): AI advertising solutions have covered about 80% of short-drama publishers, Channels ad impressions +50%; gross margin fell 4 percentage points to 52% (front-loaded investment in technology upgrades + investment in the short-drama ecosystem)
Fintech and business services 64 billion yuan (+9%): driven by consumer loans and wealth management; payment volume increased but average transaction value fell (weak consumption); promoting Tenpay Go for overseas visitors to China
III. Financials: The Peak and Turning Point of Capital Expenditure
Revenue 207.3 billion yuan (+8%), operating profit 59.1 billion yuan (-7%); excluding non-IFRS new AI products, 83.3 billion yuan (+8%); R&D 32.4 billion yuan (+42%)
Cash capital expenditure 65.1 billion yuan (+226%), free cash flow -69.1 billion yuan, +16.9 billion yuan after excluding computing power prepayments; net debt 13.7 billion yuan (still net cash of 58.2 billion yuan at the end of June)
Key guidance: prepayments in Q2 and Q3 were abnormally high, long-term contracts have been signed, and prepayments will decline or even reverse; cash capital expenditure is already at its peak, will fluctuate at a high level for the rest of 2026 and in 2027, and will decline from 2028, with input-output clearly improving
IV. Strategy and Outlook
Organizational adjustment: use AI to reduce staff in existing businesses and free up manpower for new businesses; the Hunyuan and WorkBuddy teams are models
AI and games: AI cannot yet take over mainstream multiplayer evergreen games, but it can accelerate pre-production and promote evolution toward UGC platforms (already verified in Peacekeeper Elite)
Advertising: upside risk in 2027 outweighs downside risk
Criteria for evaluating AI returns: whether valuable businesses and products are created; Hunyuan 4 and WorkBuddy have already validated this; Yuanbao investment has been significantly cut, and search capabilities have been opened externally
- END -