11/20 2026
386
Over the years as a tech media professional, I have witnessed many of Tencent's strategic adjustments—for instance, its 2018 bet on the industrial internet, and its 2021 decision to incorporate "promoting sustainable social value innovation" into its three major investment directions.
Yet none of Tencent's strategic transformations has been as fierce and profound as the one unfolding before our eyes. And the theme this time is nothing other than AI.
On November 11, at Tencent's 28th anniversary LinkTime event, the company released rare public internal remarks from Pony Ma and Martin Lau, stating that they must "break down hilltop territories and start a new venture in the AI era," and proposing that "half the people guard core businesses, half the people fight new AI battles." In the days that followed, discussions about Tencent's rapid AI transformation swept across the Chinese internet, sparking various interpretations and endless controversy.
The reason this caused such a stir is that, in everyone's memory, the outside world once regarded Tencent as a company slow to act on AI—known for its steadiness and caution. Suddenly, the calmest mobile internet giant became the most radical AI reformer, and anyone would ask: what exactly happened?
But if we pay a little attention, we will find that this high-level positioning was not groundless. Over the past year or so, systematic changes in technology, products, and operating methods around AI have been taking place inside Tencent. This fierce statement at the strategic level is more like confirming and safeguarding the rationality of the entire system's transformation.
Since changes have already occurred, what exactly has happened inside Tencent?
A saying is popular on the internet: an egg, broken from the outside, is food; broken from the inside, it is new life.
What Tencent is going through may be exactly like this: the AI era is like an egg. How to open this egg—it is time to make a choice.
For many years, whether in technological upgrades or market actions, Tencent's performance in AI has been more about following along. That is, not falling behind, not overtaking, not overdoing. Tencent would not leave any AI opportunity blank, while at the same time it did not go all-in on any AI product or technology.
From the outside, the deepest motivation for this choice is that Tencent has an extremely high mobile internet moat, with social and gaming comfort zones. It is as if wrapped in a thick eggshell made of user numbers, ecosystem barriers, and cash cows, able to ignore wherever the AI wind outside is blowing.
But although life inside the eggshell is safe and comfortable, one can also perceive that the temperature outside is rising bit by bit. The AI wind may be turning into some kind of roasting. The most representative point is that the rise of the Agent paradigm has shaken the mobile internet moat from the product logic. If AI can directly take over computers and phones, bypass apps and directly open services and deliver results, then those "entrances" that were crucial in the mobile internet era become passages through which Agents flash by.
As the certainty of AI moving to the Agent model rises day by day, the comfort zone of the mobile internet will also wither day by day.
So, wouldn't it be enough to proactively embrace the new AI paradigm? In theory, indeed so, but how difficult it is in practice. It is not that Tencent did not do AI; on the contrary, it invested very early in AI and laid out an extremely broad footprint. But the problem is that most of the business systems Tencent retained achieved considerable success in the mobile internet era. No one wants to throw away the cake already in hand and risk everything to chase uncertain AI, and coupled with barriers between departments, this organizational chronic problem magnifies the difficulties brought to AI transformation.
While the organizational structure does not fit AI, the product and market models Tencent trusted in the mobile internet era are also failing. Tencent made many mild AI products with unclear features, but most results were mediocre. Most notably, during the Spring Festival, Yuanbao quickly accumulated 100 million monthly active users through traditional promotion methods such as red envelopes, but soon fell back to more than 30 million.
The convergence of the above conditions made Tencent see that it was time for a bone-deep transformation. Fortunately, this transformation is happening in a relatively positive period. As mentioned above, while Tencent was mildly pursuing AI, the moat of its traditional businesses remained high. According to the Q3 financial report, Tencent's foundational businesses of advertising, gaming, enterprise and financial services were all growing rapidly. Tencent Cloud service revenue increased 35% year-on-year; Tencent fintech and enterprise services revenue increased 9% year-on-year to 63.66 billion yuan; Tencent marketing services revenue increased 19% year-on-year to 43.05 billion yuan; Tencent gaming revenue was 66.5 billion yuan, of which domestic gaming revenue was 47.6 billion yuan, up 11% year-on-year.
This allows Tencent to directly convert cash cows into AI ammunition, thereby forming a positive cycle between its strong businesses and future strategic direction. There is no need to take drastic measures and fight with backs to the wall, and no need to rely on massive external resources. This is also why Tencent's AI transformation this year has been unprecedented in intensity, yet external perception is mediocre, and capital market feedback remains relatively stable.
Since AI must be done, then do it while the foundational businesses are sound and stable. What drives the change now is anxiety and strategic insight, but what is held in hand is initiative and the cost of trial and error.
In other words, pecking the eggshell open by itself now is painful, difficult, and uncomfortable, but what flies out is most likely a bird. Of course, one can also wait longer, and even if the temperature outside the shell rises, it will still be warm inside. But when the eggshell is opened from the outside, the ending may be a fried egg.
This is why Tencent Culture's article "The General Office Calls for a New Venture in the AI Era, 'Reset Mindset to Zero'" says, "The wave has already come crashing down. Some core businesses look firmly positioned, but AI is advancing rapidly, and the distant wave is already swallowing hill after hill. You cannot still sit there admiring the view."
The next question is, how will Tencent break the shell from within?
The top leadership's call for "a new venture in the AI era" was not a sudden impulse, but the result of nearly a year of exploration and layout, finally finding the right time to concentrate and release strategic will.
This transformation, which takes Tencent from a mobile internet giant with a "reset to zero" mindset toward an AI technology company, unfolds in interlocking ways at the following levels.
First is technology. A transformation of astonishing intensity always starts with technology. Because technology has always been the safest part with the lowest cost of trial and error, and at the same time the prerequisite for all change to begin. Judging from the current situation, the key step in Tencent's technical reserve for AI transformation is the arrival of Yao Shunyu and many top AI talents.
According to various sources, after Yao Shunyu took the position of chief AI scientist at Tencent's "CEO/President's Office" at the end of last year, what began was a technological layout with "rebuilding" as its underlying tone. The AI Lab, established nearly ten years ago, was dissolved and merged into the Hunyuan system; the large language model department and the multimodal model department were merged to establish the foundation model department; Tencent's AI Infra was completely restructured. These thunderous moves almost overturned and redid Tencent's accumulated AI technology system.
This released a clear signal inside Tencent: since it is not good enough, why not start over?
During Yao Shunyu's nearly one year in office, Tencent's achievements at the AI foundational technology level have also been very notable. Hunyuan Hy3 preview, Hy3, and the official version of Hy4 were launched successively. Among them, Hy4 ranked first in China in 9 of 12 scenarios in office scenario evaluations, ranked first among global open-source models in Agent Arena after release, and ranked second among Chinese models in the Artificial Analysis Intelligence Index. This has allowed Hunyuan, which had long been in the following tier, to gradually move toward a qualitative change, and Hunyuan's priority is rising in both Tencent products and external products.
The additional value in the process of technological reconstruction is that the flat management, emphasis on business collaboration, and organizational model valuing emerging talent promoted by Yao Shunyu began to gain trust.
This technological force, containing determination to rebuild, the possibility of leading, and a younger organization, began to spread inside Tencent and became the first sound of pecking open the eggshell.
What Tencent is best at is products, and this should be able to become a consensus. But how to bring product advantages back into play in the AI era? Today, as AI applications are just moving toward diversification, this is actually an unsolved puzzle placed before the whole world.
During this period oriented toward AI transformation, Tencent's choice is to let products "run" first. This running includes quickly launching products adapted to the new AI paradigm, and also breaking the barriers between traditional technology and products so that the model side and the product side can run together. At the same time, it also includes letting products run with products, crossing so-called hilltops, and completing product collaboration in a more efficient way.
The reason WorkBuddy could run so fast in the market is also because it is an AI native product born under Tencent's "small steps, quick run" model.
Eight months after launch, WorkBuddy iterated more than 100 times, becoming the single AI product with the largest token consumption in China, with the task resolution rate rising from 72% to 90%. According to the latest released data, WorkBuddy's daily active users have exceeded 10 million, with average daily token consumption exceeding 100 trillion. In order to achieve the rapid climb of product capability, the Tencent Hunyuan team sent experts to WorkBuddy's product team, completing a sharp AI product surprise attack by breaking boundaries and moving in small steps, quickly.
This staged success proved that Tencent's original product model of each fighting its own battle must change, and also can change. As long as product efficiency is raised and product teams are set running, it will be found that Tencent's dazzling array of product accumulation is a natural AI rich mine. Experts in finance and gaming can feed back into foundation model research and development, and AI office experience can be extended to advertising, gaming, social, and other fields.
You should know that the biggest problem for so-called AI native enterprises is holding the hammer of large models without knowing where to find the nails of application scenarios. In Tencent's hands are nails one after another, naturally possessing the solutions AI craves.
There is a judgment inside Tencent that AI applications are still far from the stage of large-scale explosion. Analogous to the electrical revolution, the current stage is similar to when the light bulb has appeared, but a dazzling variety of electrical appliances remain to be invented.
The fast-running product mechanism is precisely to make Tencent the inventor of these appliances, and also to make product advantages the second sound of pecking open the eggshell.
In this AI transformation, one striking point about Tencent is that its investment and expenditure in AI infrastructure are continuously widening. The Q2 financial report shows that Tencent's capital expenditure in a single quarter reached 52.8 billion yuan, up 176% year-on-year, mainly used for AI computing power procurement and data center construction. By Q3, Tencent's quarterly capital expenditure had reached 65.1 billion yuan, and R&D investment was 32.36 billion yuan, up 226% and 42% year-on-year respectively.
In addition, Tencent has also increased its investment amount in AI infrastructure industries. For example, Tencent is the largest external shareholder and largest customer of Enflame Technology, and has also invested in chip and computing infrastructure companies such as Moore Threads (GPU), Lightelligence (optical computing), Yunbao Intelligence (DPU), and Infinigence AI (AI computing power scheduling).
The procurement, deployment, self-developed technology, and investment system of AI infrastructure all indicate such a phenomenon: Tencent is bidding farewell to the asset-light, light-self-research, platform-heavy, entrance-heavy model prevalent in the internet era, and instead embracing a heavy-asset, heavy-R&D model with a solid computing foundation and a complete self-developed technology system. If saying one embraces the AI era is not entirely credible, then let the flow of money explain everything.
Real gold and silver investment in AI strategy has become the third sound of pecking open the eggshell, and this has made the outside world hear it clearly.
A change at one endpoint does not explain much, so perhaps the collective transformation of technology, products, and strategy can already explain something. The reasons and starting points for the above changes are exactly why we have recently seen Tencent publicly announce its embrace of the AI era.
A powerful strategic adjustment is extremely difficult. It needs to be verified in business, and then through top-down strategic positioning, the verified experience is promoted. This step is the value and necessity of organizational change.
Very rarely, Tencent laid all those problems that previously would most likely not be put on the table: numerous hilltop factions, lengthy middle-layer chains, strong territorial awareness, and grassroots innovation being worn away by middle layers.
And the successful experience gained from AI systems like Hunyuan and WorkBuddy—streamlining middle management layers and enabling small teams to operate with agility—happens to offer the right remedy for these problems. This is why Tencent's General Office directly called out "cadres who act as middlemen, passing messages along while refusing to move, yet still clinging to their turf," explicitly proposed that "there may not necessarily be so many management positions in the future," and mandated small teams as combat units with comprehensive organizational flattening.
Bottom-up experience and top-down determination together constitute the direction of Tencent's organizational transformation in the AI era. This direction carries three implications. First is organizational flattening—shedding bloated hierarchical mechanisms and middle-level leadership to select people who can adapt to new methods and new battlefields. Next is letting guerrilla forces "go up the mountain," breaking the inefficient model of past group-based operations, aligning the organizational structure and ways of working with the AI era's demand for fast decision-making, high flexibility, and high efficiency, and building a combat channel for small teams to accomplish big things. Most importantly, this high-level positioning, delivered with an especially stern tone, is meant to convey the determination that there is absolutely no way back. Tencent must pour all 28 years of its accumulated resources into AI and become an AI technology company—no room for negotiation, no room for reversal, no person or matter can be an exception.
The changes in business and the changes in strategy have now converged. Because changing while comfortable is too difficult, before pecking open the eggshell, Tencent has already chosen to endure all the wind and rain in the sky.
This is the zeroth sound—and also the most critical one.
Technological reconstruction, rapid product iteration, heavier asset commitments, and organizational restructuring with absolutely no turning back. The interlinking of these elements fully shows us how a mobile internet giant is to become an AI technology company.
It is clear that the whole process is extremely complex, with an interlocking logical chain. Probably no internet company has yet undergone such a major transformation. Success or failure, we shall wait and see.
Of course, Tencent's transformation currently has only the intensity of departure, not the height of results. This change is still underway—except that Tencent has used every possible means to show internally and externally that this change must continue.
The reasons for this include anticipation of risks, unease about the status quo, strong medicine for the serious illness of big-company disease, and firm trust in technological change and the next era.
Tencent, once conservative in the early AI era, has reached the turning point where extremes reverse. It chooses to go see the AI sky, to become a bird, rather than stay and slowly turn into a fried egg.
Of course, all the answers are still blowing in the wind.
Recently, a friend working at Tencent shared some new changes with me: many graduates choose to interview at Tencent because of Yao Shunyu, rather than because of the traditional big-company halo. In the Hunyuan team, graduates and even interns may lead a team, rather than in the past when leadership had to follow seniority and years of service. Although the technical team still works serious overtime, after completing a major version, they decided to give themselves a one-week holiday.
This does not sound like an internet giant, but rather like a startup team founded by several university graduates, like the studio of a Bilibili uploader.
Is this what Tencent wants? Is this what the AI era should look like? I do not know. But this is most likely what young people want.
A company's future lies less in believing in AI than in believing in the future.