Over the past two decades, Tencent has hardly ever lost many races, including social networking applications, gaming, and content. Its strategy of "holding back conservatively at first, then striking later" has proven effective time and again.

But this time, the situation in the AI battlefield is different.

According to the latest report from research firm QuestMobile, ByteDance's Doubao ranks first among domestic AI-native apps in terms of weekly active users, with DeepSeek in second place. Although Tencent's Yuanbao comes in third, the gap between it and the top two is wide. Moreover, Alibaba's Qianwen, which launched not long ago, is gaining momentum and is quickly catching up to Yuanbao, which has been online for over a year.

At Tencent's staff meeting on January 26, AI was the primary topic Pony Ma addressed. He made it clear:   "2025 was a big year for AI, and among Tencent's business lines, the only one currently worth significant investment and spending is AI."

He even unusually admitted that Tencent has been slow to act and that its AI infrastructure is a weak spot.

At the same time, Tencent has resorted to a "cash giveaway" tactic it hasn't used in years: letting Yuanbao distribute 1 billion yuan in red envelopes during the Spring Festival, aiming to replicate the glory of WeChat's red envelope campaign. In addition, it quietly launched the internal beta of an AI social product called "Yuanbao Party," which Ma described as "originally a top-secret project."

Each of these statements and moves may not seem significant on their own. But taken together, they suggest that Tencent, long known for its calm and late-mover advantage, is now shedding its old "restrictions" and preparing to go on the offensive.

This time, will Tencent succeed?

Tencent With "Restrictions"

Ma's speeches at Tencent's annual meetings have always been a window for outsiders to observe the company's strategic shifts.

In 2024, Ma remained cautious when discussing AI; by 2025, he was calling for "integration" and "investment"; and by 2026, he made the rare admission that Tencent had "moved too slowly" and strongly promoted "Yuanbao." This shows that Tencent is shifting from its previous restraint in AI to a more proactive approach.

In fact, Tencent's past restraint in AI wasn't due to slow reactions, but rather its "muscle memory."

What is the most valuable asset of Tencent? It ’ s not the content, nor the games, but WeChat — the social media giant. This is the foundation of its business empire, what Ma refers to as the "pillar of strength."

Many people ask, now that even people on the street are talking about AI, why doesn ’ t WeChat offer any particularly impressive AI experiences? Why doesn ’ t Tencent use AI to completely overhaul WeChat from top to bottom?

This isn ’ t a technical issue; it ’ s a matter of survival philosophy. For a product with over a billion monthly active users, any radical or disruptive change is a high-risk move. What if users can ’ t get used to it or find it annoying?

The strength of Tencent ’ s social network lies in network effects and user habits. Here, technology plays the role of "providing stable support," not "reckless experimentation." It ’ s not that Tencent can ’ t do it, but rather that it doesn ’ t dare to — this is the instinct of those who have something to protect.

In Ma ’ s words, Tencent ’ s style is steady and methodical, moving at its own pace.

This leads to its second characteristic: always aiming to be a late mover, which has become a habit.

At the end of last year, investor Zhu Xiaohu commented on Tencent and the current AI competition: "Tencent has always been following behind. In the past 20 years, it has never burned money to experiment at the front. It always waits until the market is clear, then leverages its strong product micro-innovation capabilities and massive traffic to catch up from fifth or sixth place, to third, to second, and eventually to first. Tencent hasn ’ t made its move yet, so the domestic landscape is still relatively early."

What he means is that Tencent is used to watching others make mistakes and hit roadblocks first. Once the market model becomes clear, it uses its strengths in product refinement and huge user base to overtake competitors.

This approach has proven effective time and again in games, social, and content sectors. It has helped Tencent avoid a lot of early-stage risks and reap the rewards of being a late mover.

During the golden era of internet innovation in China, this was a huge advantage. But this strategy is not well-suited for foundational technology research that requires long-term investment, has uncertain prospects, and doesn ’ t have a clear path to monetization for any specific business.

AI happens to be exactly this kind of field. It ’ s like building a high-speed railway — you have to invest a lot of resources and time upfront to lay the infrastructure, not wait until you need the railway to start researching the technology.

Tencent ’ s past lag in AI ’ s underlying computing power, data systems, and top research talent is a natural result of this business inertia. Pony Ma has also admitted that during last year ’ s adjustments, they realized the core issue might lie in the infrastructure.

So, the "tightening curse" on Tencent ’ s head is partly self-imposed, out of a desire for stability and fear of shaking the foundation; the other part comes from the habitual mindset formed by over twenty years of success with its late-mover strategy.

The "tightening spell" that once served as its protective charm may no longer be effective on the AI battlefield.

Changing Rules

The dramatic shift in rules and visible gaps on the AI battlefield are forcing Tencent to leave its former comfort zone.

The first force pushing it out is the brand-new, almost ruthless set of rules in the AI era.

In past internet competition, model innovation, operational efficiency, and user experience were key, giving latecomers real opportunities. But the core competitive law of AI large models is the "law of scale." Simply put, the more you invest in computing power, the broader your data, and the denser your top talent, the stronger your model becomes.

This is a classic "snowball" game: the greater your first-mover advantage, the more users and data you attract, making your model better, which in turn attracts more investment and talent, and your advantage snowballs bigger and bigger.

If Tencent still wants to rely on "late-mover advantage," it's like waiting until others' snowballs have become small mountains before starting to make its first snowball. This gap is hard to bridge with traditional product micro-innovations and traffic redirection. The risk of falling behind at every step has never been greater than it is today.

The second force is the relentless competitors and market.

According to a report by research firm QuestMobile, during the statistical period from December 8 to 14, 2025, Doubao's weekly active user base reached 155 million, ranking first in the market. DeepSeek ranked second with 81.56 million weekly active users, while Yuanbao came in third with 20.84 million.

Here comes a sobering fact: whether it was QQ in the PC internet era or WeChat in the mobile internet era, Tencent has always been the undisputed leader in user numbers. But with the AI era approaching, Tencent has, for the first time, ceded the top spot — and the gap with the leader is over 100 million active users, equivalent to the distance of six or seven "Yuanbao" in between. In addition, Alibaba's Qianwen, Lingguang, Afu, and other "AI family buckets" are also being rolled out on a large scale.

The market won't wait for Tencent to figure things out. ByteDance is pushing Doubao to the limit with free offerings, turning its AI assistant into a household application; Alibaba is stuffing its "AI suite" into e-commerce, payments, healthcare, and other sectors, building its own closed ecosystem. They ’ re not just competing for technological dominance — they ’ re aggressively educating users and shaping habits.

Users ’ understanding and expectations of AI are being rapidly defined by Tencent ’ s competitors. By the time Tencent slowly polishes its products and enters the stage, it may find that the audience has already grown accustomed to someone else ’ s show.

This time, Pony Ma commented that Alibaba ’ s "suite" model has its pros and cons — the downside being that users may not like an all-in-one package, and not every service within the ecosystem is necessarily the industry ’ s best.

This is indeed true, but on the other hand, this kind of "suite"-style large-scale ecosystem linkage is itself a powerful mechanism for trial and error and incubation. It ’ s like an internal "horse race," where different business lines compete on a unified technical foundation and traffic pool. While users might find it bloated, it does increase the chances of a dark horse emerging.

Tencent is actually very familiar with this logic. After all, wasn ’ t WeChat itself the dark horse that emerged from several internal competing projects at Tencent?

Tencent used to be the top player and biggest beneficiary of this game. But now, in this new AI "horse race," Tencent ’ s "horses" — whether it ’ s Yuanbao or other AI products — don ’ t seem to give people the feeling of a real race.

The third pressure comes from the sense of a shrinking survival space.

What is Tencent ’ s core? Social, gaming, and content. And these are precisely the core areas AI is set to reshape.

AI is creating entirely new forms of social interaction and content, and it ’ s changing how games are produced and experienced. If Tencent can ’ t secure a spot in AI-native applications and ecosystems, its massive traffic could end up merely funneling users to others. The most lucrative and imaginative value will be born in someone else ’ s AI ecosystem.

Even more pressing, competitors are moving beyond software and starting to make breakthroughs at the hardware entry point.

For example, the Doubao phone jointly launched by ByteDance and ZTE was seen as an "AI phone that penetrates deep into the mobile operating system," able to directly call up apps like WeChat. In response, WeChat quickly imposed restrictions. Behind this move lies the risk that WeChat ’ s core social scenarios and data security boundaries could be breached from the ground up by competitors.

This is the most alarming future. So for Tencent, taking off the "tightening curse" is not a compromise — it ’ s a counterattack.

Continuing to wear it and reciting the scriptures of the past may mean being gradually marginalized in the next technological era. Taking it off proactively, though it means facing great uncertainty and investment, at least preserves your eligibility to compete and the possibility of a comeback.

Pony Ma ’ s candid admission of "insufficient infrastructure" is the first step in removing the shackles: acknowledging the problem and facing reality head-on. For a company of Tencent ’ s size, being able to do this in itself signals the arrival of a strategic turning point.

How to Remove the "Shackles"?  

The determination is set, the words have been spoken, but what matters most is action. This time, Tencent ’ s move to remove the "shackles" is not just a slogan — it ’ s a comprehensive effort from the inside out, deploying three major strategies.

The first move targets the organizational structure.

By the end of 2025, Tencent upgraded its large model R&D framework — establishing three new major departments: AI Infrastructure, AI Data, and Data Computing Platform. This is not a simple upgrade, but a complete reorganization: AI R&D resources, data resources, and computing power that were previously scattered across various business groups ( such as WeChat, gaming, and cloud ) are now being extracted and reintegrated at the group ’ s unified strategic level.

The goal is clear: break down departmental silos, concentrate resources, and tackle the toughest, most expensive challenges in foundational AI technology. This is arguably Tencent ’ s most significant tech system restructuring since the "930 Reform" in 2018, sending an unmistakably strong signal: AI is the group ’ s highest strategic priority and must be pursued with the full strength of the company.

The second move focuses on talent deployment.

Structure is the skeleton; talent is the flesh and blood. Tencent ’ s previous "business-first" model wasn ’ t particularly attractive to top-tier, visionary technical talent who require long-term investment. This time, they ’ ve changed.

They ’ ve brought in global AI heavyweight and former OpenAI researcher Yao Shunyu directly into the CEO/President ’ s Office, giving him the title of Chief AI Scientist. This move is telling: the highest decision-making level wants to hear directly from the cutting edge of AI research; for those capable of defining the future, top-tier treatment will be provided. This is a remedy for the previous shortage of top research talent.

The third move targets market strategy.

Technology and talent alone aren ’ t enough — you also need users and a market. On the product side, Tencent ’ s approach has become more aggressive and more "internet-like" than before.

During the Spring Festival, Yuanbao is giving away one billion yuan in cash red envelopes, aiming to replicate the successful strategy of WeChat Red Envelopes years ago. By using the simplest and most direct approach — handing out cash — it seeks to attract users in the shortest possible time, generate buzz, and secure valuable initial users and iteration time for its AI applications. This is Tencent leveraging its most traditional advantage: launching a blitzkrieg with massive resources and operational expertise.

On the other side of the battlefield, the atmosphere is equally tense. Almost as soon as Tencent announced its cash giveaway, Baidu quickly followed suit, announcing that its Wenxin Assistant would distribute 500 million yuan in cash red envelopes. This is no longer a single company making a breakthrough, but a sweeping industry-wide competition for user attention and usage habits.

It ’ s worth noting that Yuanbao Party, this "AI + social" experiment, reveals Tencent ’ s next move. The term "Pai" is similar to a group, with each "Pai" creating an online space where AI and a group of people can coexist.

An internal beta tester told me: "From my experience, Yuanbao Party feels like a naturally sociable group member — you can @ it to banter, share memes, take notes, or even help with work tasks. It ’ s quite personable. But you have to @ it every time to interact, it tends to forget complex commands, and it can ’ t monitor things in real time. So, it ’ s a smart but limited companion."

Currently, Yuanbao Party is still in internal testing, but its direction is intriguing. It ’ s no longer content to be just a tool-based Q&A assistant; instead, it ’ s exploring the possibilities of "AI + social," aiming to turn AI into a new kind of connective link. In a way, this is Tencent combining its social networking roots with a new approach.

With these three moves — architecture, talent, and market — progress is being made on all fronts. The process of shedding its "restrictive headband" will inevitably come with pain and confusion, but at least, it ’ s moving forward.

The meeting is over, the statements have been made. The structure has been adjusted, new talent has been brought in, the new product is in internal testing, and the red envelopes are ready to be distributed.

A Tencent that once relied on late-mover advantage is now shedding its shackles, trying to become a more proactive Tencent.

Next, we ’ ll see if it can become a dark horse in the Year of the Horse.