Editor: Jiaxin

"Numerous self-media outlets claim that Tencent is lagging behind. We missed some key opportunities in AI. Do you think we're genuinely slow? What will define the next phase of the AI race?"

Recently, at the Tencent Cloud AI Industry Application Conference, Dowson Tong, Senior Executive Vice President of Tencent Group, posed this question to Yao Shunyu, Tencent’s newly appointed Chief AI Scientist.

Yao Shunyu paused briefly before responding, "Actually, that's a question I was going to ask you." He also emphasized that AI is a marathon, not a sprint: "ChatGPT and Claude Code won’t be the only game-changers. New opportunities will keep emerging."

This seemingly casual exchange unexpectedly touched on the core debate over Tencent’s AI strategy in recent years.

Since ChatGPT’s debut, Tencent’s AI efforts have often been labeled as "slow."

Compared to Baidu’s early launch of ERNIE Bot, Alibaba’s aggressive open-source ecosystem play, ByteDance’s rapid user acquisition with Doubao, and the sudden rise of DeepSeek, Tencent remained relatively quiet for an extended period.

This perception even became a consensus in capital markets.

However, on June 2, after news broke that WeChat’s AI Agent was nearing release, Tencent’s stock price surged over 10% intraday, adding roughly HK$414.8 billion to its market cap in a single day. Yet, within just a few trading sessions, the stock price retreated, erasing most of those gains.

Behind this volatility lay the market’s mixed feelings about Tencent’s AI prospects. On one hand, Tencent seemed to have missed the most dynamic phase of the large model era; on the other, it held one of the industry’s most valuable assets—WeChat.

This is the biggest question mark surrounding Tencent’s AI strategy today.

At this critical juncture, we examine Tencent’s AI journey over the past three years, attempting to answer three key questions: Where exactly is Tencent lagging? Can WeChat become its ace in the hole for a comeback? And as the AI competition enters its next phase, where does Tencent stand?

In November 2022, ChatGPT was released, reaching 1 million users in five days and 100 million monthly active users within two months.

At that time, Tencent was in a rare slump since its IPO. For the entire year, Tencent’s revenue was RMB 554.6 billion, down 1% year-on-year; net profit attributable to shareholders was RMB 188.2 billion, down 16% year-on-year. This marked the first time since its 2004 listing that Tencent experienced declines in both annual revenue and net profit.

With stricter game approvals, sustained regulatory pressure, and its stock price having plummeted by more than half from its 2021 peak to the end of 2022, Pony Ma criticized multiple businesses in an internal speech at the end of 2022: "Many businesses should be cut. Don’t blindly follow competitors."

A company focused on cost reduction and efficiency improvement finds it challenging to simultaneously respond at top speed to a technological revolution. This was directly reflected in the rollout of the Hunyuan large model.

Baidu released ERNIE Bot in March 2023, and while the product was still immature, it seized the public’s perception of China’s large model landscape. Alibaba, Huawei, and iFLYTEK followed closely with frequent releases.

In contrast, Tencent didn’t officially launch the Hunyuan large model until September 7, 2023, at the Global Digital Ecosystem Summit, opening it to external access via Tencent Cloud—nearly ten months after ChatGPT’s debut.

Moreover, Tencent’s strategy was to prioritize internal adoption before external release. At launch, over 50 Tencent businesses and products, including Tencent Cloud, Tencent Advertising, Tencent Games, Tencent FinTech, Tencent Meeting, Tencent Docs, WeChat Search, and QQ Browser, had already integrated the Hunyuan large model for testing.

This contrasted with Baidu and Alibaba’s approaches. While other companies vied for market perception as China’s ChatGPT, Tencent chose to first transform its own ecosystem with AI.

This path had inherent rationality but came at a cost. Tencent was nearly absent from the public perception battle over large models in 2023.

Entering 2024, Tencent began attempting to reach consumer-facing (C-end) users.

In May 2024, Tencent Yuanbao was launched, but the results were underwhelming. QuestMobile data showed that as of March this year, Yuanbao’s MAU (Monthly Active Users) was approximately 57.346 million, while ByteDance’s Doubao and Alibaba’s QianWen had MAUs of 345 million and 166 million, respectively.

A telling detail occurred during the 2026 Spring Festival, when Yuanbao launched a costly "Share RMB 1 Billion in Cash Red Packets" social fission campaign, primarily relying on WeChat groups for user acquisition. However, within days of the red packet links flooding everywhere, they were restricted due to triggering WeChat’s external link compliance rules.

The official WeChat account "WeChat Pie" announced: "We received user complaints that Yuanbao’s Spring Festival campaign induced sharing, harassed users, and disrupted the ecological order. The links have been restricted from opening directly within WeChat in accordance with regulations."

This fiasco, to some extent, reflected Tencent’s identity crisis in the AI era: Why can’t a super social platform with 1.4 billion monthly active users spawn a super app for the AI era?

The rise of Doubao was essentially a classical internet-style entry defense battle. ByteDance leveraged its abundant traffic dividends and App factory’s user acquisition genes to forcibly create an independent super entry—Doubao—to anchor its dominance in C-end AI traffic.

In contrast, Tencent’s launch of the standalone App Yuanbao, which encountered growth bottlenecks, was ridiculed by the industry as "begging with a gold mine."

Tencent’s greatest moat is WeChat. Asking users to leave WeChat for another independent dialog box to find AI not only pits its weakness against ByteDance’s strength in user acquisition but also represents a colossal waste of WeChat’s natural scenarios.

Tencent doesn’t need to find an entry point for AI—the entry point already exists, and no competitor can replicate it.

During the Q3 2025 earnings call, Martin Lau articulated a clear vision: "WeChat will eventually launch an AI Agent to help users accomplish many tasks within WeChat using AI."

He argued that WeChat’s ecosystem boasts a powerful communication and social ecosystem with vast data, enabling the Agent to understand user needs, intentions, and interests; a robust content ecosystem, including Official Accounts and Channels; a Mini Program ecosystem covering most internet use cases; and a commercial ecosystem allowing purchases, along with a payment ecosystem enabling nearly instant transactions.

Lau said this would essentially be users’ ideal assistant, understanding their needs and executing all tasks within the ecosystem.

At an internal executive meeting in December 2025, the WeChat team summarized several trends in its AI entry: WeChat must have built-in AI tools independent of third-party systems; WeChat essentially consists of three parts: social interactions between people, information acquisition, and efficiency tools. Among these, social relationships themselves cannot be replaced by AI, which can only add value at the information and efficiency levels.

This statement reflected the WeChat team’s understanding of its own boundaries: AI will not disrupt WeChat’s social core; its value lies in transforming the other two areas.

In 2026, WeChat’s built-in AI Agent accelerated its rollout.

In March, foreign media The Information reported that Tencent was secretly developing an AI Agent for WeChat, classified as top-secret and led by WeChat’s technical head Zhou Hao, reporting directly to Allen Zhang. A gray-scale release was planned for mid-2026, with a full rollout in Q3.

In June, reports emerged that Tencent was launching an embedded AI Agent for WeChat, having completed prototype testing, with compliance approvals for public release expected to begin as soon as this month.

A source who viewed an early demo explained that users could access the AI Agent’s dialog box by swiping right on WeChat’s main interface. In this interface, users could input instructions, and the Agent would automatically invoke WeChat’s millions of Mini Programs to complete tasks such as finding a café based on taste preferences and price requirements and placing an order.

So, as Tencent’s trump card in AI strategy, can the WeChat AI Agent enable Tencent to stage a comeback and influence the AI war among domestic giants?

From market reactions, many investors clearly want to believe this story. However, from vision to reality lies a lengthy engineering process. WeChat boasts a vast number of Mini Programs, and service quality, interface stability, merchant cooperation, payment workflows, and revenue distribution are all non-trivial matters. Meanwhile, future users, merchants, and the platform may all launch their own Agents, risking chaos without proper constraints.

For AI Agents to truly succeed, they require not just an entry point but coordination across the entire chain: stable Mini Program interfaces, merchant cooperation, seamless payment integration, controllable inference costs, and compliant data usage.

Meanwhile, competitors are not standing still. Alibaba and ByteDance are accelerating their expansion into service scenarios. QianWen has deeply integrated with Alibaba’s ecosystem, including Taobao, Alipay, Fliggy, and Gaode Maps; Doubao is also deeply integrating with Douyin’s e-commerce ecosystem.

Pony Ma said at the 2026 Q1 shareholders’ meeting: "A year ago, we thought we’d boarded the ship, but later realized it was leaking. Now we feel like we’re on board but can’t sit comfortably yet. We still hope the ship can speed up." This perfectly described Tencent’s AI predicament.

Tencent spent three years transitioning from observation to commitment. Now, it needs to prove it can make up for lost time in the second half.

Tencent isn’t blind to AI—it places heavy emphasis on return on investment.

Over the past two decades, one of Tencent’s most successful commercial practices has been rapidly amplifying its advantages after recognizing trends. Whether in gaming, payments, or industrial internet, Tencent excels at heavy investment after business models mature, rather than being the first to take risks.

This DNA is also evident in AI.

Over the past few years, while advancing Hunyuan’s R&D, Yuanbao’s launch, and internal business transformations, Tencent has maintained strategic flexibility. Around 2025, Tencent invested in tech companies like Jiyiwei Semiconductor and Xizhi Technology, while also appearing on the shareholder lists of multiple large model companies, including Moonshot AI, MiniMax, and Zhipu AI.

However, Tencent has been much more restrained in capital expenditures.

In 2025, Tencent’s capital expenditures reached RMB 79.2 billion, with R&D investment hitting RMB 85.75 billion, both record highs. In comparison, foreign media reported that ByteDance’s AI-related capital expenditures in 2025 were approximately RMB 150 billion, with about RMB 90 billion allocated to AI computing power procurement.

But this restraint doesn’t mean Tencent underestimates AI. Over the past two years, Tencent has been waiting for an answer to one question: How exactly can AI make money?

For the past two years, the core narrative in the large model industry has been model capabilities. Parameter scale, benchmark tests, reasoning abilities, download volumes, and user scales have dominated industry competition. While models continue to improve, business models remain unclear.

What truly began to change the industry was the emergence of the Agent wave.

Since late 2025, Agent frameworks like OpenClaw have rapidly gained traction. Unlike traditional chatbots, Agents don’t just answer questions—they can invoke tools, connect services, and execute tasks. From querying information to completing transactions, from content generation to automating workflows, large models are transforming from dialog tools into execution systems for the first time.

This means AI now has a relatively clear path to commercialization.

Token invocation, tool usage, task execution, enterprise subscriptions, and transaction revenue sharing could all become income streams. To some extent, Agents have shown Tencent the endpoint of AI commercialization.

This is why Tencent’s response to Agents has noticeably accelerated in 2026. Tencent Cloud quickly launched a one-click OpenClaw deployment service; the WeChat Agent entered testing; and products like Yuanbao, Enterprise WeChat, and Tencent Meeting began restructuring around Agent capabilities.

Reviewing Tencent’s product history over the past two decades reveals a recurring pattern. Tencent isn’t always the first to invent technologies but often succeeds in transforming complex technologies into mass-market products. This was true for QQ, WeChat, WeChat Pay, and Mini Programs.

Today, this logic is being replicated in the Agent era.

If the past two years of competition focused on model capabilities, entering 2026, the focus is gradually shifting toward ecosystem capabilities.

Who can integrate AI into users’ habitual scenarios, retain developers and merchants on the platform, and enable Agents to complete the full loop from understanding needs to executing tasks and finalizing transactions will have a better chance of winning the next phase.

And this is precisely the direction Tencent is betting on.