The capital market hasn't been this excited by a product rumor in a long time.
On June 2, 2026, news emerged that Tencent plans to integrate AI Agent capabilities within its WeChat ecosystem. Even though the news hasn't been officially confirmed, the market has already cast its vote. On that day, Tencent's stock price soared over 10% in a single session.
Tencent, which had been criticized for its slow pace in AI and suffered a prolonged stock decline, has finally recaptured some investor attention.
A TMT-focused securities analyst noted, "The market is aggressively buying because Tencent is about to overlay a layer of capable 'agents' onto WeChat, which has over a billion daily active users and the densest social relationship chain. The potential could support a full re-rating."
While WeChat's Agent is generating buzz, Alibaba is also quietly preparing for action.
It is understood that Alibaba is accelerating the pace of opening its Agent ecosystem. Recently, its AI model Qwen has announced its entry into high-frequency consumption scenarios like travel and dining. Companies such as China Eastern Airlines, KFC, and Luckin Coffee have begun participating in Agent cooperation tests. Meanwhile, Alibaba's own businesses like Fliggy and Taobao are also gradually integrating Agent capabilities.
Alibaba is arriving with a familiar story: it intends to redo what it has excelled at for the past two decades—matching supply and demand—this time using Agents, setting the stage for a head-to-head competition with Tencent.
A Clash Between Top-Tier Players
The WeChat Agent is moving closer to launch.
A source close to the WeChat Agent project revealed that internal testing was originally scheduled for early June, but the WeChat team is considering further refinement and optimization, potentially delaying the final unveiling. According to reports, there may be several internal test versions of the WeChat Agent.
Meanwhile, outside the gates, Alibaba is making quiet but determined efforts, preparing to open the doors to its own ecosystem.
Over the past six months, Alibaba has been using its own ecosystem for stress testing. Dozens of Alibaba-affiliated Agents for travel, shopping, government services, and more have been connected. Taobao has embedded an AI shopping Agent; its instant retail service is developing an Agent; Fliggy is running a travel planning Agent; and the second-hand marketplace Xianyu is also conducting tests.
An industry professional in the Agent field commented, "In reality, the largest commercial endpoint for AI intelligent transactions in China is Qwen." However, they added, "The supply-side offerings could still be more diverse."
As if on cue, just before the WeChat Agent's debut, Alibaba made its move first.
On June 3, Qwen announced it would fully open its arms to third-party Agents and Skills, allowing enterprises to operate their own branded Agents within the Qwen platform. Initial companies like Luckin Coffee, KFC, Mixue Bingcheng, and China Eastern Airlines are testing Agent services on Qwen and will gradually launch them.
First creating its own successful examples, then expanding into a platform to grow the ecosystem—this is Alibaba's customary playbook.
An executive within the Alibaba system explained, "The essence of an Agent is to improve service quality, solve customer problems, enhance the efficiency of customer service, and ultimately win over consumers to choose this service."
"For instance, using Qwen to hail a ride, you just tell it via voice where you want to go from and to, and that the price should be between 10 to 15 yuan, and the options appear immediately," the person said. "There's no need to open an app and wait for an input box, or fill out various fields repeatedly."
In their view, Alibaba is encouraging players to create more Agents. "The current focus should be on getting Agentic commerce to function within China's broader economic environment. Otherwise, all commerce would just be external coding."
Wu Jia, President of the Qwen Business Group, stated plainly, "What we truly envision is integrating AI into the daily life scenarios of ordinary people." He firmly believes this is something that will inevitably happen in the future.
Placing Tencent and Alibaba side by side for comparison now reveals the most fundamental difference lies in *how* users will use this new product.
Users within WeChat primarily seek social interaction, communication, and relationship maintenance. All tool-like attributes (payments, mini-programs, official accounts) are byproducts of the social scene.
Users in the Qwen app have a purer purpose: to get things done. Booking flights, ordering coffee, checking points, planning itineraries—users come to Qwen not to kill time, but to complete a specific task and leave.
In this mindset, the Agent doesn't need to compete for social attention; it just needs to execute tasks smoothly. This difference will likely determine the divergent future business models of the two products.
Tencent's Challenge: Inserting a New Role into a Social Space
Integrating an Agent into WeChat sounds straightforward, but it might not be as simple as imagined.
The density of user experience within the WeChat product is top-tier globally. A single chat window carries instant messaging, an entry to the social feed, official account subscriptions, video channels, mini-programs, payments, enterprise communication, government services... Any new function squeezed in competes with the user's already muscle-memorized interaction paths.
Furthermore, to a large extent, WeChat's core user mindset is "talking to a person," not "talking to a system." When opening WeChat, the default expectation is that the other side is a specific person—a friend, colleague, family member, or an acquaintance in a group chat.
This mindset is anchored so deeply that even the presence of mini-programs is somewhat suppressed beneath it. Most users enter mini-programs temporarily, pulled in by a specific scenario, and leave after use.
Inserting an Agent into such a product forces Tencent to answer the question: Why would a user, in the same chat window where they talk to friends, turn around and tell an AI, "Help me book a flight ticket?"
Tencent certainly has ways to circumvent this, such as not altering WeChat's main conversational framework and creating a separate AI entry point. But then the WeChat Agent would just become another among many entries, failing to fully leverage WeChat's true traffic advantage. If they do alter the framework, they must bear the risk of disrupting users' established habits.
This is what Tencent truly needs to prove: whether it can create a new, AI-native mainstream interaction within a national-level product that is already mature and saturated with features. The difficulty lies not in technology, but rather in product discipline—deciding what to change, what to leave alone, and to what extent.
The ultimate direction for the WeChat Agent appears open at this point.
It could be a hybrid of social interaction and AI, or it might lean more towards tool-like properties. Regardless of the choice, Tencent must confront the usage habits already formed by WeChat's 1.4 billion users, which are its greatest asset and also its biggest burden.
Looking back, however, Tencent has undertaken similar challenges in its history. WeChat Pay forcefully carved out a financial scenario within a熟人社交 (acquaintance-based social) environment, relying on the unique lever of digital red packets. The video channel entered the already mature short-video赛道 (track) by cutting in from the反向 (reverse direction) of WeChat's social relationship chain.
Tencent succeeded in both these endeavors, but each took three to five years. For the Agent battle, the market has already paid a premium of 410 billion Hong Kong dollars upfront; Tencent will need to deliver tangible results to justify it.
Alibaba's Challenge: Building Commercial Infrastructure for Agents
For Alibaba, it doesn't have a national-level conversational product like WeChat in its hands. This starting point dictates that Alibaba cannot take Tencent's path of "adding an AI layer to a熟人场 (acquaintance field)."
But Alibaba has its own trump cards.
In the view of industry insiders, whether an Agent can help a user book a flight depends on whether airlines are willing to open their flight systems, seat availability, change/cancellation rules, special meal preferences, and other information for the Agent to access, and whether they are willing to allocate resources to operate this link. The same applies to coffee brands,餐饮品牌 (food and beverage brands), and government service providers.
Without this kind of foundational work, Agent commerce cannot function. Large model companies aren't adept at handling payments, merchants wouldn't trust the decisions made by Agents on behalf of users, and users wouldn't dare hand over payment authority.
This is a comprehensive contest involving ToB sales networks, merchant operation systems, and supply chain integration capabilities.
Alibaba has accumulated over twenty years of experience in these areas, from connecting外贸工厂 (foreign trade factories) and overseas buyers in the B2B era, to connecting shops and consumers in the Taobao era, to connecting餐饮品牌 (food and beverage brands) and delivery users in the本地生活 (local services) era.
This ingrained DNA is not something other large model companies can develop from scratch.
Not only is the foundation important, but Alibaba has also taken a step ahead in thinking about the demand side for the Agent era.
A person close to Alibaba gave an example: If a user believes they would only be willing to buy a product priced at 10,000 yuan if it drops to 7,000 yuan, this demand was almost impossible for merchants to perceive in the past. But an Agent can place a长期挂单 (long-term pending order) on the user's behalf. "Notify me to buy when it reaches 7,000 yuan."
When more and more users express similar demands, what merchants see is no longer模糊流量 (vague traffic), but真实购买意愿 (genuine purchase intent). This resembles a form of反向团购 (reverse group buying). Agents might even help merchants discover demand.
In the past, merchants didn't know the user's true psychological price point. In the future, Agents might directly feed demand back to the supply side, giving the commercial world接近实时的需求信号 (near real-time demand signals).
It can be said that in the Agent era, understanding intent often holds the greatest value. And what Alibaba is doing is connecting merchant Agents with consumer Agents.
Looking back now, Alibaba's direction and path seem quite clear.
Over the past year, Qwen has持续投入 (continuously invested) in businesses like AI shopping and instant retail to create样板间 (showcase models) and also conduct stress tests. After streamlining these processes, it plans to open up to all merchants.
From this perspective, Alibaba is building the underlying network for the Agent era. If this story can be successfully told, the valuation of Alibaba and Qwen will likely undergo a new round of重估 (reassessment).
Qwen's current position somewhat resembles that of mini-programs in 2017. The market previously viewed it more as Alibaba's AI infrastructure investment—in other words, a "cash-burning project."
But when daily service conversations exceed 100 million, 130 million users are using Agents to get things done, and 90% of 300 million transactions come through Qwen, it already possesses some of the conditions for a platform valuation.
More notably, the logic of traffic value is changing.
After Qwen opens up to third-party Agent入驻 (entry), it will no longer be just an internal工具型产品 (tool-type product) for Alibaba, but will transform into a B+C双边平台 (two-sided platform). This入口 (entry point) could become a new channel for Alibaba's local services and e-commerce businesses to acquire incremental users, and it might also reshape the business models of Taobao, Fliggy, and instant retail services in turn.
Of course, this path has its own uncertainties: the actual timeline for large-scale Agent adoption, the possibility of Tencent using WeChat to卡位 (block the position), and whether Alibaba can maintain its own节奏 (rhythm).
But one thing is certain: the Agent battle won't be a winner-takes-all story, because Doubao, Qwen, and ChatGPT are not walking the same path at all.
Doubao leverages ByteDance's流量基因 (traffic DNA) to perfect companionship and content; ChatGPT relies on OpenAI's enterprise客户 (clients) to sell tools; Qwen relies on Alibaba's商业网络 (commercial network) to build a platform. These three paths will ultimately occupy different poles in the AI application赛道 (track).
However, when an Agent can remember your preferences, proactively提醒 (remind) you to place an order before the lunch rush,帮你算清 (help you calculate) the optimal redemption plan when your points are about to expire, and替你下单 (place an order for you) the moment your desired price appears, you'll realize that the internet's user relationship of "use and leave" from the past two decades is being replaced by an AI assistant that understands you.
And this process of replacement will most likely unfold simultaneously on Alibaba and Tencent's respective home turfs, in two completely different ways.
Comments