Tencent takes the initiative from Alibaba in China's AI agent race

In China’s race for AI services, Tencent has begun to claw back leadership from Alibaba thanks to a new focus on agentic AI — systems that do not just answer queries, but take actions on the user’s behalf. In recent days, the company has launched several products and is preparing a more important move: embedding such an agent directly into WeChat.

Why Tencent has accelerated

Until now, Alibaba looked faster: the company rolled out AI features more aggressively, added users more quickly, and promoted its open-source models more visibly. But market interest has shifted from “which chatbot is smarter” to “who will be first to embed an agent into everyday scenarios.” Here Tencent has a trump card that almost no one else in China has: WeChat, with an ecosystem of mini-apps, payments, and services inside a single app.

In one week, Tencent introduced a lineup of products aimed at the wave of interest in AI agents like OpenClaw. The company is also working on integrating its own agent into WeChat. Such a service should be able to book a restaurant, call a taxi, and carry out other tasks without making users jump across dozens of apps. The launch could come as soon as the next few weeks, if the company is not constrained by a shortage of computing capacity.

Betting on WeChat

Tencent’s main strength is not that it was the first to show an agent, but where that agent will live. Agent systems need access to user data, payments, chat history, and a broad set of applications; otherwise they quickly run into the limits of an ordinary chatbot. Tencent already has all of this assembled inside WeChat, which is used by about 1.4 billion people. So the company does not have to build a new distribution channel from scratch and can instead embed AI directly into a familiar environment.

• QClaw — an agent product for individual users
• WorkBuddy — a tool for work tasks and corporate use cases
• Development of an integration into WeChat — so the agent can act directly from the super app
• Access to payments, mini-apps, and user data within a single ecosystem
• A faster path to mass adoption without a separate “AI app”

Investors are already reacting to this pivot. After the launch of QClaw and WorkBuddy, Tencent shares began to show their best monthly performance relative to Alibaba in almost two years, and the company’s market value rose by roughly $30 billion. For Tencent, this matters not only as a stock-market story: usually reserved founder Pony Ma has personally started promoting the new products through his WeChat account, showing that the topic has become strategic for the company.

“In 2026, AI agents will become one of the key formats of interaction,” says investor

Kevin Xu of Interconnected Capital.

How Alibaba is responding

This does not mean Alibaba has suddenly fallen out of the race. On the contrary, it was Alibaba that set the pace for a long time: it launched new AI products faster, actively expanded its audience, and remains one of China’s leaders in open-source models. The problem is elsewhere: technological leadership has not yet turned into an equally obvious commercial advantage. Against the backdrop of the new wave of agentic AI, investors want to see not only models, but also a clear monetization plan.

Alibaba is trying to respond on several fronts at once. The company is preparing an enterprise agent based on Qwen, which should eventually connect with DingTalk, Taobao, and Alipay. At the same time, it is carrying out a major restructuring of its AI business to focus on profit from these products. But at that same moment the market also saw weaker signals: a prominent model developer left the company, and the reorganization itself looks like an admission that the previous structure was poorly suited to the market’s new phase.

What it means

China’s AI race is rapidly shifting from models as such to ecosystems where an agent can actually perform tasks. Tencent is betting on WeChat and distribution; Alibaba is betting on models and its own commercial infrastructure. The winner will not be the one with the stronger chatbot, but the one that first turns the AI agent into a familiar interface for everyday actions.