← Back to stories A woman interacts with a productivity app on her smartphone at a desk setup.
Photo by freestocks.org on Pexels
凤凰科技 2026-03-18

Tencent’s QQ may fully back Longxia (龙虾) to build and run in‑app communities, Ma Huateng says

What was announced

It has been reported that Ma Huateng (马化腾), chairman and CEO of Tencent (腾讯), for the first time publicly outlined the company’s “养虾” (lit. “shrimp‑farming”) concept at Tencent Holdings’ 2025 performance briefing on March 18. He said the Longxia (龙虾) application could “land” AI into a wide range of scenarios rather than limit it to chatbots, help integrate Tencent’s disparate business lines, and provide inspiration for developing WeChat (微信) Agents. Reportedly, QQ may be positioned to fully support Longxia in creating, operating and managing in‑app communities.

Why it matters

Why use another app when WeChat already dominates China’s messaging and services? QQ remains a large, distinct social product with different user demographics and content styles, and Tencent appears to be exploring multiple entry points for agent‑style AI beyond the WeChat mini‑program ecosystem. Ma said WeChat mini‑programs have long adhered to a decentralized design and that this thinking could inform Longxia’s architecture, while Tencent seeks to strike a balance between centralization and decentralization so partners aren’t short‑circuited or reduced to mere API calls.

Broader context

This push comes as Chinese tech firms race to build domestic AI stacks and agent ecosystems amid a shifting international landscape of export controls and sanctions that affect access to advanced chips and cloud services. It has been reported that Tencent’s framing — giving partners traffic access while protecting their commercial value — aims to keep ecosystems vibrant without concentrating all control internally. Who benefits? Users may see richer, integrated AI services; developers and service providers will be watching closely to see whether Tencent’s promises mean more traffic — or more dependence.

AISpace
View original source →