Tencent closed 2025 with solid headline numbers — revenue rose 14% to RMB 7,517.7bn and adjusted net income climbed 17% to RMB 2,596bn — even as quarterly profit dipped on seasonal and strategic spending. Management framed the year as the company’s most AI‑intensive to date: every major business line cited AI as a growth driver, from games and advertising to cloud and enterprise services.
The company’s capital expenditure story is instructive. Tencent’s 2025 capex totalled RMB 792bn, roughly in line with the prior year, and Q4 alone accounted for RMB 196bn as the firm continued to invest in AI infrastructure. That restraint contrasts with the large, visible capex surges seen at Microsoft, Google and Amazon, and Tencent executives attribute the difference largely to a worldwide GPU supply squeeze. To bridge the gap, Tencent leased external compute and curtailed third‑party sales of its own capacity so it could prioritise internal AI needs.
Tencent executives also committed to a near‑term step‑up in AI spending. President Liu Zhiping said last year’s RMB 180bn of AI product investment will at least double in 2026. The company is not treating base models as a binary success factor; instead it has reorganised AI teams, built new pre‑training and reinforcement‑learning infrastructure and is privately testing a new flagship model, Míngyuan (Mixed Yuan) 3.0, which Liu described as the largest jump in capability in Tencent’s model lineage.
Those investments are beginning to show in the P&L. Advertising grew faster than industry peers, Tencent Cloud reached scale profitability — a result management directly linked to rising enterprise AI demand — and gaming remained a cash engine. Domestic game revenue rose 18% to RMB 1,642bn and international game revenue jumped 33% to RMB 774bn, helped by Supercell titles, PUBG MOBILE and new contributions from the homegrown hit Mingchao, pushing Tencent’s overseas games business past the US$10bn annual threshold.
But management was candid about short‑term trade‑offs. Fourth‑quarter profit fell 8% quarter‑on‑quarter to RMB 583bn amid seasonal gaming weakness, a spike in sales and marketing spend driven by AI‑native app promotion and game launches, and foreign‑exchange headwinds. Tencent’s RMB 1bn Spring Festival spend to promote its new Yuanbao app was pointed to as a deliberate rehearsal for mass marketing; the app has surpassed 100m monthly active users, but it still trails rivals in retention and engagement.
Public‑facing experimentation with so‑called “yang xia” or “raising‑shrimp” apps — light, scenario‑focused agents that chain LLM capabilities into vertical experiences — has become emblematic of Tencent’s strategy. CEO Ma Huateng argued these small agents spread AI into many hands and scenarios rather than concentrating it within a single chatbot. He also suggested a hybrid approach that blends WeChat’s historically decentralised mini‑program model with some centralised capabilities to preserve partner economics and avoid “short‑circuiting” third‑party developers.
WeChat is where Tencent’s strategic bet is most visible and most fraught. The company is piloting multiple AI features inside WeChat as a staging ground for a future WeChat agent, which management sees as an ideal commercial and distribution arena given more than 1.4bn users. But executives admitted the technical, privacy and operational challenges are substantial: protecting user data, supporting real‑time agent operations at massive scale and integrating a host of ecosystem partners are unresolved issues with no firm launch timeline.
For international observers the story is twofold. Tencent is moving from an infrastructure‑limited posture into aggressive productisation of AI across monetisable businesses, leveraging games, ads and WeChat’s distribution to monetise models rather than compete as a pure cloud‑compute landlord. That model could make Tencent a defining force in how AI is packaged for consumers in China. However, the compute constraint and the tighter capital ramp compared with Western hyperscalers suggest Tencent will rely on a mix of rented capacity, selective capex and ecosystem distribution to compete — a strategy that carries both upside (faster product go‑to‑market) and risk (dependence on external supply and higher near‑term marketing costs).
Investors will watch 2026 for confirmation that higher AI investments translate into sustained revenue growth and improved unit economics rather than persistent margin pressure. If Tencent’s hybrid approach — heavy productisation, prioritised compute allocation and WeChat‑centred agents — works, it could define a China‑specific pathway to mass AI adoption. If compute shortages or partner frictions persist, Tencent may face a more conventional scaling challenge against Alibaba, ByteDance and global cloud incumbents.