Beyond the Top Line: What Tencent’s Q2 2026 Earnings Reveal About Its AI Transformation

The engine of the revenue beat was Marketing Services, which grew 22% year-on-year to ¥44 billion, driven by AI-enhanced ad targeting and placement . Domestic Games revenue rose 17% to ¥47.3 billion, fueled by titles such as Delta Force, while FinTech and Business Services grew 9% to ¥60.3 billion . Value-added services overall increased 8% to ¥98.4 billion .

Gross profit improved 13% to ¥118.4 billion, and on a non-IFRS basis operating profit grew 9% to ¥75.6 billion .

The AI Spending Story: CapEx Near Triples

The headline financial number drawing the most analyst attention was capital expenditure: ¥52.8 billion in the quarter, representing a 176% surge year-on-year . The spending was concentrated on computing power, data centers, and AI model training infrastructure. Operating capital expenditure alone hit ¥51.8 billion, a 190% jump .

As a direct consequence, free cash flow turned negative — to roughly -¥13.8 billion — as the cash outlay for servers and facilities exceeded operating cash generation .

The Underlying Business: Stronger Than the Headline Profit Suggests

Tencent provided a revealing counterfactual: if you exclude the revenue, costs, and expenses attributable to its new AI products — primarily the Hunyuan (Hy) model family, the Yuanbao assistant, CodeBuddy, WorkBuddy, and the Xiaowei agent — non-IFRS operating profit would have grown 19% to ¥86.1 billion, compared with the reported 9% . This suggests the core, non-AI business is accelerating, while the AI investments themselves are currently a profit drag that management expects to flip to a return.

Pony Ma’s Three-Layer AI Strategy

On the earnings call, Pony Ma laid out a structured framework for Tencent’s AI push across three layers :

1. Intelligence (Models) — The Hunyuan Hy3 production model, released shortly before earnings, is now deployed across multiple products including the Yuanbao consumer app. The model is available globally via API and through Tencent's WorkBuddy international version. A Hy4 upgrade is planned for near-term release .

2. Applications — Ma cited two early-stage AI tools gaining traction: WorkBuddy (an AI office productivity assistant) and CodeBuddy (an AI coding assistant). Both are seeing rapid user growth internally and externally . Additionally, Xiaowei, an agentic AI prototype inside the Weixin (WeChat) super-app, is now in testing .

3. Infrastructure — The company significantly increased computing power procurement, with newly added capacity prioritized for self-developed large model training and expansion of AI applications. By the end of the year and into 2027, Tencent expects to have sufficient GPU capacity to strengthen its cloud services, including bare-metal GPU leasing and Model-as-a-Service (MaaS) offerings .

Ma expressed confidence in the strategy’s financial logic: "We're comfortable in making significant investments in AI because we can foresee superior long-term returns" . He added that AI is already being deployed across Weixin, games, advertising, and cloud, generating positive early returns . Ma earlier in the year had likened Tencent’s prior AI efforts to a "leaky" ship, adding "We still hope the ship can sail faster" . With the Q2 results, the company appears to be accelerating.

What It Means for Tencent’s Trajectory

The Q2 2026 report presents a company in a deliberate investment cycle: the core gaming, advertising, and fintech businesses are performing well, while AI spending temporarily depresses profit growth and cash flow. Management’s argument, backed by the 19% ex-AI operating profit growth, is that the underlying earnings power is stronger than reported results suggest. The key question for investors is whether the AI spend — now approaching an annualized run rate near ¥200 billion in CapEx — will translate into the "superior long-term returns" Ma promises. The early product traction of WorkBuddy, CodeBuddy, and the Hy3 model deployment are the data points to watch.