Why Tencent Is Giving Up 30% Immediate Profit to Bet on Its Own Large‑Model AI

During the Q2 earnings conference call, Bernstein analyst Robin Zhu asked Tencent when it expects a tangible return on its $53 billion capital expenditure. In response, senior executives presented data showing the intense demand for AI compute and outlined two divergent paths.

Short‑term rental versus long‑term model development : Chief Strategy Officer James Mitchell noted that renting out compute could recoup depreciation almost immediately, and President Martin Lau cited offers to lease GPU capacity at a 30% premium over previous purchase prices. Despite the lucrative short‑term opportunity, Tencent rejected the rent‑out model.

Commitment to self‑developed large models : Tencent will channel the majority of new GPU resources into training and iterating its proprietary Hunyuan models and AI applications, believing that a dominant, home‑grown model ecosystem will generate far greater economic returns over time.

Commercialization pathways : The company highlighted two AI agents driving revenue:

WorkBuddy – an agent cluster that plans, executes, and delivers complex tasks end‑to‑end, monetized via token‑based billing.

CodeBuddy – a code‑generation tool that boosts internal development efficiency and accelerates customers’ cloud migration, feeding back into strong growth for Tencent Cloud.

Additionally, Tencent reported that its GPU fleet already yields significant direct monetization when used for personalized advertising.

Hunyuan model roadmap :

Hunyuan 3 – a 295‑billion‑parameter open‑weight model released in July, described by Lau as only a “very small starting point”.

Hunyuan 4 – upcoming, promising a leap in parameter count and overall performance, with a bespoke product ecosystem designed to outperform solutions built on generic models.

Hunyuan 5 – planned for a future release, targeting state‑of‑the‑art performance as the flagship model.

Q2 financial highlights :

Total revenue: $30.3 billion, up 11% YoY.

Net profit: $10.3 billion, up 9% YoY.

WeChat/Weixin MAU: 1.349 billion, with 7 million new accounts each month.

Advertising revenue: +22% YoY; domestic gaming revenue: +17% YoY.

Despite solid results, investors remain cautious about the long‑term AI narrative; the stock slipped about 3% after the release, reflecting concerns over high capex and uncertain short‑term AI monetization. The article concludes that Tencent’s willingness to forgo immediate compute‑rental profits in favor of building a proprietary AI stack offers a valuable case study for CTOs and technology‑focused CEOs.