Summary of Key Points
Tencent is shifting from using AI to enhance its existing businesses to building a new Tencent empowered by AI. In the second quarter of 2026, it invested 52.8 billion yuan in capital expenditures (mainly for purchasing computing power). AI has begun to impact the company's profits, but the successes are scattered across various areas such as advertising, gaming, cloud services, and office tools, and a systematic cycle of “model → task → revenue → reinvestment” has not yet been established. The greatest risk is not that AI does not generate profit, but that it merely makes existing businesses like gaming and advertising more efficient without providing access to new service scenarios (for example, users do not initiate requests through Tencent’s AI services). As a result, Tencent could end up becoming just another “toolbox” for other companies.
1. The 52.8 Billion Yuan in Capital Expenditures: What is Tencent Betting On?
This investment is not a one-time expense but rather the purchase of long-term assets (such as GPUs and data centers) that will depreciate over time. Tencent has the confidence to spend so much because:
- Existing businesses are still profitable: Gaming, advertising, and fintech continue to generate revenue, with operating profits of 75.6 billion yuan in the second quarter of 2026, providing funding for AI experimentation.
- There is a safety net: If Tencent doesn’t use all the computing power, it can rent it out to other companies for a profit (management estimates a 30% margin on rentals).
However, this investment is not without return; it buys “time” to connect models, products, and ecosystems rather than merely strengthening existing businesses.
2. Tencent’s AI Products: Which Have Successed, and Which Are Still in the Works?
Tencent’s AI products have mixed results:
- YuanBao (consumer AI): Promoted with a 1-billion-yuan red envelope campaign, it reached 50 million daily active users, but user retention rates without subsidies are not publicly available. Its scale is smaller compared to other products like DouBao and QianWen (QuestMobile data as of June 2026: DouBao has 380 million monthly active users, YuanBao has 49.84 million).
- WorkBuddy (office intelligence tool): Leads in desktop office AI with 20.97 million monthly visits and is now available for a monthly fee of 99 yuan, but the revenue generated is not disclosed.
- CodeBuddy (programming AI): Used by 95% of internal engineers, reducing coding time by 40%, but its market share is only 6.9% (ranking fourth).
- WeChat Mini Programs: A critical experiment for creating a “task entry point” for users, but it is still in limited testing with no public data on user scale or task success rates. The challenge lies in convincing merchants to use the AI services (for example, will using CodeBuddy to book a car result in losing customer relationships?)
- Hy3 Model: Affordable (1 yuan per million tokens) and effective for handling long texts (up to 256K characters), but it is not the best in the industry and not a preferred platform for external developers.
In summary, there are some highlights, but no single product has yet become a cornerstone of the “new Tencent.”
3. The Biggest Challenge Is Not Technology, But Power Balances Within the Organization
AI development at Tencent is not the responsibility of one department; instead, multiple executives oversee different aspects, leading to clear conflicts:
- Ma Huateng: Oversees strategy and funding, aiming to accelerate AI progress while preventing disruptions to user experience (e.g., opposing the “AI everything” approach).
- Zhang Xiaolong: Manages WeChat, focusing on protecting user privacy and keeping products simple, which slows the adoption of new features (he believes a product used by a billion people should not be overcomplicated).
- Yao Shunyu: Works on models, aiming for better performance while ensuring user satisfaction (complex models require balance between efficiency and accuracy).
- Tang Daosheng: Handles commercialization, seeking to generate revenue from corporate clients while balancing the growth of consumer products.
These conflicts are not about who is right or wrong, but about how to create a seamless cycle of “model → product → revenue” (e.g., model and product teams need to collaborate).
4. Is WeChat an Ace, or a Shackle for AI?
WeChat, with its 1.439 billion users, mini-programs, payment services, and enterprise solutions, is theoretically the best entry point for AI. For example, if a user asks to book a flight, WeChat Mini Programs could directly use relevant services through WeChat Pay. However, WeChat’s emphasis on simplicity creates constraints:
- AI needs access to chat records and functionality, but privacy must be protected.
- External intelligent tools (like DouBao) may trigger security issues when interacting with WeChat (for example, DouBao was discontinued at the end of 2025 due to such concerns).
- The ecosystem for WeChat Mini Programs is not yet well-defined: how do merchants join? How are services ranked? Who is responsible for errors?
If WeChat cannot address these issues, it may become just another tool for users, with them turning to other AI solutions first, and Tencent losing its ability to understand user intentions.
5. The Most Dangerous Form of Success: AI Making Existing Businesses More Efficient
The most likely dangerous outcome is that AI makes existing businesses (advertising, gaming, cloud services) more profitable, but without providing new entry points for users or creating unique competitive advantages. This would mean that while Tencent’s businesses perform better, it fails to establish a new model-based business model. It would still be a profitable giant, but it would have missed the opportunity to transform into a platform leader.
To prove the arrival of a “new Tencent,” the following signs are needed:
- Models and products can complement each other (e.g., user feedback from WorkBuddy improves Hy3’s performance).
- New products generate stable revenue (e.g., WorkBuddy has recurring subscriptions, YuanBao has natural user retention).
- The WeChat Mini Programs ecosystem is well-defined with merchant participation.
- The investment in computing power pays off (AI revenue growth exceeds depreciation costs).
Only when these conditions are met can Tencent truly transform into a “new Tencent” empowered by AI. Otherwise, AI will merely serve as an accelerator for its existing businesses, not the driving force for future growth.