Summary of Key Points
In its financial report for the second quarter of 2026, Tencent demonstrated a "gamble-all" approach to AI: Its capital expenditure for that quarter soared to 52.8 billion yuan (a year-on-year increase of 176%), resulting in the first negative free cash flow (-13.8 billion yuan). The AI business alone consumed 10.5 billion yuan in operating profits, halving the growth rate of Tencent's core earnings. Although there are highlights such as WorkBuddy (an office intelligence tool) and WeChat Xiaowei (an AI assistant), their commercialization has not been effectively verified, and issues like data opacity and limited functionality exist. The three main pillars supporting Tencent's AI investment—games, advertising, and fintech—are currently stable, but their long-term ability to sustain these investments is in question. The investment in computing power seems to offer both offensive and defensive advantages, yet it also carries cyclical risks. Overall, this represents a critical bet by Tencent in the changing landscape of AI: using short-term profits to secure future growth potential, with the outcome still uncertain.
Financial Pain Caused by Aggressive Investment
Tencent's aggressive investment in AI is clearly reflected in the financial report figures:
- AI consumes 10.5 billion yuan in profits: Without considering the AI business, the core operating profit growth rate for the second quarter was 19%; with AI included, the growth rate dropped to 9%—equivalent to losing 10.5 billion yuan in profits due to the AI business, halving the growth of its core operations.
- First negative cash flow: Free cash flow (the money available for the company's discretionary use) turned negative at -13.8 billion yuan, the first time this has happened since 2005. The reason is a 176% year-on-year increase in capital expenditure on hardware such as computing power and servers, which reached 52.8 billion yuan. Although Tencent claims that excluding prepaid payments for computing power leaves 37.6 billion yuan, these prepayments are essentially locked-in commitments for future usage and should be considered fixed expenses.
- Rapid depletion of cash reserves: The net cash balance at the end of June was 60% lower than at the end of March, remaining at only 58.2 billion yuan. To fund AI investments, Tencent has also cut costs: although the number of employees increased by 4.2%, total compensation decreased by 0.8%, indicating a reduction in per-employee salary.
In short, Tencent is currently tightening its budget to invest heavily in AI, facing significant financial pressure in the short term.
The Illusory Prosperity of the Application Layer
The AI application layer in the financial report appears promising, but upon closer inspection, it hides several issues:
- WorkBuddy: Data opacity and ongoing costs: Claimed to be the leading office intelligence tool in China, Tencent does not disclose key metrics such as monthly active users, paid users, or revenue. It only provides information on traffic and interaction volumes (similar to a store that only reports the number of visitors without mentioning sales). Additionally, it continues to spend money on promotion, including ads in elevators and on CCTV, with many free users receiving subsidies, resulting in low gross margins. Moreover, competitors like ByteDance and Alibaba are catching up, potentially eroding its lead.
- WeChat Xiaowei: Limited functionality and conflicts with advertising: WeChat, with 1.4 billion users, has great potential with AI integration. However, Xiaowei can only perform basic question-and-answer tasks and cannot integrate deeply with payment or social features (for example, it cannot place orders on behalf of users). Tencent claims these restrictions are for security and compliance reasons, but this reduces the user experience. If Xiaowei becomes capable of completing transactions with a single command, ad revenue from WeChat's social platforms (which is its fastest-growing segment at 22%) could be affected.
- Yuanshuo: Downgraded strategy and high trial-and-error costs: The C-end chatbot Yuanshuo, which was heavily promoted last year, is now rarely mentioned. Tencent plans to break down its functions and integrate them into other products, indicating that the previous C-end approach was unsuccessful, with the resulting costs being reflected in AI losses.
In summary, while the application layer seems active, commercialization has not been achieved, and it faces competition and internal conflicts.
The Short-Term Stability of the Main Pillars
Tencent's ability to sustain its AI investments relies on the cash flows from its three main businesses: games, advertising, and fintech. However, each has potential issues:
- Games: Dependence on product cycles: Game revenue increased by 11% in the second quarter, driven by new titles like "Delta Action" and "Kingdom of Loos: World." However, the industry is in a state of stagnation (global mobile game revenue decreased by 2%), and most new releases this year are mid-range products without major hits. If current games lose their popularity, growth could slow down.
- Advertising: Growth ceiling: Ad revenue increased by 22%, mainly due to higher loading rates on video platforms (currently 4%-5%, compared to Douyin's 15%). However, loading rates cannot increase indefinitely; too high rates may annoy users and impact the WeChat ecosystem. Executives suggest that growth cannot be expected to remain above 20% in the future.
- Fintech: Slowest growth and unstable cloud business: Fintech revenue grew by only 9%. Although the cloud business benefits from AI demand, price increases due to rising hardware costs may lead to pricing wars if hardware prices fall in the future, affecting profitability.
In short, these three businesses can provide funding for now, but none of them can sustain unlimited investment in AI in the long term.
The Reliability of Computing Power Investment as a Safety Net
Tencent claims that "even if AI fails, computing power can be rented out for profit," but this may not be a reliable safety net:
- Cyclical pricing of computing power: GPU prices are high due to supply-demand tensions, but in the next 1-2 years, as overseas production capacity increases and domestic computing power becomes more available, prices may fall. Prepaid commitments for computing power could result in losses rather than profits.
- Tencent Cloud's position in the rental market: Alibaba Cloud leads the domestic cloud market, and ByteDance's火山 engine (Huoshan Engine) is also gaining momentum. Tencent Cloud is in the second tier, prioritizing its own needs for computing power (model training and supporting WorkBuddy), making it less competitive in the rental market.
Therefore, the notion of being "offensive and defensive" with computing power investment is more of a statement to reassure investors than a true safety measure, as the actual risks are significant.
Conclusion
Tencent's current AI investment is a last-ditch effort: It missed the initial wave of AI development due to caution and must now make aggressive moves to catch up. However, the large scale of the investment puts significant financial pressure on the company in the short term, with unclear commercialization prospects for its applications and limited support from its main businesses. The success of this bet depends on whether AI can become profitable in the next 2-3 years. If it succeeds, Tencent will gain a foothold in the new AI era; otherwise, its core businesses could be undermined. For users, products like WeChat Xiaowei and WorkBuddy reflect Tencent's substantial investment in the future, but behind them lies a tense gamble with real financial stakes.