Introduction: The five stamps on the ticket are now complete, and Tencent has boarded the ship, which has officially set sail. What remains to be seen is where this vessel of the AI era will ultimately lead.

Zhang Fan/Author Lishi Business Review/Publisher

On the evening of August 12, Tencent released its Q2 2026 financial report. The next trading day, its stock price dropped by 3.29%, at one point falling over 4% intraday.

The financial report that caused the stock price to decline is not actually unimpressive—quarterly revenue reached RMB 204.785 billion, up 11% year-on-year, surpassing RMB 200 billion for the first time in history; Non-IFRS net profit was RMB 68.4 billion, up 9% year-on-year. What truly unsettled the market were two figures: free cash flow at -RMB 13.8 billion, turning negative for the first time since listing; and capital expenditures at RMB 52.784 billion, soaring 176% year-on-year, exceeding market expectations of RMB 32.1 billion by a full 60%.

That evening, investment banks were divided. Citi raised its target price to HKD 765, citing "resilience in core businesses and visible results from AI applications," while Morgan Stanley cut its target by 15%, expressing concerns about "stagnant profitability from H2 2026 to 2027." For one company and one financial report, the divergence between HKD 765 and HKD 550 represents a HKD 215 gap in expectations.

The focal point of the disagreement is clear: Is Tencent's massive spending securing a ticket for the AI era, or is it building docks for a ship that may never arrive? To answer this, we must dissect the financial report line by line.

1. One Sum of Money, Two Calculation Methods

During the earnings call, Tencent President Martin Lau first introduced a new way of calculating for investors.

He divided Tencent's business into two segments: mature core operations—gaming, advertising, social networking, and fintech—which generate robust growth and ample free cash flow; and AI-native businesses—self-developed large models, new AI applications, and supporting computing infrastructure—which require substantial upfront investment.

Under the first calculation method, Tencent's financial report appears "ugly": Non-IFRS operating profit was RMB 75.6 billion, up just 9% year-on-year, with the operating profit margin slipping from 38.5% last quarter to 37%; free cash flow turned negative.

Under the second method, the picture changes entirely: After excluding the impact of losses from AI products, Non-IFRS operating profit was approximately RMB 86.1 billion, up 19% year-on-year. The AI business incurred a quarterly loss of about RMB 10.5 billion (RMB 8.8 billion last quarter), reducing the group's profit growth by a full 10 percentage points—but this was a deliberate choice, not a deterioration of core operations.

An analogy: A person earning RMB 1 million annually spends RMB 300,000 on a part-time PhD. Their bank statements may suggest they're "getting poorer," but their salary is actually rising. Tencent is currently in this phase.

The persuasiveness of this "two-part" approach hinges on one premise: AI investments will eventually yield returns, a point management spent considerable time arguing that evening. Lau made a confident remark during the call: "A batch of computing equipment we pre-ordered months ago could now be resold at over 30% profit compared to the original purchase price."

This statement highlights the scarcity of the computing market but also conveys a deeper message: Tencent's computing investments have a guaranteed floor. Even if all AI businesses fail, reselling the equipment would still yield profit. Furthermore, "even if we halt self-developed applications and simply lease computing power, that business alone could be profitable."

A massive investment with a worst-case scenario of breaking even—this is the confidence behind Tencent's RMB 52.8 billion capital expenditure in a single quarter.

2. Why Not Lease Computing Power?

Yet Tencent has no intention of profiting from leasing computing power—at least not now.

Tencent Chief Strategy Officer James Mitchell explained this choice during the earnings call: "If we leased computing power to third parties, we could almost immediately recover equipment depreciation costs. However, we have a different strategy, pursuing a longer-term vision."

Management prioritized computing power allocation as follows: first, training the Hunyuan series of large models; second, inference for WorkBuddy; and third, cloud computing leasing.

This prioritization is itself a strategic declaration. Leasing computing power offers modest returns, while training models bets on future influence over the next decade.

The bet is paying off. The official version of Hunyuan Hy3, released in July, ranks among the top three globally in token consumption according to OpenRouter. Management unveiled an even more aggressive roadmap during the earnings call: Hy4, releasing later this year, will achieve a comprehensive leap in parameters and performance, "surpassing larger-scale competitors," while Hy5 will progressively approach and ultimately reach industry-leading levels. After achieving SOTA, Tencent plans to build a multi-tiered gradient model matrix—models of varying parameter sizes tailored to different costs, scenarios, and product lines, each layer capable of stable profitability.

This roadmap aims not just to "keep pace with the top tier" but to "stand atop it and profit through scale and cost structure."

3. Yuanbao Downgraded, WorkBuddy Upgraded

Beneath the models lie applications. Here, Tencent recently completed a crucial internal competition ruling. Management confirmed during the earnings call that "upon discovering WorkBuddy's emergence, we decisively allocated resources to it while lowering the priority of other new AI products in the portfolio."

Among the "other products" deprioritized is Yuanbao.

To clarify, Yuanbao is no failure—its DAU surpassed 50 million in February, with MAU reaching 114 million, making it one of China's largest AI assistants by user base. Logically, a product of this scale should receive more resources. Yet Tencent chose to focus on WorkBuddy.

The rationale lies in the data. Financial reports show WorkBuddy's PC-end monthly visits reached 20.97 million, ranking first among domestic AI office agents, with a Q1 user retention rate as high as 60%. More critically, its business model is viable—the gross margins of WorkBuddy's paid user and MaaS businesses now match Tencent Cloud's overall gross margin.

On one side is a C-end assistant with massive traffic but unclear monetization paths; on the other is an office agent with high retention and profitability. Tencent chose the latter, with Lau positioning it far beyond a mere tool: "WorkBuddy is essentially a new platform, a highly flexible workspace adapted for general artificial intelligence. Its core value lies in task execution." At its foundation is a scheduling framework that can deploy multiple large models to solve complex tasks—Hunyuan is one core model but not the only one.

This choice echoes Tencent's most famous self-revolution: when WeChat was launched, QQ was still thriving, yet Tencent bet the company's resources on it. The key to the "horse racing" mechanism has never been equal distribution but daring to double down on winners and cut losers once the outcome is clear.

4. WeChat's Second Amplification

If WorkBuddy is Tencent's offensive weapon in the AI era, WeChat is its deepest moat. And this moat is being re-dug by AI.

On June 20, WeChat launched its native AI assistant, "Xiaowei," powered by the self-developed visual language model WeLM. It supports text and voice interactions and can directly operate native WeChat functions—opening mini-programs, sending messages, making WeChat calls, AI searches, and group chat summaries. Currently in beta testing, it plans to expand access in Q3.

This marks WeChat's most significant redesign in six years. But what truly matters is Lau's argument during the earnings call: "In the PC era, QQ was merely a social communication tool. With the mobile internet era came WeChat, which amplified QQ's overall value tenfold. Now, with the AI era, WeChat's ecosystem faces a second massive growth opportunity, evolving into an AI-centric application and ecosystem."

This statement implies a bold thesis: each computational platform shift (PC → mobile → AI) represents an order-of-magnitude amplification of ecosystem value. If this holds, WeChat's AI transformation is not a feature upgrade but a revaluation with 10x leverage.

Lau also outlined a longer-term vision: "In the future, users will simply issue a complex instruction to their dedicated agent, which will automatically complete the entire transaction process. Most WeChat mini-program merchants will deploy dedicated merchant agents, enabling direct connections between merchant and user agents over time."

Imagine this scenario: You say, "Book a four-person dinner for Saturday night," and your agent automatically opens the Dianping mini-program, compares prices, reserves a table, places the order, and completes payment—all without human intervention. Between WeChat's tens of millions of mini-programs and its 1.4 billion monthly active users, humans will no longer serve as the "operating interface."

By then, the commercial value of WeChat's mini-program ecosystem will be reactivated in ways unimaginable today.

5. The Five Stamps on the Ticket

Returning to the original question: Why say Tencent has secured its ticket for the AI era?

Combining the financial report and management's statements reveals a complete picture. Competition in the AI era essentially requires five capabilities: models, applications, computing power, scenarios, and capital. Most companies possess only one or two: OpenAI has models but lacks scenarios; ByteDance has scenarios but is building models; startups have applications but lack computing power. Tencent is among the few with all five.

In models, Hunyuan Hy3 ranks globally in the top three, with clear roadmaps for Hy4 and Hy5; in applications, WorkBuddy leads in retention and profitability, while CodeBuddy tops China's AI programming tools; in computing power, early-locked high-end GPUs already yield a 30% unrealized profit in a tight market, with NPO super-nodes deploying by year-end; in scenarios, WeChat's 1.4 billion MAUs represent the world's largest AI application distribution channel; in capital, gaming revenue reached RMB 65.9 billion in a single quarter (up 11% year-on-year), and advertising grew double-digits for the 11th consecutive quarter—these mature businesses generate annual operating cash flows that fund the AI arms race.

Without any one of these five "stamps," the ticket would be incomplete. Interestingly, four of them—applications, computing power, scenarios, and capital—are not new to this quarter but legacy assets accumulated over Tencent's two-decade history. The AI wave has suddenly given these assets new purposes. Securing this "ticket" is half due to this year's aggressive RMB 100 billion spending and half due to two decades of preparation.

6. A Ticket ≠ the Destination

Of course, this ticket is not a panacea.

During the earnings call, James Mitchell, when asked about advertising growth, unusually took the initiative to temper expectations: "Advertising revenue growth has always fluctuated; we advise against linear extrapolation of any single quarter's performance." This restraint suggests management understands that the 22% advertising growth stems partly from AI-driven ad tools and partly from WeChat Channels' traffic release—the former sustainable, the latter with limits.

The true test lies over a longer horizon. Lau promised that capital expenditures for AI-native businesses would be "primarily concentrated in 2026 and 2027," with inference computing power only added later if "substantial returns" are achievable. This implies that by around 2027, Tencent must demonstrate to the market how AI revenue can succeed core operations; otherwise, claims of "one-time investments" will face credibility discounts. The gaps between Hunyuan and global top models, WorkBuddy's expansion from office to general scenarios, and WeChat Xiaowei's refinement from beta to full rollout—none can falter.

But these are challenges for "reaching the destination," not questions about "boarding the ship."

The 19th-century railway bubble burst multiple times, yet railways transformed America; the 2000 dot-com bubble destroyed countless companies but allowed survivors to dominate the next era. History repeatedly shows that massive capital expenditures during technological revolutions are enemies of short-term profits but the price of long-term admission. The real question is never how much is spent but whether the spending acquires irreplaceable assets.

What Tencent has acquired this time are globally top-three models, a commercially viable agent, computing power locked in until next year, and a WeChat ecosystem on the brink of AI transformation for 1.4 billion users.

The five stamps on the ticket are now complete, and Tencent has boarded the ship, which has officially set sail. What remains worthy of long-term attention is where this vessel will ultimately lead.