On August 12, Tencent Holdings released its second-quarter report for 2026. The financial data showed that total revenue for the quarter was approximately 204.785 billion yuan, up 11% year-on-year. Meanwhile, Tencent's single-quarter capital expenditure hit about 52.8 billion yuan, reaching a new record high, with total capital expenditure in the first half of the year already surpassing the entire 2025 total.
Alongside these investments, the impact of new AI products—including Hunyuan, Yuanbao, CodeBuddy, WorkBuddy, and Xiaowei—on profits has become evident. In the second quarter, Tencent's non-IFRS operating profit was approximately 75.636 billion yuan, up 9% year-on-year but nearly flat compared to the 75.627 billion yuan in the first quarter. Excluding the influence of new AI products, this figure would have been 86.1 billion yuan, up from 84.4 billion yuan in the previous quarter. The gap between the two metrics expanded from about 8.8 billion yuan in the first quarter to roughly 10.5 billion yuan in the second quarter.
With the deployment of 52.8 billion yuan in capital expenditure, the market is asking when the returns will materialize. Tencent's resource allocation may offer a clue: WorkBuddy has received priority funding, while the priority for other new AI products is being lowered. As for when monetization will catch up with investment, the company did not provide specific guidance during the earnings call. Tencent consistently emphasized that this is a long-term business requiring sustained investment, and that economic benefits will gradually emerge over time.
Capital Expenditure Doubled in Q2, Matching the Total of the Past Year
According to the financial report, Tencent's capital expenditure in the second quarter was approximately 52.784 billion yuan, a 176% increase year-on-year and a 65% rise quarter-on-quarter. This single-quarter spending exceeded the total annual capital expenditure for any year between 2019 and 2023. As of June 30, 2026, Tencent's capital expenditure stood at about 84.72 billion yuan, compared to annual totals of 79.198 billion yuan in 2024 and 76.76 billion yuan in 2025.
The impact of capital expenditure on profits is directly reflected in depreciation data. Depreciation on property, equipment, and investment properties increased from 8.334 billion yuan in the first quarter to 9.553 billion yuan in the second quarter, a sequential rise of about 14.6%. During the earnings call, an analyst noted that if annualized based on second-quarter capital expenditure, the figure would exceed 200 billion yuan, raising the question of whether this cost could be offset by additional revenue from AI investments or if it might erode company profits in the coming quarters.
In response, Tencent President Martin Lau stated that the company's capital expenditure is divided into two parts: first, spending on existing businesses, which continue to generate strong cash flow; and second, spending on AI-native businesses, used for model training, inference needs, and building AI computing and cloud services. He described the capital expenditure for AI business as "more like a concentrated investment in this year and next," and said it should not be assumed that such a large investment would occur annually, as the model-building component is more of a fixed cost. Lau also mentioned that if existing computing power were used for Tencent Cloud's leasing services, "the prepayments and computing orders made a few months ago could now be sold with a profit margin of over 30%." However, Tencent has chosen to prioritize using computing power to build its own models, then develop applications, and finally lease out the remaining capacity, aiming to gradually establish "a large-scale, highly profitable, and cash-flow-generating proprietary AI-native business."
As of the end of the second quarter, Tencent held approximately 511.154 billion yuan in cash and time deposits, with net cash (after deducting loans and notes payable) of about 58.2 billion yuan, a significant narrowing from 146.86 billion yuan at the end of the first quarter.
The Impact of New AI Products on Profit Is Expanding
The financial report shows that non-IFRS operating profit, excluding new AI products, was 84.4 billion yuan in the first quarter and 86.1 billion yuan in the second quarter, showing a slight sequential increase. However, non-IFRS operating profit remained nearly flat, at approximately 75.627 billion yuan and 75.636 billion yuan, respectively. The gap between the two metrics widened from about 8.8 billion yuan in the first quarter to roughly 10.5 billion yuan in the second quarter.
With increased capital expenditure and the impact of new AI products on profits, three of Tencent's key indicators showed a sequential decline in the second quarter. Free cash flow turned from 56.7 billion yuan in the first quarter to -13.8 billion yuan (or 37.6 billion yuan excluding prepayments for computing power procurement). Operating profit was approximately 67.276 billion yuan, slightly down from 67.375 billion yuan in the first quarter. Profit attributable to equity holders was 56.022 billion yuan, a sequential decrease of about 4%. Tencent explained that operating cash flow includes large prepayments related to AI computing needs, used to support the upgrade of the Hunyuan model, inference requirements for WorkBuddy and CodeBuddy, WeChat AI project construction, AI capability development across products and services, and meeting growing demand from external customers for cloud services.
Monetization of AI-Native Products Still Needs Verification
The financial report and earnings call repeatedly mentioned that growth in Tencent's financial technology and business services, as well as marketing services segments, is linked to AI-related demand or AI-driven initiatives. AI investments are already yielding returns in transforming parts of the company's existing operations. In the second quarter, revenue from financial technology and business services reached 60.3 billion yuan, up 9% year-on-year. Management disclosed during the call that cloud business revenue growth accelerated from "nearly 20%" in the first quarter to "just over 20%" in the second quarter, driven mainly by AI-related demand. However, they noted that the company is still "overcoming capacity constraints."
Marketing services saw the highest revenue growth, up 22% year-on-year to 43.6 billion yuan. The financial report attributed this to the continued optimization of AI-driven ad recommendation models, the upgrade of the intelligent advertising product matrix "Tencent Marketing AIM+," and the integration of closed-loop marketing capabilities within the WeChat ecosystem. The 22% growth was not entirely driven by AI. During the call, management stated that about two percentage points of this growth came from in-game advertising revenue. The Miniclip mobile game Arrows-Puzzle Escape generates revenue through in-app advertising, which is included in the marketing services segment. Management emphasized that if in-app advertising and game revenue were included in both historical and current periods, the year-on-year growth rate for international game revenue would be four percentage points higher than the disclosed figure.
Beyond existing businesses, the monetization of AI-native products such as Hunyuan, Yuanbao, and WorkBuddy remains to be proven. At the model layer, Hy3 has entered the top tier of the global OpenRouter rankings. According to the OpenRouter leaderboard for the week ending August 13, Hy3's weekly token processing volume was 10.1T, up 89% year-on-year, ranking second on the platform, only behind DeepSeek V4 Flash 0731's 10.5T. The financial report disclosed that, based on token consumption on OpenRouter, Hy3 has consistently ranked among the top three globally since its launch on July 7.
At the application layer, divergence has emerged. While office scenario products have seen paid conversions, consumer-facing applications are still searching for user retention. For the consumer product Yuanbao, according to QuestMobile data, its monthly active users (MAU) in June 2026 were approximately 49.84 million, ranking fourth among AI-native apps, with a 100.9% year-on-year increase. The top three MAU figures were: Doubao with about 382 million, up 172.1%; Qianwen with about 167 million, up 5792.9%; and DeepSeek with about 130 million, down 20.3%. In contrast, the office product WorkBuddy, according to Analysys data, led the Chinese desktop AI-native office intelligent agent platform with 20.97 million monthly visits in June 2026, followed by ByteDance's TRAE IDE domestic version with 12.79 million visits and Alibaba's QoderWork with 7.88 million visits.
Tencent Chief Strategy Officer James Mitchell stated that WorkBuddy's gross margin for paying users is already comparable to the overall gross margin of Tencent Cloud, and the model service gross margin has reached the same level. However, WorkBuddy's overall gross margin remains low due to subsidies for free users. "Domestic token production costs are far lower than commonly perceived and estimated, and the token business can maintain positive gross margins even at current low prices, because the costs are low enough."
The new consumer entry point, Xiaowei, has also begun grayscale testing, driven by the WeLM model. It expanded to the WeChat Moments scene in August, carrying "AI writing assistance" and "AI commenting" features. Capital allocation is also shifting. James Mitchell indicated that WorkBuddy is experiencing explosive growth, and Tencent is prioritizing investment in WorkBuddy while lowering the priority of other products in the new AI product portfolio. Martin Lau also noted that the investment scale for Xiaowei is "smaller than the scale of investment in Yuanbao over the past year."
During the earnings call, analysts continued to press for the timeline of monetization: Will near-term profit growth be minimal, and when will AI-inclusive profit growth outpace the growth excluding AI? In response, Lau said Tencent currently does not intend to provide specific performance guidance. "If we see a great opportunity to build a large-scale and highly profitable business for the future, we will decisively invest." He added that if more computing power is allocated to computing, the company could very quickly generate revenue, profit, and returns. He referred to this as a "fallback plan."
"We firmly believe this is a long-term business, so we will invest with a long-term perspective. Over time, economic benefits will gradually emerge, and at some point, profitability will naturally be achieved," Lau concluded.