In the new round of AI (Artificial Intelligence) competition, how to transform technological advantages into tangible commercial value has become investors' focal point for tech giants.
On August 13, TENCENT (HK00700, stock price HK$586, market cap HK$5.36 trillion) released its Q2 2025 financial results, achieving revenue of 184.5 billion yuan in the quarter, up 15% year-on-year; operating profit (Non-IFRS) reached 69.25 billion yuan, up 18% year-on-year.
On the earnings release day, TENCENT's stock price rallied throughout the trading session, ultimately closing up 4.74% at HK$586 per share, hitting its highest level since June 2021.
TENCENT Chairman and CEO Pony Ma stated that in Q2 2025, TENCENT continued to invest in AI and benefited from it, with both revenue and non-IFRS operating profit achieving double-digit percentage year-on-year growth.
"We are committed to driving usage of our AI-native application Yuanbao by empowering more application scenarios within WeChat, upgrading our Hunyuan foundation model capabilities, and bringing further AI benefits to users and enterprises."
During the Q2 2025 earnings conference call held on the evening of August 13, TENCENT President Martin Lau revealed that AI has already contributed to TENCENT's existing business growth and monetization in various ways, and the company may explore whether there are other viable monetization pathways for To C (consumer-facing) AI in the future.
Meanwhile, TENCENT will closely track AI development progress. The company has already invested considerably in AI and will continue to do so. However, TENCENT must invest wisely rather than blindly going "All in."
**Continuing AI Investment But Not Blindly "All In," Managing Costs in a Relatively Refined Manner**
Currently, TENCENT's AI applications are accelerating their deployment. Yuanbao has achieved core capability upgrades in search, document understanding, and image processing, and added video account interpretation features, deeply integrating into WeChat, QQ and other scenarios; tools like TENCENT Ima and QQ Browser continue to iterate, with WeChat's AI feature matrix constantly enriching.
TENCENT has already invested considerably in AI and will continue to do so. However, Lau pointed out that TENCENT must invest wisely rather than blindly going "All in." For instance, immediately purchasing large quantities of chips, hiring personnel, or conducting large-scale marketing must be done at an appropriate pace.
He cited Ima as an example, noting it remains an innovative product currently exploring better product positioning and use cases. Therefore, before achieving this, the company shouldn't invest too heavily in marketing expenses for this product.
Yuanbao, on the other hand, is already relatively mature, so considerable resources have been invested in its promotion, especially in Q1, while Q2 focused on product improvements.
Of course, while increasing AI investment, Lau stated that TENCENT is also managing costs in a relatively refined manner. For example, developing and using more small models costs much less than using flagship models, while also further improving inference efficiency.
Regarding To C AI monetization, currently it's difficult to adopt user payment models domestically, although this might be common in US AI tools. However, over time, the company will explore whether there are other viable monetization pathways.
Nevertheless, AI has already contributed to TENCENT's existing business growth and monetization in various ways, so the company can also subsidize To C AI costs through growth in other businesses.
In Q2, TENCENT's R&D investment reached 20.25 billion yuan, up 17% year-on-year; capital expenditure was 19.11 billion yuan, up 119% year-on-year.
An investor asked: "If users increasingly use AI search, bypassing websites and official accounts to directly obtain information from AI systems, would this impact the mobile internet ecosystem?"
In response, Lau believes there hasn't been much impact so far. Direct searches in AI systems only display content without guiding people to visit related pages. If they want to explore a topic more deeply, they will click on different links and articles. For instance, in the WeChat ecosystem, many page views actually come from channels unrelated to search, such as subscription accounts, which are actually the main traffic sources.
**Cloud Business Doesn't Rely on GPU, Will Continue Improving Large Language Models**
Among TENCENT's Q2 revenue of 184.5 billion yuan, marketing services business showed impressive growth, reaching 35.8 billion yuan, up 20% year-on-year, hitting a historical high.
This growth was mainly driven by AI-powered advertising platform improvements and enhancements in WeChat's transaction ecosystem, driving strong advertiser demand for Video Accounts, Mini Programs, and WeChat Search.
TENCENT upgraded its advertising foundation models, improving ad effectiveness across various traffic platforms, and further expanded AI applications in ad creation, placement, recommendation, and effectiveness analysis, thereby improving ad click-through rates and conversion rates, advertisers' return on investment, and driving platform marketing services revenue.
TENCENT Chief Strategy Officer James Mitchell stated that TENCENT is currently satisfied with its advertising business growth rate and still hopes to extend this business's growth runway.
"If AI deployment costs (including GPU depreciation) suddenly rise or become burdensome, it might accelerate advertising monetization pace, but currently we don't think this is necessary."
Purchasing GPU (Graphics Processing Unit) computing resources is an important expenditure for internet giants continuously increasing AI investment. However, TENCENT management maintains the attitude from Q1 earnings call—currently having sufficient chips to train large models and will reduce chip dependency.
James Mitchell also revealed that over the past two to three years, TENCENT effectively adjusted its cloud business by scaling back or even exiting some low-margin, low-value-added business projects. Now the cloud business not only has a more sustainable foundation but also improved supply chain cost competitiveness, enabling TENCENT to achieve rapid growth in enterprise services revenue without over-relying on GPU supply fluctuations.
"So if we have sufficient GPUs, we'll lease more in the cloud, but our cloud business strategy doesn't rely on GPUs."
Regarding large models, Lau stated that TENCENT hopes to continuously improve large language models, achieving higher quality pre-training and post-training. When large language models improve, they will become the foundation for all of TENCENT's AI services, improving search capabilities and productivity.
Meanwhile, TENCENT is also accelerating the development of intelligent agents, hoping they can better integrate with the WeChat ecosystem, which is crucial for WeChat ecosystem's future.
Lau added that the company will closely track AI development progress from multiple aspects, such as how AI can better help existing business growth, large language model performance and quality improvements, and monitoring how many users are using AI applications and their growth trends.
**Disclaimer: This content and data are for reference only and do not constitute investment advice. Please verify before use. Operations based on this information are at your own risk.**
Comments