Tencent Q1 Conference Call: The effects of AI have manifested in three aspects

Tencent CEO Jack Ma stated that in the first quarter of this year, Tencent's high-quality revenue maintained a solid growth trend. AI capabilities have already made substantial contributions to performance advertising and evergreen games, and increased investment in new AI opportunities such as Yuanbao applications and AI within WeChat.

Q: Regarding Agentic AI. Can management discuss our prospects and positioning in the market compared to peers? Additionally, please introduce our strategies for different AI business models, such as advertising, transactions, GPU leasing, and subscription services.

A: Regarding Agentic AI, this is indeed a very popular concept. The core idea is that AI can help users complete complex tasks that involve multiple steps, require tool invocation, and even connect to other applications. From this perspective, it can be divided into two categories: one is "general" Agentic AI, which anyone can create—a proxy that completes tasks for users in the external world; the other is Agentic AI embedded in the WeChat ecosystem, based on WeChat's unique environment, which are actually different products.

For general agents, we have developed this capability in some native AI products (such as "Yuanbao"). Initially, they could quickly answer questions; then we added "chain reasoning" long reasoning models to handle more complex inquiries; subsequently, they could perform more complex tasks, gradually evolving "embodied intelligence" capabilities to interact with other applications, programs, and even external APIs to assist users—this process will continue to evolve, and their capabilities are not fundamentally different from those of peers' general Agentic AI.

On the other hand, Agentic AI in the WeChat ecosystem has the opportunity to be very unique because it can connect social relationship chains, communication and community capabilities, content systems such as public accounts and video accounts, as well as millions of mini-programs. These components cover information, transaction, and operational capabilities across multiple verticals, thus allowing for the construction of differentiated products that are distinct from general Agentic AI.

As for AI business models: advertising first directly benefits from AI, as AI can significantly enhance advertising targeting effectiveness, leading to increased advertising revenue due to better performance—this is a huge opportunity we have already seen in performance advertising and still have further development potential. Transactions are closely related to advertising: when advertising directly drives transactions, the value of advertising will significantly increase, which is another path for us to enhance advertising revenue.

GPU leasing is essentially equivalent to a redistribution model in cloud business. Given the current tight GPU supply, GPU leasing is a lower priority for us. As for the subscription model, it is not the most likely business model for AI in China—AI capabilities are generally offered for free, so the subscription charging model popular abroad is not mainstream in China.

Q: Regarding e-commerce: Can you explain the recent organizational restructuring of the e-commerce department, especially the latest developments in small stores? What should we expect in terms of future strategy and key metrics? How will we create synergy with our own ecosystem and differentiate ourselves from other live e-commerce platforms? In addition, please talk about our preparations for the "6·18" promotion.**

A: In terms of e-commerce, this organizational adjustment is actually quite small. Our e-commerce team was initially incubated within the open platform department, and as the scale expanded, we are now just formally splitting the two, making the e-commerce team an independent department, but still managed by the same person. So this is not a significant change and should not be over-interpreted.

Q: I would like to understand the latest progress of "Yuanbao" after its integration into the WeChat ecosystem. Can you introduce the changes in user behavior we have observed after the integration? Additionally, how do you view the potential synergies between various initiatives within the ecosystem (including small shops, short content, video accounts, etc.)? What are the next milestones? Will we integrate both private and public domain capabilities into "Yuanbao" itself? Thank you.

A: Frankly speaking, we are still in a very early stage. From what we have observed, users have already started using it: they use it to ask questions, have conversations, and also send content to "Yuanbao" for summarization and analysis—these are the initial use cases we see at the moment. As user interactions with "Yuanbao" increase, the usage frequency is expected to continue to rise. In the future, we will definitely connect more capabilities from the WeChat ecosystem with the "Yuanbao" chat partner; we will conduct various experiments, and perhaps in one or two quarters, we will be able to provide more information at the system level. It is still too early to make a systematic summary.

Q: As the usage rates on both the user and enterprise sides increase, how do we balance the speed of investment with revenue potential, especially in the early scaling phase?

A: At this stage, our idea is to move forward at full speed and stimulate demand as much as possible. Frankly, we would only consider slowing down if there is excessive demand that our existing GPUs cannot handle; that would be a "happy trouble." We are not at that point yet.

Q: This quarter, the domestic gaming business has performed very strongly. Can management help us analyze the growth logic of this sector from a longer-term perspective: is this impressive performance a release of "compensatory demand" for the relatively weak period from 2021 to 2023, or is it due to new structural factors emerging recently, such as consumer attention and spending concentrating on Tencent's advantages in top global games, as well as faster content iteration driven by AI?

A: Regarding the question about domestic gaming, it should first be noted that in the first quarter of this year, we indeed benefited from the "year-on-year bonus" brought by the low base in the first quarter of 2024, which is somewhat favorable for us, but this situation will not occur in every future quarter. Nevertheless, we still believe that both domestically and internationally, gaming revenue has a long-term growth runway. There are many reasons for this, and here are three points to explain:

• Operational structure and team adjustments have shown results

  • At this time last year, we had in-depth discussions about the vision, operation methods, and team structure of several core domestic games. Now we can see that these changes are meeting expectations and are expected to continue to have a positive impact

• AI empowers large-scale competitive multiplayer games. We believe that AI is particularly beneficial for large-scale competitive multiplayer games, which account for a significant portion of domestic gaming revenue. We have begun deploying AI in various aspects,

  • For example: using AI to coach novice players; accompanying existing players; preventing cheating and hacks.
  • These features are especially critical in competitive games, effectively enhancing experience and retention.

• Aligning with the trend of Chinese players shifting towards first-person action games

  • In the past, the proportion of first-person action (FPS/TPS) games in the Chinese market was far below the global average. We believe this gap will be bridged. The multiple games we highlighted this quarter—"Peacekeeper Elite," "Crossfire Mobile," "Call of Duty Mobile," "Valorant," and "Delta Force"—are all leading titles in this category, and Tencent is the leader in this vertical. Although the growth rates of each game vary, the overall category is growing rapidly, with "Valorant" and "Delta Force" being particularly outstanding. They not only supported this quarter's revenue but also laid the groundwork for future growth.

Q: Additionally, there has been much discussion globally about regulatory intervention and the redistribution of economic benefits between app stores and applications. Tencent has been paying attention to this topic for many years. In terms of the video game sector, what stage is China in this process of value redistribution? If value distribution shifts further, what would it mean for Tencent? Thank you.

A: Regarding the app store revenue-sharing issue, we are currently in a period of adjustment. In the past, digital content creators—especially game companies—were "punished" by relatively harsh revenue-sharing terms, while merchants selling physical goods hardly had to pay for the digital distribution system.

As a company that operates both an app store and produces digital content, we believe this distribution is overly skewed towards the store side. China has been reforming this aspect for several years, and now digital content providers (including Tencent itself) can obtain a fairer share of user payments. In Western markets, this transformation has previously progressed slowly or even stagnated, but recent lawsuits and regulatory actions are driving change. While it is difficult to predict a short-term timeline, we believe that in the long run, the world will move towards a more balanced cooperative relationship, allowing digital content creators to retain more of the value they create, rather than continuing to subsidize the entire app store ecosystem for free. This is a direction already evident in China and will gradually manifest globally.

Q: In the past two to three months, as AI has been more deeply integrated into Tencent's various business applications, what significant changes in user behavior has management observed? For example, how have users performed in using "Yuanbao," have these AI innovations been noticeably perceived by business partners, and how does Tencent anticipate that the current observations can help the company further enhance user value and unlock future monetization potential?

A: In fact, I think we have partially answered this question earlier. It may still be too early to conduct a systematic analysis of changes in user behavior. At this stage, we are working hard to create new features and user experiences that leverage AI, observing which features will be genuinely accepted by users and which may not

As I mentioned earlier, users like to ask questions and engage in multiple rounds of interaction with AI. When we launched new features like "let AI analyze photos," some new users would try them out. We have already rolled out many features, and we are starting to see that some features are very popular with users, while others are not used as much. This is still an exploratory process.

Overall, we can see that user interaction with our AI assistant is continuously increasing—users are gradually developing a sense of familiarity and are exploring what value AI can bring to them. Therefore, from the perspective of overall usage, it is indeed on the rise.

Q: Regarding games, I know that some top-performing games are currently benefiting from the integration of AI. In the next few quarters, will we also integrate AI into smaller games to enhance the monetization capabilities of these mid-sized games or reactivate user engagement? Thank you.

A: I briefly mentioned how we are deploying "game AI" in large competitive games when answering William's question. Currently, this is still in a very early deployment stage, and we believe the biggest opportunity is to continue deepening along this path. At the same time, there is another approach: to introduce more generative AI into content-driven games.

This way, it may first allow game studios to produce content more quickly; then, over time, a certain degree of user-generated content (UGC) will emerge, bringing different quality experiences and ultimately achieving dynamic generation—when players choose to leave the preset map, the game itself can instantly generate a new environment. However, I believe all of this is an area to explore in the coming years. For now, the most substantial and tangible opportunity remains to apply what we call "game AI" in large multiplayer competitive games.

Q: Regarding high-end GPUs. Recently, the U.S. has introduced licensing requirements for high-end GPUs, and we understand that the primary task is to prioritize these GPUs for internal company use. Can management discuss their views on this matter? How will this affect our capital expenditures, AI R&D, and product releases?

A: The GPU situation is changing very rapidly. Since the last quarterly earnings call, we have experienced the emergence of the H20 chip, followed by new regulatory guidelines issued overnight by the U.S. Bureau of Industry and Security (BIS)—the situation is very dynamic. We must manage this situation while ensuring full compliance and finding the best solutions to ensure that the company's AI strategy can still be implemented.

There are two good pieces of news:

• Sufficient spot inventory—we previously procured a batch of chips, and the inventory is relatively ample, which is crucial for executing our AI strategy
• Clear usage priority — These chips will first be deployed in scenarios that can immediately generate returns, such as advertising placement and content recommendation. In advertising and content recommendation products, we need a large number of GPUs to directly produce effects and revenue.

In terms of large model training, this falls under the second priority, as training typically requires higher-end chips. Fortunately, there has been a shift in industry thinking over the past few months: people are beginning to move away from the so-called "Scaling Law" of American tech companies — the mindset that training clusters must continuously expand. Practice has shown that even smaller clusters can achieve good training results; models still have great potential in the post-training (fine-tuning) phase, which does not necessarily require ultra-large-scale clusters. In other words, with the existing inventory of high-end chips, we are fully capable of being "sufficient" in training the next several generations of models. The greater demand for GPUs mainly comes from the inference side, especially as user-side demand grows, or when we shift to "chain reasoning" inference models (which require more tokens to answer complex questions) and even future Agentic AI (which uses more tokens), the volume of inference will rise sharply. We have various means to address the pressure on the inference side:

• Software optimization: Continuously improve inference efficiency. If efficiency doubles, it is equivalent to doubling inference capability, which is akin to "virtually" increasing the number of GPUs.
• Model customization and distillation: Tailor model sizes for different scenarios or perform model distillation; many applications do not require ultra-large models. This meets performance needs while saving on inference consumption.
• Diversified chips: Under regulatory compliance, flexibly use local Chinese or other compliant chips that can be imported, as well as ASICs, FPGAs, etc., which can also play a role in smaller model inference.

In summary, we can meet the continuously expanding inference demand through multiple paths, putting more effort into the software layer rather than simply "throwing GPUs" at the problem.

Q: The company mentioned that after adopting a more conservative strategy in the consumer loan business last year, it achieved year-on-year revenue growth this year. Considering the macro environment still has uncertainties, could management share frontline observations — whether on the client side or the advertiser side — regarding the impact?

A: This year we have noticed that the overall credit quality of loans facilitated through the platform is steadily improving. This is partly due to our stricter borrower screening and partly attributed to macro-level changes — over the past few years, the overall savings rate of Chinese residents has increased. Clearly, we prefer to lend funds to consumers with ample savings rather than those with insufficient savings.

Therefore, while cautiously controlling risks, we have allowed the balance of facilitated loans to grow at a steady pace. Looking ahead, if we compare the balance of loans facilitated on our platform with the largest peers in the industry, it is still only a small portion of their scale. This means that there is still ample growth space for high-margin loan facilitation revenue for many years to come — provided that we continue to maintain prudent growth

Q: Regarding increasing AI reinvestment: In the opening speech, management mentioned that the gap between future revenue growth and operating profit growth will narrow. Could you provide more details on the magnitude and duration of this gap? Besides capital expenditures, what other expense items will widen (or narrow) the gap between the two?

A: Currently, the entire AI field is in a "no man's land" phase, which is true not only for Tencent but for the entire industry, making it difficult to predict with 100% accuracy. However, looking at history, Tencent has often experienced a "time mismatch" between the investment period and the harvest period when nurturing new products (invest first, monetize later), typically ranging from 1 to 2 years. In the future, this will still depend on: the actions of peer companies in the Chinese market; the evolution of user habits and advertisers' spending habits. Therefore, using "approximately one to two years" as a reference range is relatively reasonable.

We only emphasize "narrowing," which means that the phenomenon of "revenue growth exceeding operating profit growth" observed this quarter will not continue entirely; however, we do not expect operating leverage to turn negative.

Expense items other than capital expenditures:

• Depreciation and amortization (from GPU/server CapEx) — the largest proportion.
• Marketing expenses related to Yuanbao — although increased, the scale is much smaller than depreciation.
• AI engineer salaries — AI talent is scarce, and individual salaries are relatively high; however, this is an optimization of the labor structure, not a significant increase in overall headcount.

Overall, CapEx and its depreciation remain the decisive factors affecting the gap.

Q: Regarding the domestic first-person shooter (FPS) category: This quarter, "Delta Force" performed excellently, and "Crossfire Mobile" was also mentioned. Are we witnessing a structural rise of FPS games? Among the multiple FPS games you mentioned, is it possible to see several "evergreen" products? Thank you.

A: Chinese player behavior is aligning with global trends: in markets like Europe and the U.S., FPS is a "super category," consistently accounting for 40-50% of player time and revenue; whereas in China, it was only 10-20% in the past. Now we observe:

• Younger players (ages 20-30) show a particularly strong preference for FPS;
• For example, "Valorant" has precisely targeted this demographic and performed well.

New and old FPS products driving growth

• New releases: such as "Valorant" and "Delta Force" are experiencing rapid growth;
• Existing evergreen titles: such as "Peacekeeper Elite," "Call of Duty Mobile," and "Crossfire Mobile," continue to grow.

Gameplay iteration broadening the audience

• A few years ago, the "battle royale" mode popularized Fortnite, CoD Warzone, and PUBG in the West, and then spread to China, expanding the overall FPS market;

• Recently, "Extraction Shooter" has brought in new growth:

  • The increase in DAU and revenue for "Peace Elite" this quarter mainly comes from this mode;
  • The users with the highest retention rate for "Delta Force" are also concentrated in the evacuation shooting gameplay.

Every few years, new modes emerge in FPS games, and these new modes expand more than they cannibalize, thereby continuously enlarging the overall user base. In summary, we are optimistic about the prospects of multiple FPS games forming a "blooming everywhere and evergreen" situation in the Chinese market.

Q: We hear management mention that AI technology is one of the driving factors for the growth of the advertising business every quarter. So, how much growth potential can further application of AI technology bring to your advertising business? Regarding video account advertising: what are the current ad load and the latest level of eCPM? Regarding WeChat search: what is the current market share of WeChat search in search queries? How much revenue does WeChat search currently contribute to the advertising business? After you added AI search functionality to WeChat, have you seen a shift in user search behavior from traditional search to AI search? Thank you.

A:

1. How much incremental revenue can AI bring to advertising?

This is actually a common question across the global advertising industry. If I could answer this precisely, it would greatly benefit the understanding of the advertising prospects for Meta, Google, and indeed all other companies. To simplify the framework, the main enhancement AI brings to advertising revenue can largely (though not entirely) be quantified as an increase in click-through rates (CTR).

Historically: The click-through rate for banner ads is about 0.1%; the click-through rate for feed ads is about 1.0%.

With AI: We see that the click-through rates for certain ad inventories can be boosted to around 3.0%.

The next question is: what is the theoretical upper limit of the click-through rate? Currently, no one can provide an answer; this has almost evolved into a philosophical question. If you have comprehensive insights into consumer information, can infer what they truly need, and deliver ads to them accurately, then what is the theoretical upper limit of CTR, X% or Y%? It is difficult to define. Of course, AI enhances advertising revenue in ways beyond this. AI can be used to deliver more appealing content to users, extending their stay in the information feed, thus generating more ad exposure. But I believe that "advertising click-through rate" remains the most core metric.

2. Video account advertising load and eCPM

Load rate: It has remained relatively stable over the past six months, maintaining around 3-4%.

eCPM: It is still at a very ideal level, benefiting from both high click-through rates and fierce competition among advertisers for limited inventory

3. Performance of WeChat Search & AI Search Migration

• Search query share: We have not disclosed specific numbers yet, but overall it shows a good upward trend.

• User response to AI search:

  • It is still in the early stages.
  • The concept of AI search itself is somewhat vague: the boundary between users entering prompts in a purely AI dialogue box and entering queries in traditional search engines, which then return results via traditional algorithms or large language models, is continuously merging.
  • As a latecomer in the search field, this merging is beneficial for us as it gives us the opportunity to capture user share and ultimately increase revenue share.

Q: Regarding the performance of the advertising business: The year-on-year growth rate of advertising revenue in the fourth quarter is 17%, while it accelerated to 20% in the first quarter. To what extent is this acceleration driven by the recovery of the macro environment, technological improvements, and the release of advertising inventory? Based on previous discussions, I suspect the impact of inventory release should be minimal, and even if quantified precisely, it would be quite difficult. Could the management please provide a rough ranking of the three factors in terms of their impact on revenue acceleration?

A:

• Regarding the driving factors of advertising growth

• I believe there is no need to overinterpret the fluctuations in advertising revenue between quarters—it has always remained within a “band.” Frankly speaking, the performance in the first quarter of this year is at the upper edge of that “band,” and we do not necessarily wish for “excessive” growth. For us, it is more important to maintain a sufficiently long growth runway, allowing this “band” to sustain for years rather than just a few quarters.

• If we find ourselves too close to the upper limit of the “band,” we may slow down the increase of the advertising load rate of existing products or the pace of deploying ads in new AI products, prioritizing the optimization of user duration and experience instead. This can avoid a short-term breakthrough of the limit, which would shorten the runway.

• As for the specific reasons for the upward growth this quarter: it is definitely not due to inventory release—we did not stockpile or release inventory in advance; whether macro factors dominate still needs to be verified as industry peers gradually release their data.

• The improvement of AI-driven advertising technology (ad tech) is not something we meticulously plan on a quarterly basis, so it is inappropriate to extrapolate single-quarter fluctuations. Ultimately, this performance still belongs to the normal fluctuations within the “band,” and there is no need for overinterpretation or to extrapolate future trends based on it.

• Additionally, it should be noted that organic traffic growth is also a key driver, keeping the overall growth “band” in the range of “dozens of percentage points” rather than single-digit positive or negative growth.

• For the long-term strategy of the advertising business, what we value most is extending the growth runway of the advertising business. Specific practices include:

• Continuously enhance ad tech capabilities (especially AI-related capabilities) to improve click-through rates (CTR) and delivery efficiency;

• Increase quality traffic - access to advertising placements such as video accounts, mini-games/mini-programs, public accounts, and WeChat search;

• Utilize AI algorithms to push more suitable content to users, extend viewing time, thereby naturally increasing revenue without raising ad load;

• Gradually build a more complete transaction ecosystem, allowing merchants to complete conversions more smoothly after placing ads, enhancing the value of each click.

• With these systematic measures, we believe that the advertising business has a very long growth runway, and our task is to continuously "extend" this runway rather than pushing growth to the extreme in a single quarter.

Q: Regarding AI product forms: Your company has been actively driving traffic for "Yuanbao" for some time. Can you share feedback from active users, retention, and other data? More importantly, from a product perspective, how does management view whether chatbots will become the ultimate form of generative AI? Thank you.

A: In the past quarter, we significantly expanded the user base of "Yuanbao" and invested a lot of effort in user retention, achieving good results.

The current mainstream form is still the "chat" interface. However, as functionalities continue to enrich, when executing more complex tasks, "Yuanbao" may present itself more like a "project" rather than just a chat window, requiring long-term retention of task cards.

Next phase focus:

• Continue to add functionalities based on the existing scale, further enhance retention and activation, and attract new users;
• Deeply leverage Tencent's unique ecological advantages - social, content, mini-programs, etc. - to create differentiated experiences from other AI products on the market;
• Continuously upgrade self-developed large models, strengthening the "dual model" or multimodal strategy.

Regarding "whether chatbots are the ultimate form of Generative AI": we believe that forms will change with functionalities, ultimately flexibly switching between various interaction modes such as chat, project management, and long-term card retention based on user task needs.

Q: Regarding our WeChat e-commerce strategy, what is the company's core KPI? Are there plans to launch a shelf-like display form in the future, and what is the timeline? The current main forms are still live streaming, search, etc.

A: First of all, regarding the e-commerce strategy, we do not operate by setting very rigid KPIs. This is a long-term business with a very long growth runway. I repeatedly mention the term "runway" because for us, enabling the business to sustain growth for many years is more important than focusing on specific KPIs for a short-term sprint. To lay this runway, we are advancing several key tasks:

1. Improve the basic shopping experience for consumers

• Provide higher quality products.
• Establish a convenient return policy.
• Create a high-quality customer service process.
• Offer consumers competitive prices.

1. Introduce more high-quality supply

• Attract more quality brands and actively managed merchants to join our ecosystem, continuously enhancing product supply.

1. Bring more traffic to merchants

• Provide consumers with more opportunities to interact with products and merchants.

• The current primary method is live streaming sales, which we have already validated and is a "low-regret" approach, proven effective by other platforms.

• At the same time, we hope to create unique connection methods within the WeChat ecosystem:

  • Connect users with products through mini-programs.
  • Many users have completed purchases in mini-programs; if we can shift this portion of transactions to small shops, it will amplify transaction volume due to the more complete e-commerce infrastructure of small shops, benefiting both users and merchants.
  • There are also social and communication components, such as the "gift party" feature in Moments, search, group chats, and public accounts, all of which can connect with products.

• As these connections gradually take shape, they will uniquely bring more transactions to merchants in the WeChat ecosystem.

Overall, these initiatives are our long-term efforts to promote the growth of small shops and the WeChat e-commerce ecosystem. It is encouraging that even in the "first inning," our small shop transaction GMV continues to show strong growth.

Q: Regarding payment business: Management mentioned a slight decline in payment transaction volume in the first quarter. Is this mainly due to macro factors or payment strategies? Additionally, you mentioned that there has been improvement in April; could you explain the reasons? Thank you.

A: Our market share may fluctuate due to factors such as the intensity of subsidies and the degree of openness of credit card funding channels, but we do not make drastic adjustments on a monthly basis, nor have we significantly increased competitive intensity from the first quarter to April. Overall, consumer spending has shown volatility in recent quarters, and the latest data points indicate an upward trend; this may reflect that consumer confidence and spending activities are stabilizing, and we hope this stabilization will eventually translate into sustained growth.

To add some details: In the first quarter, we observed a continuous increase in transaction volume, but the average transaction value decreased. Our judgment is that as demand begins to warm up, supply-side competition may be entering its final stage. The trend in April provides some validation for this hypothesis—before the implementation of additional tariffs. Without tariff factors, consumption patterns have begun to gradually recover; what we need to observe now is the specific enforcement of tariffs. On the macro front, the government has indicated that it will continue to introduce stimulus measures, and China still has considerable room in this regard. Therefore, what needs to be focused on in the future is the impact of tariffs on the economy and the counteracting effects of stimulus policies; these two factors will jointly determine the subsequent trends The market has risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investment based on this is at one's own risk