Tencent: Has enough chips for AI training and model upgrades, with multiple options for AI inference chips

On August 13, Tencent Holdings held a conference call for its second-quarter earnings.

The financial report released on the same day showed that Tencent's Q2 revenue increased by 15% year-on-year to 184.5 billion yuan, exceeding expectations, with net profit rising by 17% year-on-year. The gaming business showed a strong recovery, with domestic market game revenue increasing by 17% year-on-year.

The company's advertising business achieved a 20% year-on-year growth driven by AI technology, and management expressed optimism about the long-term growth prospects of advertising revenue. The company is enhancing advertising accuracy and conversion efficiency through AI technology while deploying AI capabilities across multiple business lines. It has sufficient chips for AI training and model upgrades.

Tencent's management stated that the growth in advertising revenue is mainly attributed to the increase in revenue per exposure brought by AI technology, as well as the growth in traffic from video accounts and search. Currently, the loading rate of short video ads is only in the low single digits, indicating significant room for improvement compared to the industry average of over 10%.

In terms of AI business layout, Tencent continues to promote native AI applications such as Yuanbao and integrates AI capabilities into products like WeChat Search, Tencent Meeting, and Tencent Docs. Management noted that although the cost of AI services is high, overall inference costs can be controlled by adopting small models and improving efficiency.

Enterprise service revenue saw accelerated growth this quarter, mainly benefiting from increased demand from enterprise clients for GPU leasing and API Token usage. Management stated that they will continue to expand cloud computing services but will not overly rely on fluctuations in GPU supply.

Regarding the performance of products like "Delta Force," Tencent's management categorized games into two types: buyout AAA games and long-term operational service games. The company believes that the Chinese market is still dominated by the latter, with "Delta Force" seeing continuous growth in user numbers and revenue since its launch nine months ago, aligning with expectations for evergreen games.

The following is a transcript of the conference call:

Tencent Management:

We achieved the highest double-digit revenue and non-International Financial Reporting Standards (IFRS) operating profit growth. While increasing our investment in artificial intelligence (AI), we also fully benefited from the application of AI. Our games performed well in terms of user numbers and revenue, with evergreen games like "Honor of Kings" and "Peacekeeper Elite" gradually evolving into platform products, while increasing the use of AI. New games like "Delta Force" have also made breakthroughs. Our marketing services revenue has maintained rapid growth, benefiting from the upgrade of our advertising base model, which has improved advertising performance across Tencent's various traffic platforms. We are working hard to explore more AI application scenarios within WeChat, promote the use of the native AI application "Yuanbao," and continuously upgrade the "Hunyuan" base model to bring more value to consumers and enterprises through AI.

From a financial perspective, total revenue for the second quarter was 185 billion yuan, an increase of 15% year-on-year. Gross profit was 105 billion yuan, up 22% year-on-year. Non-IFRS operating profit was 69 billion yuan, an increase of 80% year-on-year; non-IFRS net profit attributable to shareholders was 63 billion yuan, up 10% year-on-year. Excluding the profit contributions from joint ventures for this quarter and the same period last year (last year saw higher profits due to improved profitability of related joint ventures and one-time retrospective adjustments), our core net profit still grew by 20% year-on-year Turning to our core services—communication and social networking. WeChat and WeChat's combined monthly active accounts have achieved year-on-year and quarter-on-quarter growth, reaching 1.4 billion. In terms of digital content, Tencent Music Entertainment Group has consolidated its leadership position in online music streaming. In gaming, we have made progress in gameplay-focused products such as "Peacekeeper Elite" and "Delta Force," while content-driven games like "Jian Wang 3" and "Ming Chao" have also performed outstandingly.

In cloud services, Gartner has named Tencent Cloud one of the world's top communication platform as a service (CPaaS) providers for three consecutive years. Next, I will hand over the business review to Martin.

Tencent Management:

In the second quarter of 2025, our total revenue grew by 15% year-on-year. Value-added services (VAS) accounted for 50% of total revenue, with the social networking segment contributing 18%, domestic gaming contributing 22%, and international gaming contributing 10%. Marketing services accounted for 19%; financial technology and enterprise services accounted for 30%. In terms of gross profit, the gross profit for the second quarter increased by 22% year-on-year, reaching 105 billion yuan. Notably, the gross profit of our three major business segments all grew by over 20% year-on-year. The gross profit from value-added services increased by 23% year-on-year to 55 billion yuan, contributing 53% of total gross profit; marketing services grew by 24% year-on-year to 21 billion yuan, accounting for 20%; financial technology and enterprise services grew by 21% year-on-year to 29 billion yuan, accounting for 28%.

Specifically, in each major business segment, value-added service revenue was 91 billion yuan, a year-on-year increase of 16%. Social networking revenue grew by 6% year-on-year, mainly benefiting from a 17% year-on-year increase in "in-app game item sales," video accounts, live streaming services, and music subscription revenue. The number of annual paying members exceeded 50 million, benefiting from new privileges such as artist card collections, merchandise, and early ticket purchases for performances. Long video subscription revenue decreased by 2% year-on-year, and the number of video subscription users decreased by 3% year-on-year to 214 million, mainly due to a reduction in the scheduling of top content. Even in a quarter with limited content supply, our self-produced drama "Beauty in Prison" became the most popular long video series across all platforms in May.

Domestic gaming revenue grew by 17% year-on-year, mainly benefiting from the strong performance of "Delta Force" and evergreen games such as "Honor of Kings," "Valorant," and "Peacekeeper Elite." International gaming revenue grew by 35% year-on-year, with a 33% increase at constant exchange rates, mainly driven by the star product "PUBG Mobile" and the new title "Tuning Awakening."

Further into the communication and social networking business, WeChat Mini Programs are increasingly becoming a powerful platform for users to connect with merchants and content providers. In the second quarter of 2024, the total transaction volume (GMV) of Mini Programs grew by over 10% year-on-year, benefiting from the support of use cases, offline dining orders, and travel scenarios. The total revenue of mini-games grew by 20% year-on-year. We upgraded the mini-game engine, enhancing compatibility, image rendering effects, and optimizing loading speed, making it easier for developers to import more complex app games such as simulation management and large-scale MMORPGs onto the mini-game platform. For small stores, we support brands and merchants in bulk importing Mini Program product libraries and unifying membership systems across both platforms In addition, we continue to strengthen the social e-commerce experience unique to WeChat, distinguishing it from traditional e-commerce platforms. First, we expanded the "gift" feature of Mini Shops to Moments, Video Accounts, and Official Accounts; secondly, we launched the "shop with friends" feature, encouraging users to share their favorite products and stores through chat and Moments, and participate in group buying activities together. In terms of AI, we added an AI citation feature to the content, allowing users to obtain AI intelligent interpretations of relevant content and background information while reading Official Account articles or commenting on Video Accounts.

We upgraded the customer service of Mini Shops, incorporating large language model capabilities, providing scenarios for mini-games such as venue, dine-in ordering, and transportation, resulting in a 20% year-on-year increase in total collections for mini-games.

We upgraded the technical architecture of mini-games, expanding the compatibility of the game engine, enhancing graphics rendering capabilities, and reducing low types, making it easier for developers to port complex app-level games to the mini-game platform. The newly added game types include simulation and massively multiplayer online role-playing games (MMORPGs).

Regarding Mini Shops, we allow brands and merchants to migrate their product libraries (SKUs) from mini-programs to Mini Shops, achieving integration of the membership systems of the two platforms. Additionally, we continue to enhance the social e-commerce experience, which is a feature that distinguishes WeChat from traditional e-commerce platforms. First, we expanded the gifting feature of Mini Shops to mini-programs, Video Accounts, and Official Accounts. Secondly, we launched the "shop with friends" feature, encouraging users to share their favorite products and stores through chat and Moments, and participate in group buying discounts together.

In terms of AI, we added an AI-driven citation feature to the content, allowing users to activate contextual AI recommendations or comments on relevant content while reading Official Account articles or commenting on Video Accounts. At the same time, we upgraded the Mini Shops customer service system, integrating large language model capabilities.

We are increasingly applying AI tools in major game content production to enhance production speed and scale. AI enables us to provide more lifelike virtual teammates in competitive PVP games and create more realistic non-player characters (NPCs) in narrative PVE games. We also use AI to improve the efficiency of marketing activities, targeting ads more precisely to users who are most likely to engage and retain.

In the international gaming sector, in PUBG Mobile, an increasing number of daily active users are starting to play in the user-generated content (UGC) scenarios of Wonder Sandbox. The popularity of the subway extraction shooting mode and the Wonder UGC environment demonstrates that we are upgrading PUBG Mobile from a "service-type game" to a "platform-type game." In April 2024, driven by ancient Egyptian-themed costumes, PUBG Mobile's monthly total revenue reached an all-time high. In Clash Royale, Supercell accelerated the pace of content updates and community activities, optimized the reward mechanism, and enhanced live streaming influence and DAU, with Clash Royale's monthly revenue reaching a seven-year high in June 2024 Do you know about "Awakening"? This is an open-world survival game developed and published by our Norwegian subsidiary Funcom. After its launch in early June, it became the top-grossing new buyout game on the global Steam platform within a week. The survival game and June novel player communities have responded positively to this title, and Funcom has prepared a rich content update plan to expand the game world.

In terms of marketing services, revenue this quarter increased by 20% year-on-year, reaching 36 billion yuan, thanks to AI-driven advertising technology upgrades and new advertising revenue from the WeChat transactional ecosystem. Specifically, the expansion of AI capabilities involves advertising creativity, placement, recommendations, and performance analysis, effectively improving click-through rates, conversions, and advertisers' ROI. Specifically, we achieved real-time insights into user interests and optimized advertising performance by deploying large models to upgrade the advertising platform architecture, analyzing advertising clicks and transactions across applications, and examining user interactions among text, images, and video content.

Within the video account, due to increased traffic and transaction activity, marketing service revenue grew by approximately 50% year-on-year; marketing service revenue within mini-programs also increased by about 50%. The interaction between mini-games and mini-theaters has created a flywheel effect, driving more developers to adopt our closed-loop marketing solutions to promote their services. WeChat search revenue increased by approximately 60% year-on-year, benefiting from enhanced advertiser appeal in mini-program search results and improved understanding of product and consumption intentions brought by large language models.

In the fintech and enterprise services sector, revenue was 56 billion yuan, a year-on-year increase of 10%. Fintech service revenue saw a high single-digit growth year-on-year, mainly driven by commercial payment and consumer loan businesses. In terms of commercial payments, benefiting from a narrowing decline in single transaction amounts and an increase in transaction volume, total payment volume turned positive year-on-year in Q2 2024. Online payments maintained healthy growth, while the weak trend in offline payments improved due to increased consumption in retail and dining. Enterprise service revenue grew by over 10% year-on-year, with accelerated growth in cloud service revenue, mainly driven by increased income from AI-related GPUs and API Tokens. Transactions related to Mini Shop grew rapidly, and the gross margin of enterprise services increased year-on-year due to efficiency improvements and structural optimization. International cloud revenue saw significant growth, and we are increasingly experienced in assisting large international clients in cloud migration and improving IT efficiency. For example, we helped Gojek and Tokopedia Group complete the migration of their SaaS systems, one of the most complex cloud migration projects in Southeast Asia. We also enhanced data quality and diversity through data augmentation and synthesis, and we are continuously conducting more efficient AI model pre-training and post-training.

The Hunyuan 3D model has become the top-ranked 3D generation model on HuggingFace, excelling in geometric accuracy, texture restoration, and instruction-based 3D alignment capabilities. An increasing number of game developers, 3D printing companies, and designers are adopting Hunyuan 3D models to generate digital assets. Now, I will hand it over to John to introduce the financial data.

Tencent Management:

In Q2 2024, the company's total revenue was 184.5 billion yuan, a year-on-year increase of 15%. Gross profit was 105 billion yuan, a year-on-year increase of 22%. Other losses amounted to 3.6 billion yuan, compared to a profit of 1.5 billion yuan in the same period last year, mainly due to a reduction in subsidies and tax refunds this quarter, as well as provisions for losses on certain receivables Operating profit was 60.1 billion yuan, a year-on-year increase of 18%; interest income was 4.1 billion yuan, a year-on-year increase of 7%; financial expenses were 3.9 billion yuan, a year-on-year increase of 27%, mainly due to increased interest expenses and expanded foreign exchange losses. Dividends from associates/joint ventures amounted to 4.5 billion yuan, compared to 7.7 billion yuan in the same period last year. Under non-International Financial Reporting Standards, this figure was 6.3 billion yuan, down from 9.9 billion yuan in the same period last year, mainly affected by reduced profits from a large associate. Income tax expenses were 11.4 billion yuan, a year-on-year increase of 12%, reflecting the growth in operating profit. Diluted earnings per share were 6.793 yuan, a year-on-year increase of 13%, due to a reduction in total share capital from share buybacks, with net profit growth under non-International standards leading. The weighted average number of shares used in the diluted EPS calculation decreased by 2% year-on-year.

Under non-International standards, operating profit was 69.2 billion yuan, a year-on-year increase of 18%; net profit attributable to shareholders was 63.1 billion yuan, a year-on-year increase of 10%. Excluding contributions from associates and joint ventures, net profit for this quarter would have increased by 20% year-on-year. In terms of revenue, 56.8% came from core business operations.

By business segment, the overall gross profit margin was 57%, an increase of 4 percentage points year-on-year. The gross profit margin for the gaming and music segments reached 60%, a year-on-year increase of 5 percentage points, driven by an increased share of high-margin domestic game revenue and growth in music subscription service ARPU and users. The gross profit margin for marketing services was 58%, an increase of 2 percentage points year-on-year, benefiting from the growth of high-margin video account and search revenue. The gross profit margin for fintech and enterprise services was 52%, an increase of 5 percentage points year-on-year, mainly due to a shift in revenue structure towards wealth management and consumer credit, as well as cost reduction and efficiency improvement in payment and cloud services.

During the second quarter, sales and marketing expenses were 9.4 billion yuan, a year-on-year increase of 3%, reflecting our investment in promoting the growth of AI-native applications, with expenses accounting for 5% of revenue, lower than 6% in the same period last year. R&D expenses were 24.3 billion yuan, a year-on-year increase of 17%, focused on AI-native applications. Excluding R&D, management and administrative expenses were 11.6 billion yuan, a year-on-year increase of 14%, reflecting some performance bonuses and the expansion of overseas subsidiaries. By the end of the quarter, the number of employees was approximately 111,000, a year-on-year increase of 5% and a quarter-on-quarter increase of 2%, mainly due to steady recruitment.

Under non-International standards, the operating profit margin for the second quarter was 38%, an increase of 1 percentage point year-on-year. Finally, a brief review of cash flow and the balance sheet:

Capital expenditures for operations were 17.9 billion yuan, a year-on-year increase of 149%, mainly due to increased investment in GPUs and servers to enhance AI capabilities. Non-operating capital expenditures were 1.2 billion yuan, a year-on-year decrease of 20%. Total capital expenditures were 19.1 billion yuan, a year-on-year increase of 119%. Free cash flow was 43 billion yuan, a year-on-year increase of 7%, driven by increased game collections. Quarter-on-quarter, free cash flow decreased by 9%, due to seasonal declines in game revenue after the Spring Festival. The net cash holding was 74.6 billion yuan, a quarter-on-quarter decrease of 17% or 15.6 billion yuan, mainly due to the final dividend distribution of 37.35 billion yuan at the end of this fiscal year.

From the questioner at Jefferies:

Good evening, thank you to the management for accepting my question, and congratulations on the company's strong performance.

My question is about the advertising business. We see that Tencent and some overseas peers are leveraging artificial intelligence (AI) to drive growth in advertising revenue. Can management comment on the growth potential of marketing services (advertising business) in the coming quarters and even in the coming years? Specifically, from the perspective of advertising growth drivers, how should we view the traffic growth, conversion efficiency, and ad load rate from WeChat Channels? Thank you.

Tencent Management's Response:

Thank you, Thomas. Regarding our advertising business and its potential, we have always believed that we have a long-term and still expanding growth space that can support our advertising revenue to continue growing at a fairly healthy rate.

The breadth of this growth space is reflected in several key variables that determine our marketing services revenue, all of which have significant room for improvement, including:

• Click-through rate: Artificial intelligence (AI) has brought more precise targeting, thereby improving the click-through rate.
• Traffic: We have seen growth in traffic from WeChat Channels and search traffic, and in the future, our AI-native in-app experiences will also bring new traffic.
• Revenue per click: On one hand, the use of generative AI to create advertising materials has led to more advertising demand; on the other hand, the increase in closed-loop transactions in e-commerce has also brought more advertising demand, both of which have jointly pushed up revenue per click.
• Ad load: Finally, the ad load rate. Currently, our short video ad load rate is only in the low single digits (e.g., 3%-6%), while our peers are at the teens (e.g., 13%-16%).

Therefore, based on these reasons, we believe our advertising revenue has very broad growth prospects.

As for the priority of these driving factors, in the second quarter, the growth of our advertising revenue mainly came from the increase in "revenue per impression." The increase in "revenue per impression" is primarily attributed to two aspects: first, we deployed AI technology, which brought higher click-through rates; second, there were more closed-loop transactions in our mini-program store (Tencent HuiJu) and mini-games, leading to higher revenue per click.

The second driving factor is the increase in "impressions," which is mainly due to the growth in traffic from WeChat Channels and search. During this period, the ad load rate remained stable and was not a major driving factor.

That concludes the analysis of the growth drivers for the second quarter. Thank you.

From the JP Morgan questioner:

Thank you for allowing me to ask a question. My first question is, you have been continuously integrating more AI features into a large number of mobile internet applications this year. Besides launching the native on-chain AI application Yuanbao, what behavioral changes have you observed in users when using these mobile internet applications with embedded AI features? And what are the ecological impacts of these user behaviors? For example, if more and more people use AI search, they might bypass websites or public accounts and consume content directly in AI summaries. Would this have an impact on the mobile internet ecosystem?

My second question is, as you continue to provide users with more AI features for free, the delivery of these AI features is actually much more expensive than traditional mobile internet services, which may affect Tencent's cost structure Is the management considering monetizing AI features aimed at the C-end directly in the next year or two? If so, which AI features/applications have the most monetization potential? Thank you.

Management's Response:

First of all, regarding AI features, we can roughly categorize them into several types. One is obviously our Yuanbao, which is a native AI application. Secondly, there are AI-enhanced search features related to search, which have been implemented in our browser and WeChat search.

Additionally, even in games, we have many AI-enabled features, such as AI assistance for players; in productivity tools like the new Tencent Meeting, there are meeting content summaries, and in Tencent Docs, there is AI-assisted writing, etc. These all belong to different application scenarios of AI features.

I believe we are still in the early stages of observing user behavior. Clearly, from user feedback, users are generally more satisfied when using these AI features. Whether it's content creation or achieving desired content, efficiency has improved, and overall user experience has been enhanced.

Regarding your question about whether AI search will harm the search business, for example, if AI directly provides answers and users no longer click through to pages, I think we haven't seen any significant negative impact so far. Although some users may directly obtain answers, if they want to delve deeper into a topic, they will still click on relevant links and articles. Overall, this impact is very limited.

At the same time, from the perspective of WeChat's content ecosystem, many page views actually do not come from search; more are from users directly following certain authors, so mainstream content browsing still primarily takes this form. Therefore, at present, the impact on specific creators is also very limited.

Regarding your question about costs, we are currently quite meticulous in controlling the costs of AI services. In many scenarios, if we can use small models, we will opt for small models, which are much cheaper than large models. In scenarios where small models can be used, and with continuous upgrades in software efficiency, the overall inference costs remain controllable.

As for whether we will directly monetize C-end AI features in the future, I believe there will ultimately be some commercialization. However, in China, the model of users paying directly is actually difficult to popularize like in the U.S. AI tools. Therefore, we will continue to explore, such as whether we can support the monetization of these AI features through advertising.

I would also like to add that AI itself is already helping to drive the growth and monetization of our existing businesses in various ways. Therefore, we can also understand that the growth of other businesses subsidizes part of the costs for AI user usage. Thank you.

Questioner asks:

Hello, management, thank you for answering my question. I would like to ask about the impact of the new advertising law on gaming companies, as well as marketing issues under the new advertising regulations (which will take effect in July). According to the new regulations, if sales and marketing expenses exceed 15% of revenue, an additional 25% tax must be paid. How will this affect our advertising revenue, especially in the mini-game business? Because mini-games heavily rely on traffic acquisition, their sales and marketing expenses can easily exceed 15% of revenue. Thank you Management's Response:

Hello. We expect that this new regulation will not have a substantial impact. In fact, our advertising business has become quite diversified. From the second quarter, although there were some negative impacts, mainly due to some food delivery companies and certain e-commerce enterprises increasing their investment in food delivery services, thereby reducing their advertising spending and allocating more funds for subsidies. Nevertheless, our advertising revenue still grew by 20% year-on-year. Therefore, we believe that advertising investments in various sub-industries will fluctuate, which is quite normal. However, more important than these fluctuations is actually our application and layout of AI in the advertising business. Thank you.

Questioner:

First, I would like to ask a few questions about games. We can see now that, for example, "Delta Force" has launched very successfully. So I would like to ask the management to share their views on the growth of AAA games in the Chinese market? Including PC, content, and tools, does this mean a new market opportunity for your company, or will it replace part of the current market? The second question is about capital allocation for AI. Compared to our current incremental strategy, what do you think about significantly increasing investments in new upstream entry points like Yuanbao and IMA to create new dominant traffic entry points?

Management's Response:

Thank you for your question. Just to confirm, your first question is about AAA games, not specifically about "Delta Force," right?

Questioner:

Yes, I want to ask about the market opportunities reflected in the rapid growth of products like "Delta Force."

Management's Response:

We define AAA games as "buyout" games, such as "Blacksmith" and "Wukong." We believe "Delta Force" is more like a "long-term operation" or "game as a service" product type.

We are confident in both types, but we believe that most of the market in China now and in the future will still be dominated by long-term operational service games like "Delta Force." We are very satisfied with the performance of "Delta Force," not only because it ranks in the top three or five on certain charts, but more importantly, since its launch, both user numbers and revenue have continued to grow over the past nine months, which aligns perfectly with our expectations for evergreen games. This sustained growth also proves our capability to produce high-quality game content on a large scale. Currently, "Delta Force" covers three gameplay modes and two platforms, and it will be available on the console platform tomorrow, becoming three platforms, and it has already covered both the Chinese and overseas markets. We are able to achieve this thanks to the capabilities of our development team, the maturity of our tools and production processes (which also includes the application of AI technology).

So we are very satisfied with the current performance of "Delta Force." As for AAA or buyout games, the audience for such products in China is also expanding. We believe that as long as we produce truly excellent AAA games, there is an opportunity for considerable profitability—though the standards for achieving "excellence" are very high. Overall, since Chinese players are accustomed to free game models, AAA or buyout games remain niche in the Chinese market, but their emergence brings vitality and breakthroughs to the market Questioner:

The second question is about AI capital allocation. What is your view on the trade-off between increasing investment to create new traffic entry points (such as Yuanbao or IMA) and the current steady progress?

Management's Response:

In terms of product promotion, especially in the first quarter, we did invest quite a bit to promote our products. This led to significant user growth. The focus in the second quarter is to optimize the product and improve user experience to better retain the large number of users we have acquired. If we can enhance the user experience sufficiently, we will indeed strengthen subsequent promotions. However, we do not intend to simply "spend money" on market placements to acquire new users. It is important to note that we already have many existing platform resources, which is a significant advantage for us. Deeply integrating Yuanbao with our existing platform will become a key competitive strength, and we will continue to deepen this integration in the coming months. At the same time, we will appropriately increase the market promotion of our products. So, thank you very much, everyone. Thank you, Robin.

Questioner from Citigroup:

Hello, management, good evening. Thank you for answering my question. I have two questions.

The first question is about the growth of enterprise service revenue mentioned by management. We have seen business acceleration this quarter, mainly benefiting from the increased demand from enterprise clients for GPU leasing and API Token usage. Now we are starting to see the actual impact of AI in internal applications. Will Tencent invest more GPU resources in the future to support the growing external demand from enterprise clients? If so, is there a prospect for sustained accelerated growth in revenue from this business line in the coming quarters?

The second question is a follow-up on the advertising business. With the upgrade of foundational models and the continuous expansion of AI capabilities, I remember management mentioned last quarter that the market service line achieved a 20% year-on-year growth, which is already at a high growth range. With the empowerment of AI technology, is it possible for this 20% to become a new lower limit for growth, especially if AI continues to improve conversion rates and CPA/ECPN performance, particularly given that our advertising load levels are still relatively low compared to peers? Thank you.

Management's Response:

I will answer your two questions. First, regarding the growth of enterprise service revenue, we have indeed seen revenue growth from GPU and API Token leasing this quarter, but overall, this is actually a broad-based growth. Over the past two to three years, we have specifically optimized our cloud business, reducing low-margin and low-value-added activities. Now, we have been able to operate our cloud business on a more sustainable basis, and the cost competitiveness of our supply chain has also improved, allowing us to refocus on a faster revenue growth pace without overly relying on short-term fluctuations in GPU supply. In other words, if GPU supply is sufficient, we will further enhance GPU leasing in cloud computing, but our cloud business strategy does not solely depend on GPUs; we are continuously expanding in areas such as CPUs, storage, databases, and CDN.

Regarding your question about the advertising business, we are currently satisfied with the growth rate achieved. In the short term, our strategy remains unchanged, still focusing on laying a solid foundation for future growth. Of course, if AI costs, including GPU depreciation, suddenly increase significantly, putting a lot of pressure on us, we may accelerate the pace of advertising monetization, but as of now, there is no need for that Thank you.

Management's Response:

In fact, we still have a lot of work to do, right? I think we can categorize it into several major aspects. First is the large language model itself; we certainly hope to continuously enhance the capabilities of the LLM. This actually includes multiple different dimensions, such as improving the quality and coverage of data to make it higher quality and more comprehensive; making the pre-training process more efficient and effective, including optimizing our pre-training models; and also improving the post-training and reinforcement learning processes to better tap into the capabilities of the pre-trained models. In addition, we are also continuously improving our infrastructure, so that we can train more efficiently on one hand, and perform inference more efficiently on the other. By enhancing inference efficiency, we can better utilize GPUs and reduce the costs of providing services to users and clients.

As our LLM capabilities enhance, it becomes the cornerstone of all AI services, especially driving our progress in search and productivity-related services. Of course, we also hope to further enhance the model's multimodal capabilities, which can provide more customized features for Yuan Bao users.

Currently, in Yuan Bao, users not only use it for search and productivity-related scenarios but also for a variety of multimodal functions, such as speech-to-text, image-to-text, and text-to-image generation. There are a large number of multimodal conversion scenarios within Yuan Bao, which requires us to have very strong multimodal processing capabilities.

The third major direction is coding and agents. If we can continuously improve in this area, we can provide a better coding environment for ourselves and our enterprise clients. At the same time, this will also enable our agents to have stronger instruction understanding and execution capabilities, which is particularly important for the future of WeChat. We hope to become a personalized assistant for users when building intelligent assistants for WeChat. Therefore, we are continuously enhancing our capabilities across the above dimensions.

Continuously updating...

Risk Warning and Disclaimer

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 are suitable for their specific circumstances. Investing based on this is at one's own risk