Tencent (Minutes): Adjust AI investment dynamically according to demand, balance capital allocation

The following is the $TENCENT(00700.HK)$Tencent(TCEHY.US) 4Q24 earnings conference call minutes. For financial report interpretation, please refer to《 Tencent: Crazy for "AI", is this stock king love or hate? 》

1. Review of Core Financial Information

1. Annual and Fourth Quarter Financial Performance: The revenue growth rate improved in 2024, achieving double-digit growth in the fourth quarter. Annual growth and operating profit growth outpaced revenue due to a shift towards high-quality revenue streams. Total revenue in the fourth quarter was RMB 172.4 billion, a year-on-year increase of 11%; gross profit was RMB 90.7 billion, a year-on-year increase of 17%; non-International Financial Reporting Standards operating profit was RMB 59.5 billion, a year-on-year increase of 21%; non-International Financial Reporting Standards net profit attributable to shareholders was RMB 55.3 billion, a year-on-year increase of 30%.

2. Shareholder Returns: Focused on delivering returns to shareholders, paid HKD 32 billion in cash dividends in 2024, and repurchased shares worth HKD 112 billion. Plans to repurchase at least HKD 80 billion worth of shares in 2025, and if approved by the shareholders' meeting, will propose a cash dividend of HKD 4.5 per share, an increase of 32% from the previous year, with dividends to be paid to shareholders on May 30, 2025.

3. Gross Margin and Expense Situation: The overall gross margin in the fourth quarter was 53%, an increase of 3 percentage points year-on-year.

The gross margin for value-added services was 56%, an increase of 2 percentage points year-on-year, due to an increased proportion of high-margin domestic game revenue, improved profit margins from music subscription revenue, and a decreased proportion of low-margin live streaming revenue.

The gross margin for marketing services was 58%, an increase of 1 percentage point year-on-year, benefiting from the growth of high-margin video accounts and WeChat search revenue.

The gross margin for financial technology and enterprise services was 47%, an increase of 3 percentage points year-on-year, due to an increased proportion of high-margin revenue and improved cost efficiency in cloud services.

Sales and marketing expenses in the fourth quarter were RMB 10.3 billion, a decrease of 6% from the high base of the same period last year, accounting for 6% of revenue, down from 7% in the same period last year. Total administrative and management expenses were RMB 31.4 billion, a year-on-year increase of 16%, mainly driven by increased R&D expenses, which rose 21% year-on-year to RMB 19.8 billion Mainly due to increased employee costs and GPU service depreciation related to artificial intelligence initiatives. Administrative and management expenses after deducting R&D increased by 8% year-on-year to RMB 11.6 billion. The number of employees at the end of the quarter was approximately 111,000, an increase of 5% year-on-year and 2% quarter-on-quarter.

The operating profit margin for the fourth quarter under non-International Financial Reporting Standards was 34%, an increase of 3 percentage points year-on-year, consistent with the expansion of gross profit margin.

4. Cash Flow and Capital Expenditure: Capital expenditure for the fourth quarter was RMB 34.9 billion, an increase of 421% year-on-year, mainly due to increased investment in GPUs and services to enhance artificial intelligence capabilities. Non-operating capital expenditure was RMB 1.7 billion, an increase of 103% year-on-year, mainly driven by the construction of data centers. Total capital expenditure was RMB 36.6 billion, an increase of 386% year-on-year.

Free cash flow was RMB 4.5 billion, a decrease of 87% year-on-year, mainly due to increased capital expenditure on GPUs, servers, and data centers. Quarter-on-quarter, free cash flow decreased by 92% due to timing differences in the settlement of certain accrued expenses, increased capital expenditure on GPUs and services, and lower seasonal total revenue from games.

Net cash was RMB 76.8 billion, a decrease of 20% quarter-on-quarter, reflecting cash outflows related to capital expenditure and share buybacks.

II. Detailed Content of the Earnings Call

2.1 Key Information from Executives' Statements

1. Business Development Highlights

Strengthening Core Service Capabilities: In 2024, the long-term operational value of key businesses was strengthened by launching a mini-program store and upgrading WeChat search, enhancing transaction capabilities. The usage duration of video account users rapidly increased due to enhanced recommendation algorithms, and the evergreen game portfolio increased from 12 titles in 2023 to 14 titles in 2024, with new games launched that have evergreen potential. Tencent Video had three works ranked among the top five in the industry in 2024, and Tencent Music expanded its industry-leading position with 121 million subscription users.

Marketing service revenue growth outpaced the industry, upgrading the advertising technology platform and increasing user traffic. In fintech, risk control was upgraded and payment funding costs were optimized, enhancing the overall fintech franchise and profitability.

Progress in Artificial Intelligence: Advancing the capabilities of the Hunyuan large model, applying artificial intelligence to internal use cases, and achieving breakthrough growth in consumer-level artificial intelligence use cases.

2. AI Strategy and Progress

1) Research and Development of Basic Large Models: Tencent's artificial intelligence initiatives date back to the establishment of its artificial intelligence laboratory in 2016. Since the beginning of 2023, significant investments have been made in the self-developed Hunyuan large model, which provides solutions in infrastructure, algorithms, training, alignment, and data management.

The Hunyuan T1 deep thinking model was also released, which is the first open-source model supporting text, image to 3D generation, and actively contributes to the open-source community by open-sourcing a series of models in the Hunyuan series for 3D generation, video generation, large language, and image generation, some of which are widely popular among global developers.

2) Consumer-Facing AI Products: A multimodal strategy is adopted to provide users with the best artificial intelligence experience, utilizing existing models to meet different user needs. AI-native applications can access multiple models, including thinking chain reasoning models such as Hunyuan T1 and DeepSeek R1, as well as the Hunyuan Turbo S rapid thinking model It can also choose to integrate web search results, allowing direct access to high-quality proprietary content within the Tencent ecosystem. From February to March, daily active users grew 20 times, becoming the third highest AI-native mobile application in China in terms of daily active users. In addition, testing of artificial intelligence features has begun in WeChat to enhance user experience, with more features to be added in the future.

3) AI services for enterprises: Accelerating the integration of artificial intelligence into cloud business infrastructure, platforms, and software-as-a-service solutions. Through infrastructure-as-a-service solutions, enterprise customers can achieve large-scale high-performance AI training and inference capabilities, while developers can access and deploy mainstream foundational models.

The TI platform flexibly supports model fine-tuning and inference needs, providing powerful solutions for enterprise customers and developers. SaaS products benefit from AI tools, with real-time transcription and meeting summary features seeing monthly active users double year-on-year to 15 million, while Tencent Docs has also enhanced user productivity and content generation processing capabilities.

In 2024, AI cloud revenue is expected to roughly double year-on-year, but the increased allocation of GPUs for internal use cases has limited the provision of GPUs to external customers, thereby constraining the growth of cloud service revenue. Since the fourth quarter of 2024, GPU procurement has increased, and overall cloud service revenue growth is expected to accelerate with the deployment of GPUs.

4) Empowering existing businesses: In the advertising business, since 2015, neural network AI capabilities have enhanced the advertising system, and in 2020, large model capabilities were used to rebuild the technology platform. In 2023, large model capabilities were added to achieve more precise ad placement and creative generation, with marketing service business revenue expected to grow by 20% in 2024. In the gaming business, since 2017, machine learning technology has been adopted in PVP games, and large language model capabilities are now being integrated to optimize matching experiences, improve 3D content production efficiency, and empower in-game chatbots. In video and music services, artificial intelligence is used to enhance content creation productivity, with content recommendation algorithms driven by AI effectively promoting content discovery.

3. Future investment plans: As the capabilities and advantages of artificial intelligence become increasingly apparent, investments in AI have been increased to meet internal business needs, train foundational models, and support user inference demands. The AI team has been reorganized to focus more on rapid product innovation and deep model research. Annual capital expenditures for 2024 are expected to increase to $107 billion, approximately 12% of revenue, with plans to further increase capital expenditures in 2025, expected to account for a low double-digit percentage of revenue. In R&D, investments will continue in proprietary models, accelerating the development of AI applications across various business groups, and marketing investments will be made to raise user awareness of new AI products and promote their adoption.

4. Performance of various business segments

(1) Value-added services: In the fourth quarter, value-added service revenue reached RMB 79 billion, a year-on-year increase of 14%. Social network revenue grew by 6% year-on-year, driven by increased sales of in-app game items, music subscriptions, and service fees from mini-game platforms Music subscription revenue increased by 18% year-on-year, with the number of subscribers rising to 121 million. Tencent Music deepened its cooperation with record companies and artists and added super VIP privileges. Long video subscription revenue grew by 3% year-on-year, with subscribers increasing to 113 million. The self-produced drama "Eternal Night Star River" became the most-watched series across all online platforms in China in November.

Domestic game revenue increased by 23% year-on-year, driven by evergreen games such as "Honor of Kings," "Peacekeeper Elite," and "Valorant," as well as newly released games "Dungeon & Fighter Mobile" and "Delta Force." International game revenue grew by 15% (16% growth at constant exchange rates), with "PUBG Mobile" performing strongly.

Communications and social networks: The mini-program store gradually added features to stimulate new and repeat transactions, such as the launch of a gifting function, and leveraging WeChat social interactions to promote transactions, while building a delivery location map to facilitate consumer shopping and merchant delivery. The search query volume and revenue on WeChat continued to grow rapidly, integrating Tencent's Hunyuan and DeepSeek large language model capabilities to enhance the relevance and quality of search results, with Tencent's model-driven results covering over 90% of question-based searches.

Gaming business: In the fourth quarter, several evergreen games benefited from IP collaborations and high-profile events. "Honor of Kings" achieved double-digit revenue growth year-on-year due to high daily active users and popular outfits based on the anime "Detective Conan." "Valorant" saw its total revenue double due to collaborations with world champion team EDG and the Riot animated series "Arcane Season 2." "Teamfight Tactics" also achieved double-digit revenue growth. The newly released game "Delta Force" generated over 1 billion RMB in total revenue on PC and mobile in China during the fourth quarter.

Game reserves include highly anticipated titles such as "Assassin's Creed: The Hidden Ones," "Goddess of Victory: New Hope," and "Valorant Mobile." During the Spring Festival of 2025, the five highest-grossing games had daily active users exceeding those during the Spring Festival of 2024. In the international gaming sector, Finland's Supercell's "Brawl Stars" became the third-highest mobile game by daily active users outside of China in 2024, with total revenue in the fourth quarter increasing several times year-on-year. New Zealand subsidiary Grinding Gear Games' "Path of Exile 2" ranked first in Steam's paid game revenue for six consecutive weeks after its early access release in December. Canada's Digital Extremes' "Warframe" released a major update "Warframe 1999" in the fourth quarter, achieving record highs in daily active users and total revenue in 11 years, with total revenue for 2024 increasing by over 30% year-on-year.

(2) Marketing services: Fourth-quarter marketing service revenue grew by 17% to 35 billion RMB, benefiting from increased user engagement and ongoing AI upgrades of the advertising platform. Enhanced AI models provided a more comprehensive understanding of user interests and responses to advertisements, resulting in more relevant ad recommendations, with revenue growth across most industries. By inventory classification, revenue from marketing services on video accounts grew by over 60% year-on-year, marketing service revenue from mini-programs grew rapidly, and WeChat search revenue more than doubled year-on-year, benefiting from more commercial queries and AI-optimized ad placements and formats, which improved click-through rates (3) Financial Technology and Enterprise Services: In the fourth quarter, this sector's revenue was RMB 56.1 billion, a year-on-year increase of 3%. Financial technology service revenue saw a recovery to low single-digit year-on-year growth this quarter, benefiting from increased commercial payment transaction volumes and growth in wealth management and consumer loan service revenues. Commercial payment revenue remained stable year-on-year, with transaction volume growth in the fourth quarter outpacing that of the third quarter. Enterprise service revenue saw moderate year-on-year growth in the fourth quarter, benefiting from increased cloud service revenue and higher technology service fees due to rising e-commerce transaction volumes. However, as mentioned earlier, allocating more GPUs for Tencent's own needs temporarily limited the growth of cloud service revenue. The gross margin for enterprise services increased year-on-year, with WeChat Work revenue more than doubling year-on-year, and Tencent Meeting revenue growing over 40% year-on-year, benefiting from increased enterprise adoption.

2.2 Q&A

Q: With increased capital expenditure in artificial intelligence, additional depreciation and R&D expenses will inevitably impact profit margins. In the past few years, we have focused on high-quality growth, significantly improving profit margins. How should we balance growth and profitability enhancement in the future?

A: We do not believe that R&D expenditure itself will pressure profit margins. Historically, Tencent has increased R&D spending on various projects each year. Capital expenditure is a more nuanced topic; last year's fourth quarter saw capital expenditure rise to a new higher stable level. Over time, incremental capital expenditure will translate into incremental depreciation in the coming years. However, it is essential to understand the direction of capital expenditure to assess whether depreciation will put pressure on profit margins.

The uses of capital expenditure mainly include the following aspects:

First, it is used to support GPUs for advertising technology and gaming businesses. The results show that these capital expenditures can generate good profit margins and high returns;

Second, it is used for GPUs for training large language models. With breakthroughs in DeepSeek, the industry and we can now achieve higher productivity in training large language models on existing GPUs without needing to increase additional GPUs at the previously expected pace;

Third, capital expenditure related to cloud business, where we purchase GPU services and lease them to customers, can generate returns. Although it is not the highest-return business in our portfolio, it can cover GPU costs and related depreciation;

Fourth, capital expenditure for consumer-facing inference may bring pressure in the short term, but we believe this is a controllable cost pressure, as it is only a part of total capital expenditure. We optimistically believe that over time, consumer-facing inference activities will become profitable through advertising revenue and value-added services.

Overall, we are confident in continuing to drive business growth while protecting profit margins. Additionally, in consumer-facing product inference, we can reduce unit costs through technology means such as software and better algorithms.

Q: The WeChat gifting feature launched in the third quarter of last year has been very popular among users. What are the strategies and key initiatives for the e-commerce business this year?

A: The mini-program store is a long-term plan for us, and the current initiatives are just one of many measures to build this ecosystem. The mini-program store is a unified platform connecting various components of the WeChat ecosystem, with standardized and indexed product information and data structures, facilitating the free flow of product information among different components of the WeChat ecosystem, aiming to help consumers find quality products and merchants The WeChat ecosystem includes multiple components such as social infrastructure, content, search, mini-programs, and transaction platforms, where the mini-program store can operate.

The gifting feature is one of the functions within the social component of the ecosystem. From initial feedback, this feature was very popular during the Spring Festival, generating a good word-of-mouth effect and helping to complete transactions. Currently, this is just one component of the WeChat ecosystem, and we will patiently continue to build this ecosystem, viewing it as a marathon rather than a sprint. In the fourth quarter of last year, the gross merchandise volume (GMV) of the mini-program store continued to grow rapidly.

Q: What is the current demand and supply situation for cloud services, and when will the service demand growth be met?

A: Currently, the demand for cloud services is very strong, but supply is limited. One reason for the significant increase in capital expenditures in the fourth quarter is the large number of urgent orders for GPUs used for inference and cloud services. Only by installing these GPUs in data centers can we meet the substantial growth in service demand, which will take some time. The first quarter may not fully capture the demand growth, but as GPUs arrive and are installed, we will be able to meet a considerable portion of the demand in the future.

Q: What software solutions does Tencent offer, and how are they performing?

A: Tencent offers products such as Tencent Meeting and Tencent Docs, with Tencent Docs being the largest revenue-generating SaaS product, achieving double-digit year-on-year revenue growth last quarter. Additionally, there are security PaaS software and audio-video PaaS software, including real-time communication and live streaming software sold to cloud customers, many of which can provide additional value to customers through AI.

Q: How are Tencent's consumer-facing applications performing?

A: Tencent has a large number of consumer-facing applications, and there will be more in the future. AI is still in its early stages, making it difficult to predict the final state. Each product will continue to evolve, providing users with more practical and powerful features. For example, Yuanbao can become a powerful native assistant, and ima Copilot can serve as a knowledge base for personal and team collaboration, while WeChat will introduce more features. Other products like QQ and the browser will also have AI experiences. Each product will seek unique use cases, leveraging AI to provide a good user experience, and different products will collaborate to help AI products expand their user base.

Q: What is Tencent's strategy for further expanding the Yuanbao user base in a competitive market, and which part of Yuanbao as a hybrid AI and AI search provider does Tencent favor?

A: Currently, Yuanbao is a chatbot and search application, and it will develop into an AI assistant with multiple functions serving different groups of people in the future, with a wide range of application scenarios. Its unique advantage lies in innovation, continuously adding features to meet user needs, and it can also connect to Tencent's content ecosystem, such as public accounts and video accounts as quality information sources.

The multi-model strategy helps users access the best models, using a combination of models to meet complex needs. In the future, many high DAU products will add AI features, and some products will establish pathways with Yuanbao, with products collaborating.

Q: What share of search queries in the overall search market will AI search ultimately occupy in the future, and how will AI search be monetized? A: Different people have different views on this, and it requires time to verify. Historically, looking at how online search replaced online directories, AI search has the potential to replace traditional search because AI search brings new technologies, new efficiencies, and trading capabilities. In terms of monetization, the West first adopted a subscription model, which will later develop into performance advertising; in China, it is expected to start with performance advertising and then develop value-added services.

Q: Tencent's advertising business saw a healthy growth of 17% in the fourth quarter. Can this growth rate be considered a reference for the advertising business growth rate in 2025? With more AI applications, is there a possibility of accelerated growth in marketing services?

A: The company is very satisfied with the growth rate of its advertising business in the fourth quarter, which significantly exceeds the industry level and is a relatively natural growth that covers almost all industries monitored by the company. This is because the AI-enhanced features deployed across various industries have brought advertisers higher returns on investment than before and through other channels. Compared to some global peers, who adopted higher advertising loading rates early in product development, the company will gradually increase the advertising loading rate for new products like short videos. As long as there are no significant changes in the macro environment, the company is quite confident in its advertising business.

Q: How do you view the growth prospects of Tencent's domestic and international gaming business this year, especially considering the high base effect of the domestic gaming business in the second half of this year? What impact does AI have on the gaming business?

A: From observable facts, the company's deferred revenue (mostly from the gaming business) grew by more than ten percent year-on-year at the end of last year, and this deferred revenue will convert into reported revenue in the first half and second half of 2025, and even part of 2026. During this year's Spring Festival, the daily active users of the company's five highest-grossing games grew year-on-year, which is a positive leading indicator.

Subjectively, the company has several games that are expected to become evergreen games, such as "Delta Force"; the company also has several new games in preparation. AI technology has direct and indirect benefits for the gaming business.

Directly, game developers can use AI to create more content faster and serve more users more effectively; indirectly, from a longer-term perspective, if people use AI more widely, they will have more time and willingness to engage in high-autonomy activities, and games, as an interactive entertainment form, are a great way to express this. Additionally, from a competitive dynamics perspective, AI can make evergreen games even more evergreen, allowing for high-quality content to be produced in a shorter time during game development and frequent updates; in PVE experiences, it can make games more exciting, and in PVP, it can better match, balance, and guide new users, making popular large games more appealing to users.

Q: What is the trend of commercial payment activities in the first quarter of 2025?

A: There has been an observed further increase in transaction volume, but the average selling price (ASP) still faces pressure. From a total revenue perspective, commercial payments are flat compared to last year. Consumer spending tendencies have rebounded, but there is still significant pricing pressure on the merchant side. If consumer demand continues to improve, over time, price competition among merchants is expected to ease, leading to potential value growth, but this still requires further observation.

Q: What are the reasons for maintaining the capital expenditure (CapEx) to revenue ratio at a low double-digit level in 2025? Is it due to the expected slowdown in demand for generative AI, or is the significant investment in 2024 sufficient to meet the demand in 2025?

A: The increase in investment at the end of 2024 should be sufficient to meet the demand for generative AI and other needs in 2025. There is a time lag from ordering GPU services to fully deploying them in data centers, which will be in this phase for part of the fourth quarter of 2024 and the first quarter of 2025, but by the end of the first quarter, the deployed GPUs will meet internal inference needs, Yuanbao demand, and generate direct revenue from Tencent Cloud renting to customers.

Overall, the company's capital expenditure was relatively large compared to its Chinese and global peers for a period last year, and it had the largest capital expenditure among Chinese tech companies in the fourth quarter. However, the capital expenditure to revenue ratio of Chinese tech companies is generally lower than that of some Western peers because Chinese companies focus more on the efficient utilization of GPU servers, which does not affect the ultimate results of technological research and development, as evidenced by the success of DeepSeek. However, this situation is dynamic; if search demand suddenly increases, it will lead to increased GPU orders, and the company will respond flexibly to market dynamics.

Q: Which vertical industries benefit the most from the advertising business? How have advertising client behaviors changed, and how has this promoted advertising revenue growth?

A: Advertising categories that achieved year-on-year growth include e-commerce, finance, fast-moving consumer goods, gaming, local services, education, and healthcare. With the application of AI technology, it is first deployed in video accounts and then gradually penetrates other WeChat businesses, Tencent's other businesses, and the advertising network. Over time, the benefits of the advertising business will become increasingly widespread.

Q: What are the company's priorities in capital allocation? How do you view merger and acquisition investments in the current context of strong demand in the AI industry?

A: The primary principle of capital allocation is to create returns for the company and its shareholders, with different investment focuses at different times. In the past, a significant amount was invested in ecosystems and partnerships, establishing a large portfolio, and then returning cash to shareholders through dividends and stock buybacks. Now, there is great potential in the AI field, and we hope to formulate investment strategies for future investments while providing current returns through dividends and buybacks.

The company has strong financial capabilities, robust operating cash flow, and a large and valuable portfolio, a significant portion of which is highly liquid, providing sufficient resources for future investments, including in the AI field and necessary investment activities. This year, it has been announced that HKD 41 billion will be paid in cash dividends, and stock buybacks have decreased from HKD 100 billion last year to HKD 80 billion, reserving HKD 20 billion as flexible buffer funds to observe whether further investment in AI is needed. Overall, the company is striving to balance current returns and future investments, with ample financial resources to support this.

Q: How do you view the profit margin expansion situation in 2025 and the coming years?

A: Over the past two years, the company has enjoyed operational profit margin leverage, primarily due to the improvement in gross profit margins. The comprehensive gross profit margin of the core business is about 50%, while some new revenue streams contribute significantly to revenue growth, with gross profit margins reaching 70% - 80%. This shift in business structure has raised the overall gross profit margin, allowing gross profit to grow faster than revenue In the long term, as the overall gross margin approaches the gross margin of new businesses, the rate of increase in overall gross margin will gradually slow down. However, the company's high-quality revenue streams, such as video account advertising, search advertising, certain value-added financial services, e-commerce commissions, and proprietary games, typically grow faster than overall revenue and contribute profit margins that are significantly higher than the overall profit margin. Therefore, the company believes it will continue to gradually enjoy gross profit leverage, thereby achieving growth in operating profit relative to revenue. Of course, investments in artificial intelligence may have some counteracting effects that need to be balanced.

Q: We see many companies releasing their large models or AI agents. What does management think about the future core competitive factors in this field, and where are our advantages? How does the company's management view the key competitive factors for large language models, what is the company's competitive advantage, and how will we enhance user engagement by adding features in the future?

A: The market will have various types of AI agents in the future, which essentially utilize model capabilities and connect different software tools to complete complex tasks. The company can leverage high-quality models to build independent AI agents while relying on a large user base across different software platforms, such as browsers.

AI agents can also be set up within applications like WeChat and QQ, which can utilize the in-app ecosystem to complete complex tasks for users and provide high-quality services. Taking WeChat as an example, its advantages are first, a large user base with long daily usage times and high frequency of opening; secondly, the activities within WeChat are very diverse, covering socializing, communication, content, work, learning, transactions, and more, along with numerous mini-programs that can easily connect to many mini-programs based on the model to help users complete activities and complex tasks. However, the company will cautiously and patiently build these experiences while focusing on user data security comfort, and safety.

The risk disclosure and statement of this article: Dolphin Research Disclaimer and General Disclosure