Tencent Q4 2025 earnings conference call
Key Takeaways (AI-Generated)
Financial Performance
- Total revenue increased 13% year-over-year to 194.4 billion RMB in Q4 2025
- Gross profit grew 19% year-over-year to 108.3 billion RMB with margin improving to 56%
- Non-IFRS net profit attributable to equity holders reached 64.7 billion RMB, up 17% year-over-year
- Free cash flow was 34 billion RMB, increasing over six times year-over-year
Business Highlights
- Delta Force sustained top three industry ranking, surpassing 50 million peak daily active users in February 2026
- Hunyuan 3.0 in internal testing represents bigger capability step versus Hunyuan 2.0 than previous generation
- Tencent Cloud achieved 5 billion RMB adjusted operating profit in 2025 after breaking even in 2024
- Mini programs user time spent increased over 20% year-over-year driven by productivity tools and games
Financial Guidance
- Expect to more than double investments in Hunyuan, Yuanbao and other AI products in 2026
- AI spending was 7 billion RMB in Q4 2025 and 18 billion RMB for full year 2025
- Revenue may grow faster than profit in 2026 due to stepped up AI investment
- Proposed annual dividend of 5.3 Hong Kong dollars per share, reflecting 18% year-over-year increase
Opportunities
- Developing Hunyuan 3.0 foundation model with enhanced multimodal capabilities including 3D text-to-image features
- AI agents in Weixin to deliver productivity benefits and boost ecosystem activity
- Deepened collaboration with e-commerce platforms facilitating merchant advertising within Tencent ecosystem
- AI integration enhancing existing services including content recommendation, customer service, and fraud protection
Full Transcript (AI-Generated)
Operator
Of customer service history for merchants, our enterprise software products Recom and Tencent meeting are leaders in their categories in China in terms of usage and revenue. For Fintech, we utilize lightweight AI models to enhance credit scoring processes and facilitate fraud protection, contributing to us sustaining better than industry non performing loan rates.
Martin Lau
Now that our businesses are benefiting operationally and financially from integrating AI, we believe we are in a position of strength to add development of new AI products to our priorities. At the foundation model layer we see substantial opportunities from combining a strong foundation model with configuration for core user cases such as chatbot, coding, multimodal and agentic applications.
Although we're not the first mover in large language models, having already revamped our team, improved our data quality and rebuild our AI infrastructure for pre training and reinforcement learning, we're now iterating more intelligent models at a faster pace. Hunyuan 3.0 in internal testing currently represents a bigger step in capabilities versus Hunyuan 2.0 than 2.0 was versus 1.0.
From multimodal capabilities our 3D text to image and world models are early category leaders and will increasingly benefit from leveraging our proprietary data and abundant use cases. Some observers in China tech are single mindedly focused on AI chat bots as the only means for bringing AI to users. We believe this mindset is overly simplistic because AI can help people in a multitude of ways beyond powering an information advice app.
We believe that AI chatbot applications are largely competing with search applications rather than with every other application. For Yuanbao, our own AI chatbot app, we focus on finding product market fit and use cases which belong in Chatbot AI app by rapidly iterating Yuanbao to enhance its user experience by providing better search integration, improved speech recognition, easier access to multimodal capabilities and exploration around group chat which we believe will increase usage and user retention of the app.
In coming months, as we deploy Hunyuan 3.0 in Yuanbao, we believe the core user experience will step up further. In addition, we've also integrated AI to enhance a range of existing user experiences with innovation, including content consumption, information retrieval and merchandise recommendation and customer service.
We're building AI agents, which are to autonomously interact on behalf of users with innovation functionalities, especially mini programs. Excitement around claw bots illustrates that people recognize AI can unlock computer use capabilities to improve their daily lives, but also illustrate the risks around unleashing unsupervised AI.
We want AI agents in Weixin to deliver AI productivity that's beneficial to the general public as well as early adopters, and which will boost ecosystem activity and naturally generate revenue. AI agents are currently powered by a multiplicity of foundation models, and we expect that users at the application level will continue to have access to range models.
However, improving the performance of Hunyuan will enable us to offer new unique to Weixin agent capabilities, so the Weixin and Hunyuan teams will work increasingly closely together going forward. Speaking of claw bots, we have introduced a number of AI tools for enhancing productivity, including Work Buddy, Q Claw and Tencent Cloud Lighthouse and we provide downloadable skills to easily put these tools to use from our skill hub.
Claw bots are upgrading AI from thinking to doing via autonomous workflows and continuous task execution. Users control this new generation AI tools through command line interfaces in their existing communication tabs, which generally means Weixin and QQ in China as it's efficient for users to interact with digital agents in a place and format where they are already interacting with human contacts.
The new AI products that I described require substantial and increasing investment which we believe would generate significant return for us over the long run. Our spending on our two biggest new AI products Hunyuan and Yuanbao was 7 billion RMB in the fourth quarter of 2025 and 18 billion RMB for the full year.
These figures are only for Hunyuan and Yuanbao and exclude AI initiatives supporting our existing products and services, as well as exclude costs arising from providing GPUs to external customers via Tencent Cloud. We expect to more than double these investments in Hunyuan, Yuanbao and other new AI products in 2026, which we intend to fund from increasing earnings from our core businesses.
In this transformational period, we are breaking out our investment in new AI products because we view these strategic investments conceptually similar to investment in affiliates or to CapEx. These are upfront investments required to build the necessary foundation to unlock new value as opposed to ongoing operational expenses.
As such, we believe the impact of these investments should be viewed separately from the profits generated by our existing businesses. Over time, we're confident that monetization will follow usage for these new AI products.
Lastly, I would like to present a case study on Tencent Cloud as the latest example on how we develop our services into market leaders with economic returns over time, and that would follow games, payments and long form video and we expect it will be the same for our new AI products.
Tencent Cloud was a relative late entrant in cloud services. However, we committed to a patient and long term investment strategy, believing that it had scale from the start due to Tencent itself being the biggest single end user for a range of technology infrastructure in China and that it could provide differentiated services arising from Tencent's unique insights, ecosystem and capabilities.
For example, we believe that we were the first cloud service provider in China to fully recognize the stepped up capabilities of AMD's recent generations of CPUs, becoming AMD's largest partner in the country and that our cloud video streaming service is the industry leader in terms of streaming quality.
After a period where Tencent Cloud prioritized revenue growth somewhat misguided by other industry participants, in 2022, we aggressively restructured Tencent Cloud to focus on high quality services rather than chasing high revenue but low value added activities such as reselling and customizing projects.
This pivot cost us several quarters of revenue growth, but it enabled Tencent Cloud to achieve operating profit break even in 2024, up from significant losses in prior years. During 2025, although Tencent Cloud continued to face revenue headwinds due to limited availability of GPU for external customers as we prioritize our internal needs, it grew revenue and sharply improved earnings, achieving 5 billion RMB adjusted operating profit.
In recent months, we're seeing a better pricing environment, especially for CPU, which along with robust AI demand and overseas expansion, allowing Tencent Cloud to grow revenue at a faster rate. Moving through the year, we have ordered a substantially higher volume of compute, which should also facilitate revenue growth.
Overall, we think Tencent Cloud is becoming another example of how Tencent competes on our own terms and pace and how our incubation investment cycle works. We view the initial losses in Tencent Cloud as a fixed sum of cash investment necessary to incubate a successful new business, but ultimately generating good economic returns and we view the initial investment in new AI products in the same way. With that, let me pass to James.
James Mitchell
Thank you, Martin. For the fourth quarter of 2025, our total revenue was up 13% year on year. VAS represented 47% of our revenue within which the social network sub segment was 16%, domestic games 20% and international games 11%. Marketing services was 21% and fintech and business services 31%.
Our gross profit was up 19% year on year to 108 billion renminbi. VAS gross profit increased 21%, marketing services increased 22% and fintech and business services increased 17%. Turning to business segments, value added service revenue was 90 billion renminbi, up 14% year on year.
Our social network revenue grew 3% year on year to 31 billion renminbi driven by increased revenue from video accounts, live streaming and from music subscriptions. Music subscription revenue increased 13% year on year on ARPU and subscriber growth.
Long form video subscription revenue increased 1% year on year as video subscribers grew slightly year on year, benefiting from the drama series Love's Ambition, the variety show Natural High Season 3 and the animated series Renegade Immortal. Each of these ranked first by video views in their respective genres across all video platforms in China.
For the quarter, domestic games revenue grew 15% year on year, primarily driven by Delta Force, the Valorant franchise and Wuthering Waves. International games revenue increased 32% year on year, primarily driven by Supercell's titles, PUBG Mobile and Wuthering Waves.
Moving to communications and social networks, we strengthened Weixin's commerce experience by upgrading features for users and tools for merchants in the mini shops. The upgraded e-commerce Gateway page allows users to check their shopping carts, see what friends are recommending and receive notifications from their favorite shops and generated substantial GMV during the quarter.
Through the new Likes for Discounts feature users can discover items liked by friends and receive and share discounts via the ecommerce gateway page, chat and moments. For mini programs, total user time spent increased over 20% year on year driven by workplace productivity tools, mini games and novels.
We added Tencent Code Buddy to our developer toolkit, enabling developers to create mini programs using natural language input. And we provided developers of AI native mini programs with free compute resources.
For domestic games, Delta Force sustained among the top three games in the industry in the quarter. In February 2026, the game surpassed 50 million peak daily active users and achieved lifetime high monthly gross receipts. Delta Force leverages AI coding for development efficiency and deploys AI powered companions to enhance user engagement.
Honor of Kings PC increased its gross receipts more than 30% year on year and achieved record high average DAU in the quarter, benefiting from Flowers meets magic, Miss Bloom skins limited time loads and E sports events.
Valorant Mobile was the most successful new mobile game industry wide by gross receipts in 2025, bringing PC quality shooting experience and a distinctive art style that appeals to younger players. The game achieved lifetime high gross receipts in February as we released outfits to integrate traditional Chinese aesthetics with contemporary design.
In January, we launched CrossFire Future, a multi platform FPS game built on Unreal Engine, which has attracted several million DAUs. Among our international games, Clash Royale ranks third among mobile games industry wide by DAU in the fourth quarter.
Its average DAU and gross receipts more than tripled year on year, reaching lifetime highs. The game launched 10th anniversary events in March, including a limited time PvP mode with random modifiers powering up players cards, providing a more dynamic competitive experience.
Wuthering Waves won The Players Voice award at the Game Awards ceremony in 2025. In the fourth quarter, the game posted rapid year on year growth in gross receipts and DAU driven by a new storyline, urban ruins, maps and new characters.
Warframe launched a major update, The Duviri Paradox, featuring a new storyline, 2 new game modes and new Warframe. Its average DAU and gross receipts reached lifetime highs in December 2025.
For marketing Services, revenue increased 17% year on year to 41 billion RMB. We experienced rapid growth from the Internet services and local services categories, partially offset by slower growth from the ecommerce category due to platforms temporarily shifting budget from marketing to subsidies and also from the financial services category due to the impact of policy changes affecting online lending during the quarter.
Growth drivers included improved ad targeting, expanding our closed loop marketing services and tailoring ad formats for specific advertiser use cases such as ads that are playable previews of the mini games being advertised.
Entering 2026, we've deepened collaboration with e-commerce platforms facilitating their merchants advertising within Tencent. We've increased the inventory for rewarded video ads and video accounts which have contributed to faster year on year marketing services revenue growth in the first quarter quarter to date versus in the fourth quarter of last year.
At a product level, video accounts total time spent increased due to upgrades to the content recommendation algorithm, enabling faster growth in AD impressions. While our ad load remained lower than peers, better conversion rates contributed to more marketing spending.
For mini shops, merchants for mini programs, consumers engaging more with mini games and mini dramas attracted more marketing spend from the mini game and mini drama studios. And Weixin search overall query volume grew at a rapid rate due to AI enhancements to search results driving growth in commercial query volume, while search pricing also increased.
On fintech and business services, segment revenue was 61 billion RMB, up 8%. We grew Fintech services revenue by a single digit percentage year on year and Fintech gross profit at a higher rate driven by wealth management and commercial payment services.
Commercial payment volumes sustained positive year on year growth, supported by a higher number of transactions and a narrow decline in value per transaction. For wealth management, which is the second biggest contributor to Fintech revenue, average assets per user and number of users each increased year on year.
Turning to business services, revenue in the fourth quarter grew 22% year on year, driven by higher cloud services revenue and increased technology service fees generated from higher mini shops e-commerce transaction volumes.
Cloud services revenue accelerated its year on year growth rate due to increased demand and a better pricing environment amid tight supply of memory and CPU industry wide. Revenue from our cloud media services grew notably as short video platforms and AI video generation services are increasingly using media processing solutions for streaming video and audio from processing in the cloud to playback on device reflecting our industry leading streaming quality and our competitive pricing. And now I'll pass to John.
John Lo
Thank you, James. Hello everyone. For quarter four 2025, total revenue was 194.4 billion renminbi, up 13% year on year. Gross profit was 108.3 billion renminbi, up 19% year on year.
Other gains were 1.3 billion renminbi, compared with other gains of 2.5 billion renminbi in the same period last year due to lower subsidies and tax rebates. Operating profit was 60.3 billion renminbi, up 17% year on year.
Interest income was 4.8 billion RMB, up 22% year on year driven in part by growth in cash reserves. Finance costs were 3.6 billion RMB compared with 2.5 billion RMB in the same quarter last year, primarily due to FX loss this quarter versus FX gain in the same quarter last year.
Share profit of associates and joint ventures was 6.8 billion RMB compared with 9.3 billion RMB in the same quarter last year. On a Non IFRS basis share profit was 9.1 billion renminbi, up from 7.7 billion renminbi in the same quarter last year with the increase from improved performance of certain domestic associates due to operational efficiencies and business growth.
Income tax expense increased by 7% year on year to 12.5 billion RMB. On a Non IFRS basis, diluted EPS was 6.966 RMB, up 18% year on year, outpacing non IFRS net profit growth due to reduced share count from share buybacks.
On Q4 non IFRS financial figures operating profit was 69.5 billion RMB, up 17% year on year. Net profit attributable to equity holders was 64.7 billion RMB, up 17% year on year.
Moving on to gross margin for the fourth quarter, overall gross margin was 56%, up three percentage points year on year. By segment, VAS gross margin was 60%, up four percentage points year on year, primarily driven by greater contribution from internally developed high margin games.
Marketing services gross margin was 60%, up two percentage points year on year as AI powered marketing services drove strong growth in high margin revenue streams, particularly video accounts and Weixin search.
Fintech and business services gross margin was 51%, up four percentage points year on year, benefiting from growing scale of cloud services and improved revenue mix in fintech services alongside enhanced cost efficiency.
On quarter 4 operating expenses, selling and marketing expenses were 13 billion RMB, up 26% year on year, reflecting increased promotional efforts to support the growth of our AI native applications and games.
R&D expenses rose by 20% year on year to 23.8 billion renminbi primarily due to higher staff costs and increased depreciation expenses driven by AI investments. G&A excluding R&D expenses increased by 8% year on year to 12.5 billion renminbi due to higher staff costs.
At quarter end, we had approximately 116,000 employees, up 5% year on year or 1% Q on Q primarily reflecting headcount additions to games and our technology platforms including AI related roles.
For fourth quarter, Non IFRS operating margin was 36%, up one percentage point year on year. For fourth quarter, operating CapEx was 16.9 billion renminbi, increasing 41% quarter on quarter as we accelerated investment in server infrastructure.
Year on year operating CapEx decreased by 51% reflecting concentrated CapEx spending in fourth quarter 2024, leading to a high base effect. Non operating CapEx was 2.7 billion RMB, up 60% year on year due to higher facility related investments.
Free cash flow was 34 billion RMB increasing over six times year on year reflecting stronger operating cash flow generation this quarter as well as lower CapEx spending versus Q4 2024. As I mentioned earlier on a Q on Q basis, free cash flow decreased by 42% due to seasonally lower game gross receipts and seasonal settlement of certain accounts payable.
Net cash position was 107.1 billion RMB, up 5% quarter on quarter or 4.7 billion RMB, mainly driven by free cash flow generation, partially offset by share repurchases of 19.6 billion RMB and net cash outflows of 6.9 billion RMB, primarily relating to investing in other corporations.
For the full year of 2025, we repurchased 153 million shares, with the total consideration of 80 billion Hong Kong dollars. The weighted average number of shares for calculating 2025 diluted EPS decreased by 4% year on year.
Given we see high return opportunities from investing in AI, we will likely buy back lower value of our shares versus 2025 to fund investment in AI, while increasing our dividends. Subject to the shareholders approval at the upcoming AGM, we are proposing an annual dividend of 5.3 Hong Kong dollars per share, reflecting 18% year on year increase. This will be payable to shareholders on the 1st of June, 2026. Thank you.
Operator
Thank you, John. We shall now open the floor for questions. If you are dialing by phone, please press 5 to raise a question, then press 6 to unmute yourself. If you are accessing from the Tencent Meeting or VooV Meeting application, please click the Raise Hand button at the bottom. We will take one question from each analyst in the interest of time. So the first question comes from Alicia Yap from Citigroup.
Alicia Yap
Hi, good evening management. Thanks for taking my questions. I have a question on the AI front versus the margin. In our prepared remark, we expect to increase profit from our existing business to more than cover incremental AI investment. Understand we need to look at this AI long term investment separately.
But as OpEx continue to increase into this year, how should we think about the profit margin or the gap between revenue and profit growth into 2026? And my second question is also on AI, how we strategically prioritize given the ongoing constraint in GPU and AI talent as we previously emphasized prioritizing internal AI deployment, but given the recent market development has management view shifted how we prioritize allocating resources or KPI that we monitor? Is that development of foundation language model or user engagement or token growth in B2B solutions. Thank you.
James Mitchell
Alicia, why don't I start and then you know Martin may complement. So, yeah, I think it's implicit in our opening remarks that, you know, it is possible that our revenue would grow faster than our profit in 2026 due to the stepped up investment in new AI products.
And you know, if that's what eventuates, we're very comfortable with that outcome because we can see that these new AI products, you know, represents an opportunity for us to expand our footprint, you know, deliver new value to users. And we can also see, you know from the user enthusiasm around some of these products that you know, that there's a very good opportunity for product market fit.
In terms of your second question around resource constraints on, you know, talent and GPUs, then as far as talent is concerned, you know, we've already been staffing up quite aggressively some very excellent quality talent from, you know, the world and from China.
The team and it will continue to make selective hires but we actually feel we have, you know, really a state-of-the-art AI talent team already in place. And, you know, we've been able to put it in place not only through compensation as an incentive, but also through creating the right culture for the team, through allocating, you know, the roles of the team versus each other and the role of the team within the rest of Tencent appropriately.
Through, you know, the best leaders of the team and in turn, attracting the best joiners to the team in terms of provisioning the team with ample compute and in terms of being able to offer the team, you know, use cases for the AI products they create that are somewhat differentiated and unique to Tencent.
So that's on, you know, talent where, you know, I think that we were facing a situation of scarcity and you know what, we're now, you know, much more comfortable with the setup. So we'll continue to recruit selectively.
In terms of GPU constraints then we've been quite actively provisioning more compute and that will be coming on stream progressively and increasingly quickly through this year, especially the second-half of the year. And you know that additional compute comes from leasing capacity, it comes from us purchasing higher end imported GPUs which are now becoming available again and it comes from us purchasing the increasing quantity of domestically China designed GPUs.
And then in terms of utilizing the compute for different use cases, you know, the priority right now is, you know, for Hunyuan and new AI products more generally, you know, the core products are inherently distributed in nature and, and they, they can themselves source compute, you know, from local devices, from, you know, multiple clouds, you know, from Tencent cloud.
But, but that's, that's sort of somewhat agnostic in terms of the sourcing of compute. So we are, you know, focusing our compute on Hunyuan as the core foundation model and then on the new AI products. Thank you, James.
Operator
Reminder that each analyst please only ask one question. We can take your follow up later if we have more time. Next question comes from Robin Zhu from Bernstein.
Robin Zhu
Thank you management for taking my question. I guess if I could get your thoughts on, you know, clearly we're heading into this AI cycle of investment, how should we think about your assessment of ROI and you know the timing of return. And that this, you know, how you prioritize building versus renting and which parts of the AI stack you think are most critical to be best in breed versus, you know, areas where you think eventually these things will be commoditized as AI continues to move forward. Thank you.
Martin Lau
OK, well, you know, I think from our perspective, we have already seen very good ROIs when we apply AI into our existing businesses right now. So if you look at the breakdown of our financials, you know, if you look at, you know, the financials on a combined basis and then sort of we break it out and saying, oh, you know, these are the financials with existing businesses plus the investment into AI for supporting these businesses, right, you know, the growth is actually quite strong and, and if you exclude the investment in new AI products, then then you know the operating leverage is clearly there.
So I think that's level number one right now and then level number two is an investment into new AI products. On that front, I think, you know, we would be seeing new investments first, right? You know, there's not that much of a revenue, especially in the context of China, unlike in the US where you can actually get consumers to pay subscriptions and you can get companies to pay for, you know, coding agents at a very high cost.
In China, those are not sort of available. So I think these will present themselves as investments upfront, but then over time we believe they will be able to generate revenue from these new AI products and they would generate very attractive return for us over time.
We quote Tencent Cloud as an example in which we initially actually have to invest in the business in terms of incurring losses, but over time, right, you know, it actually turns into a proper business and we believe AI will be like that. There will be a timing difference in terms of the investment and then the return for these new AI products.
In terms of building versus renting, I think you know, if we can buy, right, you know, I think you know, given how strong our balance sheet is, we would actually prefer to buy because then we don't necessarily need to pay the additional margin for leasing, but I think given the constraints in the supply chain and all the different regulations, so sometimes we just have to rent. And I think, you know, we would do that if we need to secure compute.
What was the last question, last part of your question? Did I answer all your questions? Oh, the last question was, if we think about the AI stack between kind of, you know, the models, the orchestration layer, the application layer and so on, which parts would you say are most critical for Tencent to be best in breed versus, you know, areas where we think these would be commoditized?
And you know, it's, it's OK just to say at this point of time is actually very dynamic, right. You're in a fast moving environment. I think, you know, it's very difficult for someone to say something will be more important than the others, right? You know, I think, you know, we have the resources, we have the people who have the team to actually invest in all these layers.
And especially the teams are actually very different when you are actually building the foundation model, you have to build the team from scratch once again. And now it's actually sort of as James said, you know, we have a very strong core team and we have a very strong capability that keep attracting top talent.
But then if you start getting into the application layer, right, you know, it's actually, you know, playing into our strength, right? Because then suddenly you don't even need to have that model capability, but you know, it actually plays into our strength in terms of product capability and orchestration capability and ecosystem is actually our strength and all the infrastructure services like security is also something that we have invested for a very long time and the ability to go across devices such as mobile and PC, right, you know it's actually our core strength too, right.
So I think that actually is really moving into our territory of strength so we would actually have to and also invest in all these capabilities and you know the dynamics of the market will play out itself and hopefully will become best of breed in all layers. Thank you.
Operator
We will take the next question from Piyush Mubayi from Goldman Sachs.
Piyush Mubayi
Thank you, Pony, Martin, James, John and Wendy. So I wanted to ask about the AI agentic agent potential with the recent launch of Q Claw, Work Buddy and we saw the skill hub as well. How should we view parallels of let's say Android versus what we are seeing now for OpenAI in this agentic opportunity and Tencent's position within? So you mentioned about Tencent cloud in that opportunity and how do we plan to differentiate other parts of the stack for example models? Thank you.
Martin Lau
I think claw is actually a very exciting concept, right, you know, and it actually sort of presents a decentralized model or decentralized regime for, you know, how AI works in this world. So there's some parallel to sort of how the Internet evolves, right? You know, in the very beginning when Internet first appears, right, you know, there seems to be sort of, you know, there is one entry point, which is the browser and then you know, there's sort of one distribution point, which is the search engine.
But over time, you know, you know there are different services evolve right, you know and then when mobile Internet comes suddenly you see some a multitude of applications coming up right you know, and within the applications there are applications which are completely mobile, native mobile centric mobile first and then there are also mobile applications that were the PC Internet champions who actually migrate onto the mobile Internet world, right, you know, and I think you know, this is you know how we felt the claw is right, you know, for some time, right the AI that seems to be a sort of new everybody is trying to fight to become the AI AGI monopoly.
You know there seems to be a point in which like people said, oh, if there's one model which is AGI, then you know it would rule over everybody, right. But the reality is not right. You know, you have multiple models becoming very strong and you know, they specialize in different kinds of activities, right? One in chatbot, the other one in coding and the other one in multimodal.
And you also have open source, which are pretty good. And you have a lot of other models, which is sort of, you know, fast followers too. And then there was a time in which in the 2C world there seems to be the chat bot being sort of, you know, the single entry point.
But now with Claw, you can see, you know, it opens up a completely decentralized regime where you know, many companies can have their own claw and the claw can be using all kinds of different models, right? You know, and it's supported by the infrastructure of cloud and each one of the claw has to figure out its unique value proposition, right to win the heart of the users.
And the claws also make use of not just the cloud, not just its unique model, it also sort of also make use of the tools available to them on devices and utilizes the file system right? So it becomes a much more decentralized world and we felt you know, we you know there's a lot of opportunities for us right you know, in terms of building products to cater to people's needs.
So that's why there's Q Claw. There's also Work Buddy and in the future, I think a lot of existing apps will try to come up with their own claws, right, you know, and their own agentic capabilities and different models would try to compete to win the hearts of these claws, so it becomes a much more exciting world and decentralized world for everybody to have some participation.
And you know, we just need to, as I said, right, build expertise in the different layers, you know, in the model layer, in the product layer, in the infrastructure layer. And you know, each layer would have to sort of have their own specific value proposition to win its usage. Thank you, Martin.
Operator
We will take the next question from Eddie Leung from Bank of America.
Eddie Leung
Thank you so much management for taking my question. I actually have a follow up just on the question towards the agentic era, how would management evaluate Tencent's value proposition in this new agentic era and since we are putting Hunyuan alongside with the other LLMs, you know? Kind of towards the prosumers and consumers, how do we potentially prevent from the other LLMs diluting our own foundation models value in the longer term? Thank you.
James Mitchell
I think that in terms of, you know, Tencent's unique value proposition or what we can bring to users in the claw era, you know, that there's a few sort of inherent, you know, attributes that we possess, which we think are very suitable for, you know, the agents, the deployment of claws and you know, Martin's touched on them.
But, you know, one of those attributes is that we're a company whose, you know, capabilities span across, you know, PC, mobile, cloud. We're a company whose capabilities span across applications and the World Wide Web, you know, just as the agentic claws, you know span the devices and span the sort of domain.
We're a company that, you know, operates a number of centralized apps, but also hosts some extremely, you know, decentralized yet vibrant ecosystems, most notably the mini program ecosystem. And so, you know, one framework you could think about it is that in prior years, the arrival of the mobile Internet really sort of turbocharged the, you know, application experience vis-à-vis the more centralized app experience versus the more decentralized World Wide Web experience and you know, now with these agentic capabilities and claws, then there's an opportunity for, you know, decentralized experiences such as mini programs to be turbocharged and themselves develop, you know, far more powerful capabilities than they enjoyed in the past.
So, you know, that's, you know, why we think there's inherently a natural fit between our capabilities and our interests and the deployment of agentic agents or claws and you know. That's why, you know, we're seeing one reason why we're seeing consumers and prosumers enthusiastically adopting our own agent and claw services.
In terms of the part of your question about preventing other large language models diluting our models value, I may not sort of understand the premise correctly, but I don't see that happening. You know the if you use these claws then you know you go into the interface and you have a choice.
Do you want to use, you know, Model A, which is, you know, very high performance and high price per token or you know, model Z that's medium performance and very low price per token, all models, you know, B through Y in the middle and you know, that's part of the appeal of the claws and you know Hunyuan is, you know, one of those models that is available and you know we believe with the capabilities of the Hunyuan team now in place that going forward Hunyuan will get better faster and therefore consumers will naturally increasingly opt to use Hunyuan.
But I don't think it will be a monopoly situation. You know that the claws that are successful will be claws that continue to allow consumers and prosumers to, you know, make their own choice along the price performance curve. And you know different models will sit at different places on the price performance curve. And you know we want to be one of those, but we don't intend to be the only one of those. Thank you.
Operator
We will take the next question from Alicia Yap from Citigroup.
Alicia Yap
Hi, good evening. Thanks for taking my questions. I have a questions related to the physical AI so considering the proliferation of the agentic focus AI agent across enterprises, especially the traditional industry. Do you believe this will accelerate the demand for the usage of the world models like the 3D models that you have and also what is management's assessments of Tencent's capability and also the competitive strength in the future physical AI era? Thank you.
James Mitchell
Alicia, I think your point is a reasonable one that you know that there is already a, you know, computer aided design capabilities and you know, one would naturally expect, you know, AI to supplement and eventually supercharge those abilities. And so you know, that's important in industrial design, it's important in architecture, it's actually very important and increasingly important in video games.
And, you know, we believe we can see that we're in a somewhat uniquely good position to provide the data to train the models to in turn supply, you know, those 3D tools because of the breadth and depth of 3D graphical assets within our video games.
But you know it's ultimately a sort of a big niche and you know, it's one that, you know, we are well positioned to address, but I wouldn't say it's, you know, the biggest opportunity ahead of us. You know, there's many larger more immediate opportunities. Thank you.
Operator
We will take the next question from John Choi from Daiwa.
John Choi
Thank you for taking my question. I have a question related to games and, you know, AI disruption. Have you already started to see some headcount and game development costs being impacted? And how do you think AI will impact the quality and also the overall cost side? And how should we expect, you know, Tencent to prepare this would also distribution and publishing be more important going down the road as we see more increased number of games and also if you look, you know, AI lowers the development entry barriers. Are we going to see meaningful increase in the supply of the game studios and in terms of the overall quantity of the games going down the road? Thank you.
James Mitchell
Yeah. Thank you for the question, John, so I don't know if any of you attended the Game Developers Conference last week, but it is the sort of premier event each year for game developers and you know, as you would expect, there was a number of, you know, well attended, you know, presentations about the use of AI within creating games and you know, I think a couple of broad observations.
One is that you know those presentations were pretty exclusively focused on how to use AI to, you know, upgrade content within existing games to accelerate the content creation, improve the content creation within games, but you know that there is not yet the capability to create games, you know, completely from scratch using AI for a number of reasons that we can get into.
And then, you know, the second observation is that many of the, you know, best attended presentations were by our colleagues within Tencent Interactive Entertainment Group. And, you know, they talked about how, you know, AI can be deployed in games, you know, for graphics, AI can be deployed in games for gameplay AI, can be deployed in games for, you know, user companionship and so forth.
And you know, we believe that, you know, we're at the forefront of the industry in this regard and you know the feedback from many of the people, the developers who attended the game developers conference was consistent with that belief.
In terms of the second-half of your question about whether AI will result in a flood of new games and therefore elevate the importance of publishing versus development then, you know, the sad reality of the game industry is that it's perpetually in a, you know, oversupply situation.
You know, every year, as Martin mentioned, there's 200,000 new games on mobile, there's 18,000 new games on Steam. So you know, whether that number goes from 200,000 to 2 million to 2 billion to 2 trillion, you know, has sort of diminishing incremental impact.
You know, the key is really, you know, making and then, you know, extending and rendering Evergreen the best games. And, you know, in order to do that, you need, you know, the best human beings, you know, supplemented by the best technology. And you know, we think that therefore the value balance between development and publishing, you know will remain where it is today and the critical success factors, it will continue to favor the best developers in the industry. Thank you.
Martin Lau
Just to add a couple more points, right number one when you talk about sort of AI disruption for games, right, you know that basically, sort of, you know, imply it's actually bad for the gaming industry. But I think sort of gaming is actually one of the industries that would benefit from AI, right? Because when AI proliferates, I think people would have more free time at their hands. And the demand side would actually increase significantly for the gaming industry, which I think you know is a rare certainty in the face of AI proliferation.
And number two is, you know, the availability of great tools, it would be available to new developers, but it also said it will be available to very organized teams and highly talented developers, you know that are already running big Evergreen games right You know, and I would say sort of, you know when a tool is actually available, it's going to overly benefit the people who have the resources and who have already got all the gamers around the platform and, you know, they can actually better use these tools to, you know, increase the amount of production and make new games even more Evergreen, right.
So I think that's an advantage for players who have Evergreen games and are also extremely fast and agile in embracing technology and finally right you know when there's a multitude of innovations. A lot of times, I think, you know, what we saw in the gaming industry is like, you know an idea comes around and then it's not perfect and it gets sort of iterated and polished overtime and I think the process would actually be speeding up if a lot of these games who have a lot of users look at these innovations and can iterate faster and incorporate these new experiences into their existing games and make games essentially into platforms and I think, you know that's a unique opportunity that we would see over time as well. Thank you.
Operator
We will take the next question from Alex Yao from JP Morgan.
Alex Yao
Thank you, Wendy. Thank you management for the opportunity. I want to follow up on the AI cloud side of the business given very strong demand for AI compute, but on the other hand also price inflation for the server AI servers due to the rising cost of DRAM and HBM. Can you guys help us understand Tencent cloud's pricing power and also philosophy to value capture in such a very dynamic environment? Put it another way. Do you want to fully pass through the cost inflation to your customers or partially subsidize the cost inflation and then get more market share or even potentially you know more than fully pass through the cost inflation to capture more profit? Thank you.
James Mitchell
Thank you for the question, Alex. So, you know, first of all, I'd start by saying that, you know, clearly there is a surge in demand for sort of AI compute, but it's not only for AI compute. You know, when people utilize the agentic tools that we've been discussing that they're using them and they create software and you know that software, you know, then primarily it needs to be executed.
And when it executes, most of it is not executing on GPU, it's executing on CPU. And then it creates, as it executes, it creates, you know, memory demands. So it's not just, you know, GPU DRAM, HBM where we're seeing demand picking up. It's also CPU, it's, you know, regular RAM, it's SSD, it's hard disk drive. It's across the board. There's a pickup in demand.
And you know in terms of how the industry addresses at an industry level responds with pricing then? For years, the industry has suffered because the cloud services providers in China, were operating at very low margins and one of the reasons they operated at very low margins was because you know, if there was a new entrant or if the customers wanted to source infrastructure directly, they were able to telephone the supplier and order the infrastructure that they wanted from the supplier of, you know, CPU or GPU DRAM, you know, that's no longer the case.
You know, now the supply is booked out months, in some cases years in advance. You know, the supply is prioritizing the biggest, most regular customers which are the hyperscalers such as ourselves. And therefore, you know, the customers, the smaller cloud providers, you know, no longer have certainty that they can, you know, source supply and they need to come to the hyperscalers.
And you know, the hyperscalers have been operating at low margins. And so, you know, when the demand picks up, then, you know, we almost sort of as an industry have no choice but to pass through higher prices. And you have seen a number of price increases in China Cloud in the last few months as a result.
In terms of, you know, how we sort of value capture in this dynamic environment, then, you know, one broad principle is, you know, we seek to deliver even more value through enrichment means that at a minimum, if you have compute, you can rent it out bare metal when you get a certain low price, some low margin, you know, preferably you rent it out, you subdivide it and virtualize it into tokens and then you get a higher price and higher margin per unit of compute.
And ideally you bundle it into a platform as a service or software as a service, and then you can get, you know, the best pricing and the best margins. And so that's part of the journey that we've been on and that's part of, you know how Tencent Cloud has moved from very substantial losses 4 years ago to pretty substantial profits last year. And we'll continue on that journey of, you know, moving from bare metal to token to platformization and to software. Thank you.
Operator
We will take the next question from Gary Yu from Morgan Stanley.
Gary Yu
Hi, good evening and thank you for the opportunity. I have one question regarding the comment quite a few times that we mentioned that we are not a first mover or we are even the late comer in AI. In the US, we have also observed that it's becoming very difficult for some of the latecomers to catch up, even for those that have very high resources in terms of compute talents and data, so how does management get comfortable and confident that we won't be following the same path in terms of, you know, lagging behind, not able to catch up around areas on compute model applications. Thank you.
Martin Lau
Yeah, I think that's a very good question. And I think if you are playing just one game, then basically it's hard to sort of, you know, catch up on one game, right. But then you know if you view AI as sort of, you know, a multiple of different games then you know, there are new opportunities, new frontiers that's opened all the time right.
So I think you know, it's already happening right now. If you look at the model right now, in the very beginning, everybody felt, you know, it's the chatbot and then coding comes around and then multimodal come around and you know, and then sort of you know, when everybody felt, oh, that's pretty much it and then suddenly sort of, you know, Claw came around, which basically further decentralized the whole AI landscape.
So you know, in the future we actually felt, oh, you know, there will be just like apps, right? You know there will be a lot of different permutations of how AI will be packaged from model to the product to agent and existing services will be having sort of you know different agentic capabilities. There will be new agentic capability coming around, you know, on mobile, on PC.
So you know, it's very early days in the whole AI development world so that's why, you know, just within a short period of time you can see sort of there are already a lot of proliferations and there'll be more and more coming. So that's why it's actually important to have some fundamental capability, right, you know, and we do have a lot of them.
In terms at the application layer, be it you know Weixin and be it you know our ecosystem of having communication and presence on PC and mobile and a lot of infrastructure capability including security and cloud and payment and, you know, all these elements can be packaged together, you know, in the new era of AI.
So it's not sort of one race, it's actually sort of a world of many, many races. And I think, you know, that would, you know, increasingly manifest itself and as a result, there will be a lot of opportunities for different players to come up and innovate from behind.
So I'm not sort of, you know, very worried about, you know, being late, but I'd be worried about, you know, if we're not innovating fast enough, right? You know which I think you know as we restructure our Hunyuan team and as we started to invigorate all our product teams to start innovating with products and I think, you know, that's actually happening in a very exciting way for us. Thank you.
Operator
We will take the next question from William Packer from BNP Paribas.
William Packer
Hi management. Thanks for taking my question. Press reports suggest that Apple is planning to cut App Store Commission rates by 5% for apps and two to 3% for mini games in March in China, Tencent a potential major beneficiary. To what extent should we expect these cuts to flow through to Tencent's gross margin or would they be shared with other stakeholders such as consumers, gaming partners or perhaps tax revenue? Thank you.
James Mitchell
William, so you know, happily in this case the press reports were based on the sort of objective reality of an Apple, you know, formal announcement and so this is not a, you know, a speculative hypothetical, it's a, you know, actual development that takes effect in the last few days.
And in terms of the flow through then you know there should be a good flow through. You know, when we have game development partners and we're the publisher of those games, which is now quite a small minority of our game revenue, then in the overwhelming majority of cases, the revenue share is calculated based on the gross revenue, not on the revenue net of the Apple take.
And so that, you know, flows through to us, you know, if by taxes you're referring to us paying corporate income tax on, you know, this incremental profit stream. And I suppose that's correct dependent on the extent to which we, you know, reinvest this incremental profit stream into new AI products.
I think that you talked about one part of the Apple announcement, which is the sort of quantitative part that moves from 30% to 25% and 15% to 12%. But, you know, for us, actually, the more important aspect of the Apple announcement looking forward was that you know, Apple stated that it would effectively offer developers in China equivalents with whatever the rate is that, you know, developers elsewhere in the world are paying to App Store and so you know, our view is that with the evolving industry trends it's a sort of matter of time for the tolls the App Store collects to normalize downward in different parts of the world and you know, with this, you know, declaration Apple has stated that, you know, as the take rates move down in different parts of the world, so the take rates will move down in China in synchronicity.
So we believe that this is, you know, a very positive first step but you know, it is, you know a first step on a multi step positive journey. Thank you.
Operator
We will take the last question from Alex Liu from Bank of America.
Alex Liu
Well, thank you for taking my questions. My question is really just on AI chips. So we're seeing a growing number of your tech peers are prioritizing the development of in house chip design capabilities. So I'm just curious where in house chip development fits into Tencent's AI priorities? Thank you.
Martin Lau
Yeah, thanks for your question. I think at this point of time, it's not the most critical thing that we'll be focused on. So if you look at the chip, you know there's, you know a difference between training chip and inference chip, right, you know, and for training chip, it's actually very, very difficult to design and manufacture and you actually want to have access to the most state-of-the-art training chips to the extent possible and in the most flexible way, so that, you know, you can actually sort of keep training for the best model.
And then, you know, if you're talking about inference, right, you know, I think inference it's mostly for cost and I think for cost at this point in time, there's actually a lot of different suppliers in China, which is actually very different from let's say in the training space, right, where there's essentially one or two players who can actually command a very, very high margin, right.
You know, in the inference world, people basically sort of, you know, are earning much lower margin and there are many more solutions and, you know, options. So I think you know the key for us is actually sort of leverage the best training chips to train the best model at this point in time, and there's a lot of value in being focused.
And when it comes to the inference part, right, you know, overtime, I think you know, the market would actually sort of, you know, play out in such a way that I think you know the margin in the inference chips will be actually quite manageable at this point in time.
We're very focused on leveraging the best chip to train our model and our Hunyuan 3.0 is going to be much better than Hunyuan 2.0. And that's actually just the starting point. I think, you know, over time we'll be able to iterate the training of our model faster. And you know, I'm very confident that, you know, if we focus on that, you know, we'll reach SOTA at some point in time. And I think that's actually the most important thing for us.
And the next important for us is actually really unleashing the power of our product development capability and integration and connection capability in order to design the most exciting AI products for users, I think you know when those are done right, then we think about how do we try to reduce the cost of inference? Thank you Martin.
Operator
We are now ending the webinar. Thank you for joining our results today, if you wish to check out our press release and other financial information, please visit the IR section of our company website at www.tencent.com. The replay of this webinar will also be available soon. Thank you and see you next quarter.
Details at Tencent IR
Tips: This summary was generated by an AI language model based on publicly available information, including auto-generated subtitles and publicly released audio from third-party sources. While efforts have been made to ensure accuracy, AI-generated content may contain errors or omissions. Please refer to the official materials released by the respective listed companies for the most accurate and up-to-date information. This content is for informational purposes only and does not constitute investment advice or a recommendation to buy or sell any investment products. The views and summaries presented do not reflect the official position of moomoo. Moomoo makes no warranties, express or implied, regarding the accuracy, timeliness, or completeness of this content and shall not be held liable for any losses arising from its use. The copyright of any original conference call content remains with the respective listed companies.
Disclaimer: Moomoo Technologies Inc. is providing this content for information and educational use only.Read more
Comments (51)
to post a comment
35
3