Tencent Holdings Limited (700) Earnings Call Transcript & Summary

March 18, 2026

Hong Kong Stock Exchange HK Communication Services Interactive Media and Services earnings 79 min

Earnings Call Speaker Segments

Wendy Huang

executive
#1

Wendy Huang (IR): Hello, and good evening, everyone. Thanks for joining us today. This is Tencent Holdings Limited's webinar covering the 2025 fourth quarter results announcement, and I'm Wendy Huang from the Tencent IR team. Right now, everyone is in a listen-only mode. Once management finishes its presentation, we'll move into a question and answer session. Those of you who joined by phone and would like to pose a question should press 5 on your telephone to raise your hand. If you're on the Tencent Meeting or VooV Meeting application, just click the raise hand button in the bottom left corner. Please note that today's webinar is being recorded. Before the presentation begins, we'd like to remind you that it contains forward-looking statements, which are subject to a number of risks and uncertainties and might not be realized in the future for various reasons. Information regarding general market conditions comes from a variety of sources outside of Tencent. This presentation also includes some unaudited non-IFRS financial measures, which should be considered in addition to, but not as a substitute for, measures of the group's financial performance prepared in accordance with IFRS. For a detailed discussion of risk factors and non-IFRS measures, please refer to our disclosure documents on the IR section of our websites.

Wendy Huang

executive
#2

At this point, I'd like to introduce the management team joining us on the webinar this evening. Pony Ma, our Chairman and CEO, will begin with a brief overview. A strategy review will then be given by President Martin Lau. Chief Strategy Officer James Mitchell will walk us through a business review, and Chief Financial Officer John Lo will wrap up the financial discussion ahead of opening the floor to questions. I'll now hand things over to Pony.

Pony Ma

executive
#3

Thank you, Wendy, and good evening, everyone — thanks for being with us. During 2025, we delivered high-quality growth by way of our evergreen products and services, with growing support from the application of AI capabilities. We broadened that evergreen game portfolio, thanks to the breakout success of Delta Force, and we strengthened existing evergreen titles including Honor of Kings and Peacekeeper Elite. AI makes a meaningful contribution across game content development, user engagement, and marketing efficiency. Total time spent on Video Accounts rose over 20%, driven by upgraded recommendation algorithms and a richer content ecosystem. Revenue from our marketing services gross outperformed the industry, helped by our upgraded ad tech model and the newly introduced automatic campaign solution, AI Marketing Plus. On fintech, we kept up healthy revenue growth by deepening cooperation with licensed financial institutions while upholding prudent risk management. In cloud, we reached profit at scale, reflecting increased enterprise demand for our industry-leading PaaS and SaaS products along with supply chain optimization. Outside China, our international games business exceeded $10 billion in annual revenue for the first time, powered by sustained growth of evergreen games and the rapid expansion of content-driven games. Cloud revenue accelerated in international markets as we broaden partnerships with key clients and drove higher adoption of our flagship cloud products.

Pony Ma

executive
#4

The strong operating leverage and cash flows our core businesses produce let us increase our investment in AI. Over the course of the year, we strengthened the team with top-tier AI talent and established processes to improve foundation model intelligence in a systematic manner. We started rolling out new AI capabilities across services such as Yuanbao and Weixin as well as cloud-type products. Turning to our financial figures for the fourth quarter, total revenue came in at RMB194 billion, a rise of 13% year on year. Gross profit reached RMB108 billion, an increase of 19% year on year. Non-IFRS operating profit stood at RMB70 billion, up 17% year on year. Non-IFRS net profit attributable to equity holders was RMB65 billion, also up 17% year on year. I will now turn things over to Martin.

Martin Lau

analyst
#5

Thanks, Pony, and good evening and good morning to everyone joining the call. In this next section, I'll walk you through how we view AI as a transformational force, beginning with how resilient our existing businesses are in the context of AI, then moving on to how we're deploying AI within our existing businesses to strengthen them, after that discussing brand-new products and opportunities that AI makes possible, and how we're investing in order to capture them. I'd like to talk about how our key franchises are very resilient in the age of AI. As we know, AI will affect every part of the technology industry, but some products and services are inherently more resilient than others. We believe some of the characteristics of resilience will include: network effects arising from consumer-to-consumer, consumer-to-content-creator, and consumer-to-business interactions in descending order of strength, that's number one; number two, deep supply chain integration linking the world of bits with the world of atoms; number three, stringent regulatory and licensing requirements; number four, scarce or unique resources including physical and intellectual properties; number five, take rates that are low compared to value provided or cost of switching; and number six, private data that is closed and interactive in nature.

Martin Lau

analyst
#6

So, applying those criteria, when we look across our major existing businesses, we conclude — and usage trends back this up — that each of them has a high degree of inherent resistance. Specifically, with our communication services, including Weixin, QQ, and Tencent Meeting, people use them to connect and interact with others — largely their families, friends, colleagues, and business partners. We believe this need for human interaction, along with the network effects and the closed nature of the data that arises from these interactions, has made communication services extremely sticky when facing competing non-AI services in the past, and they will keep being resilient versus AI-based services going forward. Then moving on to our games, those are very resilient too, since our multiplayer games, especially PvP games, also enjoy network effects. Similar to sports, they are team-based in nature, and players play with and against other players. Just as people would rather participate themselves or watch the teams they support compete in sports instead of watching AI sports, game players keep enjoying the interaction with other humans that our games provide. Our games also cultivate strong IPs. While AI will enable more games to be made faster, the game industry is already in a position of excess supply, with 200,000 new games on mobile and 18,000 new games released on Steam every year.

Martin Lau

analyst
#7

What holds us back is that any new game has to be higher quality and more innovative than the strongest games already out there, and that demands human creativity layered on top of the most advanced technology. Gaming stands to gain naturally from AI spreading, including in cases where people find themselves with extra time. Our fintech services are resilient too, since they rest on licenses that are hard to obtain and keep, which are finite by design and also define the limits on how innovation can come into an industry. We've also put in decades of work constructing a payment network of rails that are hard to replicate, reaching partner banks and merchants and linking them to over a billion consumers, which generates its own network effects. Our mobile payment take rates already rank among the lowest anywhere in the world, and we think that makes competing with us on price deeply uneconomical. Next, we aim to show that we lead in reinforcing our core businesses with AI. When generative AI first appeared, we made it a priority to use AI to bolster our core businesses, reasoning that if we can make them stronger, we'll be better positioned to put money into new products that AI makes possible. We think that across each of our core businesses, we now stand at the forefront of our respective industries in China and frequently worldwide when it comes to using AI, with encouraging early results shown by user engagement and revenue trends.

Martin Lau

analyst
#8

When it comes to games, we're putting generative AI to work so in-game content can be produced faster, which lets us turn out more content across our big titles. Generative AI also helps us bring in new users and keep the ones we have, through things like targeted advertising and personalized daily highlight reels. We're making the core gameplay richer with AI-driven features — virtual teammates in PvP titles and lifelike non-player characters in PvE titles, for example. These efforts are part of why Tencent's games keep becoming more evergreen, and why our 22% revenue growth in 2025 beat the 7% growth recorded by the global games industry. On the marketing services side, we scaled up our advertising foundation model so that more relevant ads reach more precisely targeted users, which lifts ad conversions for advertisers while also delivering better user experiences. We offer generative AI-powered ad creative solutions that let advertisers produce more ads, more efficiently, and make them more relevant to a smaller set of users. We launched our automated ad campaign solution, Aim Plus, which lets advertisers automate targeting, bidding, and placement, improving the return on their marketing investments and leading them to allocate more budget to us. These efforts are a substantial contributor to the 19% growth in Tencent's marketing services revenue in 2025, which surpassed the 14% growth of the overall China ad industry.

Martin Lau

analyst
#9

On Video Accounts, we're rolling out a longer-sequence AI model that picks up more of a user's signals, which sharpens content recommendations and is driving growth in users, engagement, and content distribution. In 2025, total time spent on Video Accounts rose by over 20%, and by DAU, Video Accounts now ranks as China's second-largest short video service. On the digital content side, we apply AI to production, which lifts workflow efficiency and delivers visually striking special effects. AI also assists distribution by making recommendations across music, videos, and literature more intelligent. In enterprise software, we're applying AI to offer capabilities like AI agents that take notes and summarize concurrent meetings for users, plus AI agents that produce intelligent summaries of customer service history for merchants. By usage and revenue, our enterprise software offerings, WeCom and Tencent Meeting, lead their categories in China. In fintech, we employ lightweight AI models to strengthen credit scoring and support fraud detection, which helps us keep non-performing loan rates better than the industry's. With our core businesses now gaining operationally and financially from integrating AI, we believe we're strongly positioned to make developing new AI products one of our priorities.

Martin Lau

analyst
#10

When it comes to the foundation model layer, we’re seeing significant opportunities from pairing a strong foundation model with configuration for core user cases like chatbot, coding, multimodal, and agentic applications. Even though we weren’t the first mover in large language models, we’ve already revamped our team, improved our data quality, and rebuilt our AI infrastructure for pre-training and reinforcement learning, and now we’re iterating more intelligent models at a faster pace. Hunyuan 3.0 is in internal testing and currently represents a bigger step up in capabilities versus Hunyuan 2.0 than 2.0 was versus 1.0. When it comes to multimodal capabilities, our 3D, text-to-image, and world models are early category leaders, and they’ll increasingly benefit from leveraging our proprietary data and abundant use cases. Some observers in China tech are single-mindedly focused on AI chatbots as the only means for bringing AI to users. We think this mindset is overly simplistic, because AI can help people in a multitude of ways beyond powering an information and advice app. We believe that AI chatbot applications are largely competing with search applications rather than with every other application. As for Yuanbao, our own AI chatbot app, we focus on finding product-market fit and use cases which belong in chatbot AI app.

Martin Lau

analyst
#11

We're moving fast on Yuanbao, iterating to improve how people use it — better search integration, stronger speech recognition, simpler access to multimodal features, and exploration into group chat. We think all of that will drive more usage and keep users coming back to the app. Over the next few months, with Hunyuan 3.0 rolling out in Yuanbao, we expect the core user experience to take another step up. Beyond that, we've brought AI into a variety of existing experiences across Weixin — things like content consumption, information retrieval, and merchandise recommendation and customer service. We're building AI agents that act autonomously for users, interacting with Weixin features, mini programs in particular. The enthusiasm around clawbots shows people see that AI can unlock computer-use abilities to make daily life better, but it also shows the risks of letting AI run unsupervised. Our aim is for AI agents in Weixin to deliver AI productivity that benefits the general public and early adopters alike, and that will lift ecosystem activity and naturally bring in revenue. Right now AI agents are powered by many different foundation models, and we expect that at the application level, users will keep having access to a range of models. Still, better Hunyuan performance will let us offer new agentic capabilities that are unique to Weixin.

Martin Lau

analyst
#12

Going forward, the Weixin and Hunyuan teams are going to collaborate more and more closely. On the subject of clawbots, we've rolled out several AI tools aimed at boosting productivity, such as WorkBuddy, QClaw, and Tencent Cloud Lighthouse, and through our Skill Hub we offer downloadable skills so people can easily start using these tools. By means of autonomous workflows and continuous task execution, clawbots are moving AI from thinking to doing. Users operate this new generation of AI tools via command-line interfaces inside the communication apps they already have, which in China generally means Weixin and QQ, since it's efficient for users to engage with digital agents in a place and format where they already talk with human contacts. The new AI products I just described demand substantial and growing investment, and we believe that over the long run this will produce significant returns for us. Our spending on Hunyuan and Yuanbao, our two biggest new AI products, came to RMB7 billion in the fourth quarter of 2025 and RMB18 billion for the full year. Those numbers cover only Hunyuan and Yuanbao, and they leave out AI initiatives that support our existing products and services, as well as costs from supplying GPUs to external customers through Tencent Cloud. In 2026, we expect to more than double these investments in Hunyuan, Yuanbao, and other new AI products, and we intend to fund that from rising earnings in our core businesses.

Martin Lau

analyst
#13

During this period of transformation, we're separating out what we spend on new AI products, since we see these strategic outlays as conceptually akin to putting money into an affiliate or to capex. They're upfront commitments needed to lay the groundwork that unlocks fresh value, rather than ongoing operating expenses. So in our view, the effect of these investments ought to be looked at apart from the profits our existing businesses bring in. In time, we have confidence that monetization will trail usage when it comes to these new AI products. Finally, I'd like to walk through a case study on Tencent Cloud — the most recent illustration of how we turn our services into market leaders that deliver economic returns over time, following in the footsteps of games, payments, and long-form video. We anticipate the same will hold for our new AI products. Tencent Cloud came into cloud services relatively late. Even so, we stuck with a patient, long-term investment approach, on the belief that it had scale right from the beginning, given that Tencent itself is the biggest single end user across a range of technology infrastructure in China, and that it could offer differentiated services stemming from Tencent's distinctive insights, ecosystem, and capabilities. As an example, we believe we were the first cloud service provider in China to fully appreciate the stepped-up capabilities of AMD's recent generations of CPUs, and that we became AMD's largest partner in the country, and that our cloud video streaming service leads the industry when it comes to streaming quality.

Martin Lau

analyst
#14

Following a stretch in which Tencent Cloud put revenue growth first — a stance we now regard as somewhat misguided, influenced by what other industry players were doing — we moved aggressively in 2022 to restructure Tencent Cloud around high-quality services instead of pursuing high-revenue, low-value-added work like reselling and customizing projects. That pivot cost us several quarters of revenue growth, yet it let Tencent Cloud reach operating profit break-even in 2024, compared with significant losses in prior years. Through 2025, Tencent Cloud kept facing revenue headwinds because GPU availability for external customers stayed limited while we prioritized our internal needs, but even so it grew revenue and sharply improved earnings, delivering RMB5 billion in adjusted operating profit. Lately we are seeing a better pricing environment, particularly for memory and CPU, and together with robust AI demand and overseas expansion, that is letting Tencent Cloud grow revenue at a faster rate. As we move through the year, we have ordered a substantially higher volume of compute, which should also help facilitate revenue growth. On the whole, we see Tencent Cloud as becoming another example of how Tencent competes on our own terms and pace, and of how our incubation investment cycle works. We look at the initial losses in Tencent Cloud as a fixed sum of cash investment required to incubate a successful new business, one that ultimately generates good economic returns.

Martin Lau

analyst
#15

Our perspective on the upfront spending for the new AI offerings — those carrying the internal codename MAP — follows the same logic. On that note, I'll turn things over to James.

James Mitchell

executive
#16

Thank you, Martin. In Q4 2025, total revenue rose 13% compared with a year earlier. VAS accounted for 47% of revenue, with the social network subsegment at 16%, domestic games at 20%, and international games at 11%. Marketing services made up 21%, while fintech and business services contributed 31%. Gross profit climbed 19% year on year to RMB108 billion. Gross profit for VAS rose 21%, marketing services was up 22%, and fintech and business services gained 17%. Now to the business segments: value-added service revenue reached RMB90 billion, 14% higher year on year. Social network revenue increased 3% year on year to RMB31 billion, helped by higher revenue from Video Accounts live streaming and music subscriptions. Music subscription revenue rose 13% year on year on ARPU and subscriber growth. Long-form video subscription revenue was up 1% year on year as video subscribers grew slightly year on year, aided by the drama series Love's Ambition, the variety show Natural High Season 3, and the animated series Renegade Immortal. For the quarter, each of these ranked first by video views in its respective genre across all video platforms in China. Domestic games revenue rose 15% year on year, mainly driven by Delta Force, the VALORANT franchise, and Wuthering Waves.

James Mitchell

executive
#17

Revenue from international games rose 32% compared with the same period a year earlier, with the main drivers being titles from Supercell, plus PUBG Mobile and Wuthering Waves. Turning to communications and social networks, we made Weixin's commerce experience stronger by improving the features available to users and the tools available to merchants within Mini Shops. The revamped e-commerce gateway page lets users view their shopping carts, check out what their friends are recommending, and get notifications from shops they favor, and it produced substantial GMV over the quarter. With the new Likes for Discounts feature, users are able to find products their friends have liked and can receive and pass along discounts through the e-commerce gateway page, chats, and Moments. On Mini Programs, overall time spent by users grew by more than 20% compared with a year earlier, powered by workplace productivity tools, mini games, and novels. We brought Tencent CodeBuddy into our developer toolkit, which lets developers build mini programs through natural language input, and for developers of AI-native mini programs we offered free compute resources. On the domestic games side, Delta Force held a place among the industry's top three games during the quarter. In February 2026, the game went past 50 million peak daily active users and hit a lifetime-high monthly gross receipts figure. Delta Force makes use of AI coding to boost development efficiency and fields AI-powered companions to strengthen user engagement.

James Mitchell

executive
#18

On the PC side, VALORANT's gross receipts rose by over 30% compared with a year earlier, and the quarter set a record for average DAU, with help from the Flowers Meet Magic: Mystbloom skins, limited-time modes, and esports events. VALORANT Mobile stood out as 2025's top new mobile game across the industry by gross receipts, delivering a PC-quality shooting experience plus a distinctive art style that draws in younger players. Lifetime-high gross receipts for the game came in February, when we rolled out outfits blending traditional Chinese aesthetics with contemporary design. January saw the launch of Assault Fire: Future, a multiplatform FPS title built on Unreal Engine, which has drawn several million DAUs. Turning to our international games, Clash Royale was the third largest mobile game industry-wide by DAU in the fourth quarter. Both its average DAU and gross receipts more than tripled versus a year earlier, hitting lifetime highs. March brought 10th anniversary events for the game, among them a limited-time PvP mode with random modifiers that power up players' cards, offering a more dynamic competitive experience. At the 2025 Game Awards ceremony, Wuthering Waves took the Players' Voice award. During the fourth quarter, the game saw rapid year-on-year gains in gross receipts and DAU, fueled by a new storyline, urban ruins maps, and new characters.

James Mitchell

executive
#19

Warframe released a big update called The Old Peace, which included a fresh storyline, 2 new game modes, and a new Warframe named Uriel — and in December 2025 its average DAU and gross receipts hit all-time highs. On marketing services, revenue rose 17% year over year to RMB41 billion. We saw fast growth in the internet services and local services categories, partly offset by slower growth in e-commerce, where platforms briefly moved budget away from marketing into subsidies, and also in financial services, where policy changes affecting online lending hit the quarter. Drivers of growth were better ad targeting, expansion of our closed-loop marketing services, and ad formats tailored to specific advertiser use cases — for example, ads that are playable previews of the mini games being promoted. As we enter 2026, we've deepened collaboration with e-commerce platforms, helping their merchants advertise inside Tencent, and we've added inventory for rewarded video ads in Video Accounts, which has helped marketing services revenue grow faster year on year in the first quarter quarter-to-date than it did in the fourth quarter of last year. At the product level, total time spent on Video Accounts rose thanks to upgrades to the content recommendation algorithm, letting ad impressions grow faster, while our ad load stayed below that of peers.

James Mitchell

executive
#20

Improved conversion rates drove higher marketing spending from Mini Shops merchants. On the Mini Programs side, greater consumer engagement with mini games and mini dramas pulled in more marketing spend from the mini game and mini drama studios. Weixin Search saw overall query volume climb quickly thanks to AI enhancements to search results, which fueled growth in commercial query volume, and search pricing went up as well. Within fintech and business services, segment revenue came to RMB61 billion, an increase of 8%. Fintech services revenue rose by a single-digit percentage year on year, and fintech gross profit grew at a faster rate, powered by wealth management and commercial payment services. Commercial payment volume kept up positive year-on-year growth, helped by a larger number of transactions and a smaller decline in value per transaction. In wealth management, the second biggest contributor to fintech revenue, both average assets per user and the number of users rose year on year. Moving to business services, revenue for the fourth quarter was up 22% year on year, lifted by higher cloud services revenue and greater technology service fees stemming from higher Mini Shops e-commerce transaction volumes. Our cloud services revenue sped up its year-on-year growth rate, reflecting increased demand and a better pricing environment against tight supply of memory and CPU across the industry.

James Mitchell

executive
#21

Our cloud media services revenue saw significant growth, driven by short video platforms and AI video generation services increasingly relying on our media processing solutions—from cloud-based processing of video and audio to playback on devices. This reflects both our industry-leading streaming quality and competitive pricing. Now, over to John.

John Lo

executive
#22

Thank you, James. Hello, everyone. For the fourth quarter of 2025, total revenue came in at RMB194.4 billion, which is up 13% year on year. Gross profit reached RMB108.3 billion, up 19% year on year. Other gains were RMB1.3 billion, versus other gains of RMB2.5 billion in the same period last year, reflecting lower subsidies and tax rebates. Operating profit stood at RMB60.3 billion, up 17% year on year. Interest income was RMB4.8 billion, up 22% year on year, helped in part by growing cash reserves. Finance costs came to RMB3.6 billion, compared with RMB2.5 billion in the same quarter last year, mainly because of a forex loss this quarter versus forex gains in the same quarter last year. Share of profit of associates and joint ventures was RMB6.8 billion, compared with RMB9.3 billion in the same quarter last year. On a non-IFRS basis, share of profit was RMB9.1 billion, up from RMB7.7 billion in the same quarter last year, with the increase coming from improved performance of certain domestic associates due to operational efficiencies and business growth. Income tax expense rose 7% year on year to RMB12.5 billion. On a non-IFRS basis, diluted EPS was RMB6.966, up 18% year on year, outpacing non-IFRS net profit growth thanks to a reduced share count after our share buybacks.

John Lo

executive
#23

Turning to non-IFRS numbers for Q4: operating profit came in at RMB69.5 billion, a rise of 17% compared with a year earlier. Net profit attributable to equity holders reached RMB64.7 billion, also up 17% year on year. Now for fourth-quarter gross margin — the overall figure stood at 56%, an increase of 3 percentage points year on year. Looking at segments, VAS gross margin was 60%, 4 percentage points higher year on year, mainly on the back of bigger contributions from high-margin games developed in-house. Marketing services gross margin came to 60%, up 2 percentage points year on year, with AI-powered marketing services fueling robust growth in high-margin revenue streams, Video Accounts and Weixin Search in particular. Fintech and business services gross margin was 51%, 4 percentage points higher year on year, helped by the expanding scale of cloud services and a better revenue mix in fintech services, together with improved cost efficiency. As for fourth-quarter operating expenses, selling and marketing expenses totaled RMB13 billion, 26% above the prior year, reflecting stepped-up promotional efforts supporting growth of our AI-native applications and games. R&D expenses climbed 20% year on year to RMB23.8 billion, largely because of higher staff costs and greater depreciation expenses stemming from our AI investments. G&A, excluding R&D expenses, was up 8% year on year at RMB12.5 billion, driven by higher staff costs.

John Lo

executive
#24

When the quarter wrapped up, our workforce stood at roughly 116,000 people. That's a 5% rise compared with a year earlier, or 1% quarter over quarter, mostly reflecting added head count in games and on our technology platform, including AI-related head count. Our non-IFRS operating margin for the fourth quarter came in at 36%, which is 1 percentage point higher than a year ago. Operating capex in Q4 was RMB16.9 billion, up 41% quarter on quarter, since we sped up investment in server infrastructure. Compared with a year earlier, operating capex fell 51%, which reflects the concentrated capex spending in Q4 2024 that created a high base effect. Non-operating capex reached RMB2.7 billion, 60% higher year on year, driven by larger facility-related investments. Free cash flow totaled RMB34 billion, more than 6 times the year-earlier figure, reflecting stronger operating cash flow generation this quarter and also lower capex spending versus Q4 '24, as I mentioned earlier. Sequentially, free cash flow dropped 42%, owing to seasonally weaker game gross receipts and the seasonal settlement of certain accounts payable. Our net cash position was RMB107.1 billion, up 5% quarter on quarter, or RMB4.7 billion, mainly driven by free cash flow generation and partly offset by RMB19.6 billion of share repurchases and RMB6.9 billion in net cash flows, primarily relating to investment in other corporations.

John Lo

executive
#25

Looking at the full year 2025, we bought back 153 million shares, and the total consideration came to HK$80 billion. The weighted average number of shares we used in calculating diluted EPS for 2025 went down 2% compared with a year earlier. Since we see strong return opportunities from putting money into AI, we will probably repurchase a lower value of our shares than we did in 2025, so as to fund AI investment while also raising our dividends. Subject to shareholder approval at the upcoming AGM, we propose an annual dividend of HK$5.3 per share, which reflects an 18% increase year on year. That dividend will be paid to shareholders on the 1st of June 2026. Thank you.

Wendy Huang

executive
#26

Thank you, John. At this point we'll begin the Q&A portion. Those of you joining by telephone should press 5 to submit a question, and then press 6 so your line is unmuted. For anyone on the Tencent Meeting or VooV Meeting app, just tap the raise hand icon located at the bottom. So that we can get through everyone, each analyst will be allowed a single question. Our first one is from Kenneth Fong at UBS.

Kenneth Fong

analyst
#27

Hello, good evening, management, and thanks for taking my questions. I have a question about AI versus margin. In our prepared remarks, we expect the increased profit from our existing business to more than cover incremental AI investment. We understand we need to look at this AI long-term investment separately. 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? My second question is also about AI: how do we strategically prioritize given the ongoing constraint in GPU and AI talent? We previously emphasized prioritizing internal AI deployment, but given the recent market development, has management's view shifted? How are we prioritizing allocating resources, or the KPIs that we monitor? Is that the development of a SOTA large language model, or user engagement, or token growth, i.e., 2B solutions? Thank you.

James Mitchell

executive
#28

Kenneth, perhaps I should begin, and then Martin can add to that. I believe it's already implied in what we said at the outset that our revenue could potentially increase more quickly than our profit in 2026, given the increased investment in new AI products. Should that turn out to be the case, we feel quite comfortable with such a result, since we can see that these new AI products give us a chance to broaden our footprint and bring new value to users, and we can also tell from how enthusiastic users are about some of these products that there's a strong opportunity for product-market fit. As for your second question, about constraints on resources when it comes to talent and GPUs, on the talent side we've already been recruiting quite aggressively, bringing in very high-quality people from around the world and from China for the Hunyuan team. We'll keep making selective hires, but in fact we feel we already have a state-of-the-art AI talent team in place. We've managed to build it not only by using compensation as an incentive, but also by fostering the right culture for the team, by appropriately allocating the roles of team members relative to one another and the team's role within the rest of Tencent, by having the team's best leaders in turn draw the best new joiners to the team, by providing the team with ample compute, and by being able to offer the team use cases for the AI products they create that are somewhat differentiated and unique to Tencent.

James Mitchell

executive
#29

On the talent side, I think we were dealing with a scarcity situation, and at this point we feel much more at ease with how things are set up, though we'll keep recruiting selectively. As for GPU constraints, we've been fairly actively provisioning additional compute, and that will arrive progressively and at an accelerating pace over the course of this year, particularly in the second half. That extra compute comes through leasing capacity, through buying higher-end imported GPUs that are now becoming available once more, and through buying a growing quantity of GPUs designed domestically in China. When it comes to using the compute across different use cases, the priority at present is Hunyuan and our new AI products more broadly. The core products are inherently distributed by nature, and they can source compute themselves from local devices, from multiple clouds, from Tencent Cloud, but they're somewhat agnostic about where the compute comes from. We're concentrating our compute on Hunyuan as the core foundation model, and then on the new AI products.

Wendy Huang

executive
#30

Thanks, James. Just a reminder for each analyst to please limit yourselves to one question. If we have more time later, we can take your follow-up then. Our next question is from Robin Zhu at Bernstein.

Robin Zhu

analyst
#31

Thanks, management, for taking my question. So I suppose I'd like to hear your thoughts on—obviously, we're moving into this investment cycle around AI. How should we frame your view of ROI and when those returns come through, and how do you weigh building against renting, and which pieces of the AI stack do you see as most essential to be best-in-breed versus the areas where, eventually, these things will get commoditized as AI keeps advancing? Thank you.

Martin Lau

analyst
#32

From an ROI standpoint, I believe we've already observed quite strong returns when AI is applied within our current businesses, correct? When you look at the financials on a combined basis, and then we separate them out and say, okay, these are the financials for the existing businesses plus the AI investment made to support those businesses, right, the growth is in fact pretty robust. If you take out the spending on new AI products, the operating leverage is plainly evident. That's kind of the first level, right? Then the second level involves investing in new AI products. On that front, I think we'd be seeing those new investments first, right? Revenue isn't really that substantial there, particularly in China's context, unlike the US, where you can actually get consumers paying for subscriptions, and you can get companies paying for coding agents at a very high cost. In China, those things aren't really that available. I think these will show up as upfront investments.

Robin Zhu

analyst
#33

Then

Martin Lau

analyst
#34

As time goes on, we think these new AI products can bring in revenue, and over time they’d deliver quite attractive returns for us. Tencent Cloud is an example we cite: at the start we actually had to put money into the business and take losses, but eventually it becomes a profitable business. We think AI will follow that pattern. With these new AI products, there will be a timing gap between the investment and the return. On building versus renting: if we’re able to buy — right — given how strong our balance sheet is, we’d actually rather buy, because then we don’t necessarily have to pay the extra margin for leasing. Given supply chain constraints and all the various regulations, sometimes we simply have to rent, and we’d do that if we need to secure compute. What was the last question? The last part of the question? Did I cover all your questions? Oh,

Robin Zhu

analyst
#35

The final question was this: when we consider the AI stack — the models, the orchestration layer, the application layer, and so forth — which parts would you say matter most for Tencent to be best-in-breed, versus areas where you think these will be commoditized and it's fine just to have something?

Martin Lau

analyst
#36

I'd say at this moment things are really dynamic, right? In a market that moves this fast, it's pretty hard for anyone to claim, oh, one layer will matter more than all the rest. I think we've got the resources, we've got the people and the team to really invest across all of these layers. The teams especially are quite different from one another, right? With Hunyuan, we actually had to rebuild the team from scratch once again. At this point, as James said, we have a very strong core team, and a very strong ability to keep drawing top talent. Then once you get into the application layer, right, that really plays to our strengths, right, because all of a sudden, you don't even need to have that model capability, but it plays to our strength when it comes to product capability and orchestration capability. Connection, which is our strength, ecosystem is really our strength, and all the infrastructure services such as security are also something we've invested in for a very long time. Being able to go across devices like mobile and PC, right, that's really a core strength of ours, too, right? I think that's really moving into territory where we're strong.

Martin Lau

analyst
#37

We'd genuinely put money into each of these capabilities as well, and market dynamics will unfold on their own, and ideally we'll end up best-of-breed across every layer.

Wendy Huang

executive
#38

Thanks. We’ll take the next question from Ronald Keung at Goldman Sachs.

Ronald Keung

analyst
#39

Thanks to Pony, Martin, James, John, and Wendy. With the recent rollout of QClaw and WorkBuddy, plus the Skill Hub too, I'd like to ask about what potential AI agentic agents hold. How ought we to think about parallels — say, Android — against what we're seeing right now with open claw in this agentic opportunity, and where we sit within it? You brought up Tencent Cloud in that opportunity, and how do we intend to set apart other portions of the stack, models being one example? Thank you.

Martin Lau

analyst
#40

Claw strikes me as a genuinely thrilling idea — it kind of lays out a decentralized framework, a decentralized order, for the way AI functions in this world. There's a parallel, I think, to the way the internet developed. Right at the start, when the internet first showed up, you had a single entry point — the browser — and then essentially a single distribution point, the search engine. As time passed, various services emerged. Then mobile internet arrived, and all at once you saw a whole multitude of applications appear. Among those applications, some are entirely mobile native, mobile-centric, mobile first, while others are mobile apps from the PC internet champions that actually migrated over into the mobile internet world. That, I think, is how we've felt about claw. For a while, everyone was battling to be the AGI hegemon, a monopoly. There seemed to be a moment when people said, oh, if a single model is AGI, then it would rule over everybody. Reality doesn't work that way. You have multiple models turning very strong, and each specializes in different types of activities.

Martin Lau

analyst
#41

One is in chatbot, another in coding, and yet another in multimodal. You've also got open source models that are pretty solid, plus plenty of other models acting as quick followers. Then at one point, in the 2C space, the chatbot looked like the lone entry point. Now, with claw, you can see a fully decentralized regime opening up, where lots of companies can each have their own claw, and that claw can draw on all kinds of different models, right? Cloud's infrastructure supports it, and every claw has to work out its own unique value proposition, right, to win users' hearts. The claws also make use of not only the cloud, not only their own unique model, but also the tools on the devices available to them, and they utilize the file system, right? It turns into a far more exciting decentralized world, and we sensed there are plenty of opportunities for us, right, when it comes to building products that cater to people's needs. That's why QClaw exists, and WorkBuddy too, and going forward, I think many existing apps will try to come up with claws of their own, right, along with their own agentic capabilities, and different models will likewise try to compete to win the hearts of these claws.

Martin Lau

analyst
#42

It turns into a far more thrilling and decentralized world where everyone can take part, and as I mentioned, we simply have to develop know-how across the various layers — the model layer, the product layer, the infrastructure layer — and each one would need to offer its own particular value proposition in order to earn its own usage.

Wendy Huang

executive
#43

Thanks, Martin. Our next question comes from Ellie Yap at Macquarie.

Ellie Jiang

analyst
#44

Thanks a lot, management, for taking my question. I do have a follow-up, and it's really about the question that came up just now regarding the agentic era. In this new agentic era, how would management assess Tencent's value proposition? Given that we're placing Hunyuan alongside the other LLMs, sort of for prosumers and consumers, how can we potentially keep the other LLMs from diluting the value of our own foundation models over the longer term? Thank you.

James Mitchell

executive
#45

When we consider Tencent's distinctive value proposition and what we can deliver to users in the claw era, a few inherent attributes we hold strike us as particularly well suited to claw deployment, and Martin has already mentioned some of these. One such attribute is that our capabilities extend across PC, mobile, and cloud. Our capabilities reach across applications and the World Wide Web, much as agentic claws reach across devices and across domains. We are a company that runs a number of centralized apps, yet also hosts some highly decentralized but vibrant ecosystems, most notably the Mini Program ecosystem. One way to frame this is that in earlier years, the arrival of the mobile internet effectively turbocharged the more centralized app experience relative to the more decentralized World Wide Web experience. Now, with these agentic capabilities and claws, an opportunity exists for decentralized experiences like Mini Programs to be turbocharged and to develop far more powerful capabilities than they had in the past. That is why we believe there is inherently a natural fit among our capabilities, our interests, and the deployment of these agents or claws, and it is one reason we see consumers and prosumers enthusiastically adopting our own agents and claw services.

James Mitchell

executive
#46

On the piece of your question about other large language models diluting the value of ours — I may not quite grasp the premise correctly, but I don't see that happening. When you use these claws, you go into them and you have a choice. Do you want Model A, which delivers very high performance at a high price per token, or Model Z, with medium performance and a very low price per token, or models B through Y in between? That's part of what makes the claws appealing. Hunyuan is one of the models that's available. With the capabilities of the Hunyuan team now in place, we believe that going forward Hunyuan will improve faster and faster, and so consumers will naturally increasingly choose to use Hunyuan. I don't think it will be a monopoly situation. The claws that succeed will be the ones that keep letting consumers and prosumers make their own choice along the price-performance curve, and different models will sit at different points on that price-performance curve, and we want to be one of those, but we don't intend to be the only one of those.

Wendy Huang

executive
#47

Thanks. Our next question comes from Alicia Yap at Citi.

Alicia Yap

analyst
#48

Hello, good evening, management, and thanks for taking my questions. My question relates to physical AI. Given how productivity-focused AI agents are proliferating across enterprises, particularly in traditional industry, do you think this will speed up demand for the use of world models, such as the 3D models that you have? And also, how does management assess Tencent's capability as well as its competitive strength in the coming physical AI era? Thank you.

James Mitchell

executive
#49

Alicia, I believe you're making a fair point — computer-aided design tools already exist, and it's only natural to expect AI to augment those capabilities and in time give them a major boost. That matters for industrial design. It matters for architecture. And in video games it's genuinely very important, and growing more so. Because our games contain such a wide and deep range of 3D graphical assets, we think we're in a rather uniquely strong position to supply the data needed for training the models that will in turn power those 3D tools. In the end it's a sizable niche, and one we're well placed to serve. I wouldn't call it the largest opportunity in front of us. There are plenty of bigger, nearer-term opportunities.

Wendy Huang

executive
#50

Thanks. Our next question comes from John Choi at Daiwa.

John Choi

analyst
#51

Thanks for taking my question. I have one about games and how AI disruption plays in. We're already beginning to see some effects on headcount and on game development costs. How do you expect AI to affect quality, and also the cost side overall? How should we expect Tencent to get ready for this? And as we see a rising number of games, would distribution and publishing matter more further down the road? Also, if AI brings down the barriers to entry for development, will we see a meaningful rise in the supply of game studios and in the total quantity of games further down the road? Thank you.

James Mitchell

executive
#52

Thanks for that question, John. I'm not sure whether any of you were at the Game Developers Conference last week, but each year it's something of a premier event for game developers. As you'd expect, there were a number of well-attended presentations on using AI in creating games. I think there are a couple of broad observations: one is that those presentations were pretty exclusively focused on how to use AI to upgrade content within existing games, to accelerate the content creation, improve the content creation within games, but that there is not yet the capability to create games completely from scratch using AI for a number of reasons that we can get into. Then the second observation is that many of the best-attended presentations were by our colleagues within Tencent's Interactive Entertainment Group. They talked about how AI can be deployed in games for graphics, AI can be deployed in games for gameplay, AI can be deployed in games for user companionship, and so forth. We believe that we're at the forefront of the industry in this regard, and the feedback from many of the developers who attended the Game Developers Conference was consistent with that belief.

James Mitchell

executive
#53

On the second part of what you asked — whether AI leads to a huge wave of new games, which would make publishing matter more relative to development — the unfortunate truth about this industry is that oversupply has always been with us. As Martin noted, each year brings 200,000 new titles on mobile and 18,000 new titles on Steam. If that figure rose from 200,000 to 2 million, to 2 billion, to 2 trillion, the extra impact would be fairly marginal. What really counts is creating the strongest games, then growing them and keeping them evergreen. Doing that requires the finest people, supported by the finest technology. So our view is that the value balance between development and publishing stays as it is today, and the keys to success will keep favoring the industry's best developers. Thank you.

Martin Lau

analyst
#54

Let me add a couple more points here. First, when people bring up AI disruption in gaming, the implication is usually that it's harmful to the industry, but my view is that gaming is actually among the sectors that stand to gain from AI. As AI spreads, people will end up with more spare time on their hands, and demand on the consumer side should rise substantially for gaming — something I see as a genuine certainty as AI proliferation unfolds. Second, great tools are going to become available to new developers, yet they'll likewise be available to well-organized teams and highly talented developers who are already running large evergreen games. When a tool does become available, it will disproportionately favor those who hold the resources and who already have all the gamers gathered around their platform, and they can use these tools more effectively to boost production output and make their games even more evergreen. I see that as an advantage for players that have evergreen games and are also extremely fast and agile when it comes to embracing technology.

Martin Lau

analyst
#55

And then, when you have a lot of innovations coming at once, right — I think what happened in gaming is that an idea shows up, and at first it isn't perfect, and then it gets iterated on and refined as time goes by. Again, I believe that process would actually get accelerated if many of these games with large user bases examine these innovations and are able to iterate more quickly, folding these new experiences into the games they already have and turning games basically into platforms. I think that's a unique opportunity we'd see develop over time as well.

Wendy Huang

executive
#56

Thanks. Our next question comes from Alex Yao at JPMorgan.

Alex Yao

analyst
#57

Thanks, Wendy. And thanks to management for the chance to speak. I'd like to pick up on the AI cloud part of the business, since demand for AI compute is very strong, yet at the same time there's price inflation on AI servers because DRAM and HBM costs keep rising. Could you help us get a handle on Tencent Cloud's pricing power, and also your philosophy on value capture in such a dynamic environment? To put it another way: do you intend to pass the full cost inflation through to customers, or subsidize part of it and gain more market share, or perhaps even pass through more than the full cost inflation to capture additional profit? Thank you.

James Mitchell

executive
#58

Thanks for that question, Alex. Let me begin by noting that obviously we're seeing a big jump in demand for AI compute, though that's not the only place demand is rising. Whenever people use those agentic tools we've talked about, they employ them, they build software, and that software above all has to run — and as it runs, the bulk of it isn't running on a GPU; it runs on CPU, and as it runs it generates memory requirements. It isn't only GPU, DRAM and HBM where demand is climbing — CPU too, ordinary RAM, SSD, and hard disk drives as well. Demand is rising everywhere you look. As for how the industry overall reacts on pricing, well, for years the industry has struggled, since cloud service providers in China were running on extremely thin margins, and a key reason those margins were so thin was that whenever a new player entered or customers chose to obtain infrastructure straight from the source, they could simply pick up the phone, call the supplier, and order whatever infrastructure they wanted from the company providing the CPU, GPU, or DRAM.

James Mitchell

executive
#59

That is no longer true. At this point, supply is committed months ahead, quarters ahead, and in certain cases years ahead. Priority goes to the largest and most consistent customers — the hyperscalers, like us — which means the smaller cloud providers can no longer be sure of getting supply, so they have to come to the hyperscalers. Hyperscalers have been running on thin margins, so once demand rises, we as an industry really have no option except to pass along higher prices, and as a result you have seen several price hikes in China Cloud over the last 24 hours. On how we value capture in this kind of dynamic environment, one broad principle is that we aim to deliver more value through enrichment. Enrichment means that at a minimum, if you have compute, you can rent it out bare metal, and you get a certain low price and low margin. Preferably, you subdivide it and virtualize it into tokens, and then you get a higher price and higher margin per unit of compute. Ideally, you bundle it into platform as a service or software as a service, and then you can get the best pricing and the best margins.

James Mitchell

executive
#60

That's a piece of the path we've traveled, and it helps explain how Tencent Cloud went from really big losses 4 years back to fairly large profits in the last year, and we'll keep going along that path of shifting from bare metal to token to platformization and on to software. Thank you.

Martin Lau

analyst
#61

Thank you.

Wendy Huang

executive
#62

Thanks. The next question comes from Gary Yu at Morgan Stanley.

Gary Yu

analyst
#63

Good evening, and thanks for taking the question. I have one thing I'd like to ask about. You've said on a number of occasions that in AI we aren't a first mover — that we're actually a latecomer. And in the US, we've seen that catching up is getting really hard for some of the late arrivals, even ones with very substantial resources when it comes to compute, talent, and data. So how does management reach a point of comfort and confidence that we won't end up on that same path — falling behind, unable to close the gap — in the areas of compute, models, and applications? Thank you.

Martin Lau

analyst
#64

That's a really good question, I think. If you're only playing a single game, it's basically tough to catch up in that one game, right? But if you see AI as a whole bunch of different games, then fresh opportunities and new frontiers keep opening up all the time, right? And I think that's already happening, right? If you look at the model, right, at the very start, everyone thought, oh, it's the chatbot, and then coding shows up, and then multimodal shows up. Then when everyone thought, oh, that's about it, suddenly claw shows up, which basically decentralized the entire AI landscape even further. Going forward, we actually thought, oh, there will just be apps, right? There will be many different permutations of how AI gets packaged, from model to product to agents, and existing services will be gaining different agentic capabilities. New agentic capability will be arriving on mobile, on PC. It's very early days in the whole AI development world. That's why in just a short period of time, you can see there's already a lot of proliferation, and more and more will keep coming.

Martin Lau

analyst
#65

So that’s why having some core capability really matters, right — and we do have plenty of those at the application layer, whether it’s Weixin or our ecosystem with communication and presence across PC and mobile, plus a lot of infrastructure capability, including security, cloud, and payment. In this new AI race, all of those pieces can be bundled together. It isn’t really a single race — it’s actually a world of many, many races, and I think that will show up more and more, and because of that, there will be plenty of opportunities for different players to emerge and innovate from behind. I’m not really worried about being late; what I would be worried about is whether we aren’t innovating quickly enough, right — which I think, as we restructure our Hunyuan team and as we’ve begun to energize all our product teams to start innovating with products, is actually happening in a very exciting way for us.

Wendy Huang

executive
#66

Thanks. The next question comes from William Packer at BNP. William, your line is open.

William Packer

analyst
#67

Hello, management. Appreciate you taking my question. Media reports indicate that Apple intends to lower App Store commission rates in China during March — by 5% on apps and down to 3% for mini games — and Tencent stands to gain significantly from that. How much of those reductions should we anticipate reaching Tencent's gross margin, or would they instead be split among other parties, like consumers, gaming partners, or possibly tax revenue? Thank you.

James Mitchell

executive
#68

Hi, William. In this instance, happily, the press coverage rested on the objective reality of a formal Apple announcement, so we aren't dealing with a speculative hypothetical. This is a real development that takes effect over the last few days. As for the flow-through, there should be a good flow-through. Where we have game development partners and we serve as publisher of those games — which is now a fairly small minority of our game revenue — the revenue share, in the overwhelming majority of cases, gets calculated on gross revenue rather than on revenue net of the Apple take. That flows through to us. If by taxes you mean us paying a teens percentage corporate income tax on this incremental profit stream, then I suppose that's correct, depending on how much of this incremental profit stream we reinvest into new AI products. I think you spoke about one part of the Apple announcement, namely the quantitative part that moves from 30% to 25% and 15% to 12%. For us, actually, the more important aspect of the Apple announcement looking forward was that Apple stated it would effectively offer developers in China equivalence with whatever lower rate developers elsewhere in the world pay to App Store.

James Mitchell

executive
#69

We think that as industry trends keep changing, it's just a question of time before the fees App Store charges drift lower across various regions. And with this announcement, Apple has made clear that as take rates come down in various regions, take rates will come down in China in step. In our view this is a highly encouraging initial move, though it is only the first move in what should be a multi-step favorable path. Thank you.

Wendy Huang

executive
#70

Thanks. Our final question comes from Alex Liu at Bank of America.

Alex Liu

analyst
#71

Oh, thanks for taking my questions. My question is really just about AI chips. We're seeing more and more of your tech peers putting priority on building their own chip design capabilities. I'm just curious how in-house chip development fits into Tencent's AI priorities. Thank you.

Martin Lau

analyst
#72

Appreciate the question. I'd say right now, that isn't really the main thing people will be zeroing in on. When you think about chips, training chips and inference chips are not the same thing, correct? On the training side, designing and manufacturing is genuinely extremely hard, and you really do want the most advanced training chips available, as much as you can get, and in the most flexible manner, so that you're able to keep training toward the strongest model. Then when it comes to inference, I'd say inference is largely a matter of cost. On cost, right now there are actually plenty of different suppliers in China, which is quite unlike, say, the training area, where basically 1 or 2 players can command an extremely high margin, correct? In inference, companies are generally making far lower margins, and there are many more solutions and choices. What matters most for us, I think, is leveraging the best training chips to train the best model right now, and there's considerable value in staying focused.

Martin Lau

analyst
#73

Now regarding inference, over time I believe the way the market plays out will make inference chip margins quite manageable. Right now, our focus is very much on using the best chip available to train our model. Hunyuan 3.0 is going to be significantly better than Hunyuan 2.0, and that is really just the starting point. As time goes on, I think we'll be able to iterate on training our model more quickly, and I'm quite confident that with that focus, we'll reach SOTA at some point. That, I think, is the most important thing for us. The next most important thing is truly unleashing our product development capability and our integration and connection capability so we can design the most exciting AI products for users. Once those are done, then we'll consider how to try to bring down the cost of inference.

Wendy Huang

executive
#74

Thanks, Martin. That brings our webinar to a close. Appreciate everyone being with us for today's results. Should you want to review our press release along with other financial materials, head to the IR section on our corporate site, www.tencent.com. A replay of this webinar is going to be posted shortly as well. Thanks again, and we'll see you in the next quarter.

Read the full transcript via the API

You're viewing the first half of this call. Get the complete Tencent Holdings Limited transcript — plus 132,000+ transcripts from 7,000+ companies, speaker segments, AI summaries and full-text search — through the EarningsCalls.dev API.

Get the API View API docs →

This call discussed

For developers and AI pipelines

Programmatic access to Tencent Holdings Limited earnings transcripts and 132,000+ others is available through the EarningsCalls.dev REST API. Plans from $39.99/month — full transcripts, speaker segments, full-text search, and the recently-added /api/v1/transcripts/recent polling endpoint for ETL pipelines.

More tools for stock research

OptionsBell: an unusual options activity scanner that tracks options flow across all US stocks and emails you when large trades appear in Tencent Holdings Limited or any ticker on your watchlist.

StockMarketScan: a stock screener with alerts: 24 pre-built screens refreshed every 5 minutes, watchlists and email notifications. Both offer API and MCP access and fair monthly pricing.