Kuaishou Technology B

Good day, ladies and gentlemen. Thank you for standing by. Welcome to the Quite Sure Technology fourth quarter and full year 2025 financial results conference call. Please note that English simultaneous interpretation will be provided for management's prepared remarks. This English line will be in listen-only mode. I will now turn the call over to Mr. Matthew Chow, VP of Capital Market and IRR at Quite Sure Technology. Thank you, Arbiter. Good evening to everyone. Welcome to Kuaishou Technology fourth quarter and full year 2025 financial results conference call. Joining us today are Mr. Chen Yixiao, co-founder, chairman, and CEO. Mr. Jinping, chief financial officer. Before we start, please note that today's discussion may contain forward-looking statements which involve a number of risks and uncertainties. Actual results and outcomes may differ from those discussed The company does not undertake any obligation to update any forward-looking information except as required by law. For all important information about this call, including forward-looking statements, please refer to the company's public information on or before the quarter and for year 2025 results. Announcement ended December 31, 2025, issued earlier today. During today's call, management will also discuss certain non-IFRS financial measures. These are provided for additional information and should not replace IFRS-based financial results. For a definition of the non-FRS financial measures, a reconciliation of IFRS to non-FRS financial results and related risk factors, please refer to our fourth quarter and full year 2025 results announcement. For today's call, management will use Chinese as the main language. A third-party interpreter will provide a simultaneous English interpretation in a preparatory mock session and a consecutive interpretation during the Q&A session. Please note that English interpretation is for the convenience purposes only. In case of any discrepancies, management statements in the original language will prevail. Lastly, a loss of the listed occurrence mentions are in R&B. I will now hand the call over to Isha. Hello, everyone. Welcome to CoinShow's fourth quarter and full year 2025 earnings conference call. Over the past year, guided by our tech-driven, user-centric philosophy, we accelerated the execution of our AI strategy across all major business areas. Our clean AI, multimodal, large video generation models maintain a global leading position, and we continue to leverage our advanced capabilities to empower quite a choice of content and commercial systems. These efforts supported a high quality growth across the user scale, revenue expansion, and profitability. In Q4, 2025, average DAUs on the Quasio app reached 408 million, representing solid year-over-year growth. Total revenues for Q4, 2025 increased by 11.8% year-over-year to 39.6 billion. Revenue from our core commercial business, including online marketing services and other services, primarily e-commerce, increased by 17.1% a year-over-year. Adjusted net profit is increased by 16.2% a year-over-year to $5.5 billion. For the full year 2025, average GAUs across Europe reached $410 million, and total revenues increased by 12.5% a year-over-year to $142.8 billion. Adjusted net profit for the full year increased by 16.5% a year-over-year to $20.6 billion. with an adjusted net margin of 14.5%. As we scaled AI investments, we continue to deliver steady improvements in the group's overall profitability. Our AI capabilities have become a core engine driving Qyto's long-term growth. Meanwhile, as disclosed in the results announcement, given the company's business performance, the board has recommended the payment of a final dividend of 0.5%. 6.69 Hong Kong dollars per share for the year ended December 31st, 2020, amounting to approximately 3 billion Hong Kong dollars in total. This reflects our confidence in the company's long-term growth prospects and solid financial position, as well as our unwavering commitment to enhancing shareholders' value and sharing the benefits of the company's strong cash flow generation. We are sincerely grateful for our investors' continued support. Quite a steady growth would not have been possible without the trust and support of our shareholders looking ahead. By staying closely aligned with the business development and marketing conditions, we will flexibly evaluate and continue advancing diversified shareholder returns, including share repurchases and dividend distributions, to deliver the fruits of our growth to all our shareholders. Next, I will walk through the detailed progress of our major business segment in Q4 2025. First, our AI strategy and the progress of our large video generation model, Culling AI. Culling AI remained committed to its core vision of empowering everyone to craft competitive stories with AI, aiming to become the premier inclusive, efficient video generation infrastructure for the AI era, while driving continuous breakthroughs in model capability, product experience, and monetization. In Q4 2025, Collinear exported a rollout of multiple model upgrades across several iterations. We launched Collinear 1, the world's first unified multimodal video model, developed on the multimodal visual language architecture. Collinear 1 transcends traditional single-task video generation models by integrating multimodal text-to-video image and subject inputs in a single generative editing engine. Collinear 1's unified architecture enables anti-anaconic creation within one model, system, allowing users to envision systematically from generation to editing and refinement without switching tools. We also released the Cling Video 2.6 model, which incorporates simultaneous audio-visual generation capabilities. The model can generate a complete video containing natural voiceover, sound effects, and ambient audio in a single process, enhancing creative efficiency across the AI video creation flow. Cling Video 2.6 also introduced a motion control feature that enables users to replicate specific movements from uploaded videos. or from the online motion library. By pairing this with the character reference image, users can generate character-specific videos with the frame-level precision in both body movements and facial expressions. In February 2026, we launched the Cling AI 3.0 model series, developed under an all-in-one product framework. Cling 3.0 supports full multimodal input and output, spanning text images, audio, and video, integrating video understanding, generation, and editing within a single streamlined AI workflow. This mode unifies multiple tasks within a native multimodal architecture, enable more complex narrative logic, automated storyboarding, and precise shot control while maintaining strong prompt adherence. Cooling AI's innovations in foundational models and product features have paved the way for widespread commercial applications across professional creative sectors, including marketing, e-commerce, film, and television, short plays, animation, and gaming, These capabilities have supported a stronger adoption among professional creators and enterprise clients globally, earning the models widespread acclaim and accelerating their monetization. In Q4-25, Kling AI achieved a revenue of $314 million. Notably, in December 2025, Kling AI's monthly revenue exceeded $20 million, corresponding to an ARR of $214 million U.S. dollars. At the same time, Cling AI's motion control feature gained significant traction across major global social media platforms, driving widespread discussion and organic distribution. This momentum broadened Cling AI's reach beyond professional creators to a broader mainstream user base. In Q4-25, we continued to deepen the impact of large AI models to empower our content and our commercial ecosystems, while driving further quality and efficiency improvements in our organizational infrastructure. In terms of strengthening the foundation of our content ecosystem, our proprietary multimodal large language model, Kuai Ke Yi, $671 billion model, has demonstrated strong video comprehension capabilities. Meanwhile, we upgraded our short video and live streaming content understanding system and launched a tech next, our next generation teching system, which enables more accurate content understanding, grabbing higher app usage time per user, and a retention rate. In content recommendation, iterated our end-to-end generative recommendation large model with the launch of One Rack V2, continuously enhancing the precision of our recommendations. For online marketing services, we further optimized our end-to-end generative recommendation technology. By deeply integrating multi-dimensional business data, we enhanced model performance and improved the precision of online marketing material recommendations. For intelligent bidding technology, we developed a unified bidding large model built on multi-scenario and multi-objective data, Together, our generative recommendation large models and intelligent bidding models drove roughly 5% of growth in domestic online marketing services revenue in Q4-25. While reducing the cost of generating online marketing materials, AIGC technology also unlocked additional budgets from our online marketing clients. In Q4, the total spending from online marketing services driven by AIGC marketing materials was nearly $4 billion. For e-commerce business scenarios, Q4, we further iterated our end-to-end generative retrieval architecture one search. We introduced addable structured semantic identifier tailored to the e-commerce business, enhancing semantic understanding for mid- to long-tail search queries. This drove a nearly 3% increase in search order volume in Shopping Morgan Q4. In addition, we expanded the application of end-to-end generative recommendation technology from pan-shelf-based e-commerce to content-driven scenarios such as live streaming rooms, and short videos propelling GMV growth in all e-commerce scenarios. For live streaming business scenarios, we further refined the universe gift customization feature to deliver better interactivity, reach a dynamic presentation, more refined visual aesthetics, significantly increasing users' willingness to send official gifts. To drive organization efficiency, we have completed the upgrade of our intelligent coding tools. Our self-developed AI programming tool, CodeFlicker, has evolved from a coding assistant to an AI engineer, with more engineers adopting the agent-based coding model. And the AI generation rate of new codes has rapidly risen to over 40%. Moreover, our AI advancements are underpinned by our investments and our optimizations in computing power infrastructure. Building on the success of our self-operated in-house wall-and-chopper data center, we are steadily advancing the construction of our new computing power center to continuously elevate server and bandwidth operating efficiency. Second, user growth and content ecosystem. In Q4-25, average DAUs on the Quest App reached 408 million, and AMA use reached 741 million, with average daily time spent per DAU on the Quest App was 126 minutes. We're committed to building a vibrant community with a distinctive quality question characteristics, continuously strengthening high-quality user growth, differentiated premium content supply, traffic mechanism optimization, and interactive scenario development to achieve a healthy, sustainable expansion in both the user skill and traffic. To drive high-quality user growth, we refined user acquisition strategies across channels to continuously optimize user segments and improve retention rates. We also leveraged AI technologies and enhanced push strategies, resulting in a higher open rate for Quadro app. In addition, we introduced innovative user engagement and retention initiatives that consistently improved ROI. Harnessing our established capabilities in content operations, we supported the growth of benchmark creators like Xinyu, the illustrious lady, and continued to cultivate high-quality top-tier content IPs with distinctive Quadro characteristics. Rural culture and entertainment activities, exemplified by the Village Gateway mini stage, enabled rural residents to transition from passive viewers to active on-stage participants, featuring diverse content ranging from intentional cultural heritage performances to agricultural technology demonstrations. These initiatives both enriched rural culture like and provided a new channel for showcasing rural culture. We produced the sixth anniversary concert for Teens in Time, which garnered over 680 million live streaming views. Leveraging live streaming, interactive features, and AI-powered creative content, We crafted a shared youthful memory that fosters a mutual ground between the fans and idols. We optimized our traffic mix to increase traffic exposure for top-tier original content, fostering a virtuous cycle between content creation and consumption. In Q4, the number of high-quality content uploads increased over more than 15% year-over-year. To further develop engagement scenarios, we continue to innovate private messaging and game events features, driving year-over-year increase of nearly three percentage points in rapidly in daily average penetration rate of private messages among users with mutual followers during the quarter. Third, online marketing services. In Q4, revenues from online marketing services to reach 23.6 billion RMB, up 14.5% year-over-year. The accelerated integration and innovative application of AI across cyber's online marketing services scenarios not only empowered our ecosystem partners, but also injected a new growth momentum into our online marketing services business. In Q4, Q5, within the life cycle lifestyle service sector, where clients primarily operate under a data-based model. We help the client to reach users more efficiently and achieve the higher user conversion rates by operating our private messaging products and optimizing our algorithms. At the same time, by continuously expanding into more industries and acquiring new clients, we've broadened our online marketing client base and generated incremental marketing placements. In addition, as the lifestyle service sector clients are a predominantly small and medium-sized business, leveraged AIGC tools to enhance their ability to produce marketing materials. These enhancements lower the barriers for marketing placement and growth for the growth in online marketing spending. In Q4, content consumption sector led by short plays, comic style short plays, mini games, along with the application sector, with a key growth revenue driver for the non-e-commerce online marketing services. In the content consumption sector, short plays continue to sustain solid growth by optimizing marketing materials exposure formats. We increased the marketing spending in short play verticals Meanwhile, empowered by the deep integration of AI technologies, comic-style short plays have advanced rapidly through continuously comprehensive support programs and bullet of the comic-style short play AI agent with further expanded high-quality and diverse content supply to capture emerging growth opportunities in the sector. Moreover, a rising market budget from clients across the application vertical will leverage our insights into industry placement pace and market trends to consistently optimize resource allocation and conversion efficiencies. effectively channeling and capturing marketing placement and spending from AI application clients. In Q4, for online marketing products, we continue to upgrade offerings, including our UX placement solutions, EIDC marketing material solutions, live stream digital human solutions, and digital employee solutions. These enhancements help to lower barriers to marketing placements, improve clients' placement experience, and drive further growth in online marketing spending. Specifically, during the quarter, UX developed a periodic delivery and account-level smart placement product These upgrades enable clients to extend managed campaign cycles and algorithmic system management from the app unit level to account level, thereby improving overall delivery efficiency, raising the ceiling for campaign scale, and providing more stable cost performance for our clients. In Q4, penetration rate of our UX placement solutions accounted for nearly 80% of the non-e-commerce marketing spending, and its penetration rate among active users exceeded 90%. For e-commerce marketing services, following our consolidation of e-commerce business and related online marketing teams in late September last year to advance traffic synergy, we established our closed-loop capability, simplifying traffic, transaction, online marketing, conversion, and merchant services This was designed to align our platform's overall revenue growth with merchant-based refinement, enabling e-commerce merchant and GPM and CPM for marketing services to improve intent in Q4. In the first half of 2025, we essentially completed capability refinement of our Omni platform marketing solution. In the second half, we focused on addressing differentiated scenario needs across diverse customer segments, effectively increasing incremental GMV generated for e-commerce merchants across Omni domain scenarios and enhancing business stability in Q4. Our Omni platform marketing solutions accounted for even greater share of our total e-commerce marketing spending, rising further to 75%. Our Omni platform product promotion achieved full coverage across products and scenarios, becoming the primary placement offering for our e-commerce marketing services. Our fully managed auto placement and product combo for small and medium-sized merchants gained broader adoption and recognition, driving a significant increase in spending by these customers. In Q4, by continuously optimizing our pen-shelf-based e-commerce scenarios and strengthening the synergy of Omni domain supply and online distribution, our e-commerce marketing services revenue pen-shelf-based scenarios increased rapidly year-over-year. Fourth, our e-commerce business. In Q4-25, our e-commerce GMV grew 12.9% year-over-year to $521.8 billion. Building on the systematic omni-domain operations strategy for their integrated pathway between public domain, traffic conversion, and private domain asset accumulation, unlocking a new growth engine for merchants and supporting their stable, sustainable operational development across diverse scenarios. During Q4, we continue to empower merchants to strengthen their private domains and operational efficiency, broadening a variety of supply. As a result, repeated purchase frequency of active e-commerce users further increased year to year. Meanwhile, by enhancing the operations of our key product categories and more precisely identifying the needs of our core user bases, we drove continued growth in ARPPU. In two parts of New Divide, we mobilized a combined strength of service provider agencies and industry zones to broaden our e-commerce supply, guided by a full lifecycle framework for new merchant development. We deepened our cost reductions and efficiency enhancement, stepped up incubation programs for new merchants, strengthened support for merchants from industrial zones and for the optimized business environment. Collectively, these measures reinforced merchants' operational stability, empowered both new merchants and small and medium-sized merchants to grow. and enhanced long-term predictability, sustainability of merchant operations during Q4. Both the newly onboarded merchants and newly onboarded active merchants grow year-over-year in quarter-over-quarter, driving our active merchant base to another record high, up 7.3% year-over-year. Furthermore, in Q4, we launched the Voyage initiative, focusing on in-depth partnerships with the top tier brands in diverse sectors. Through coordinated resource empowerment, the initiative aimed at a pioneer new model of mutually reinforcing growth for both the platform and the brands. At the end of December, we began to capture early benefits from our high-quality product and content supply, as well as merchant mix optimization. In terms of our live streaming scenario development, the pop-up follower red envelopes initiative, which was launched in Q3 to drive targeted follower growth, achieved a meaningful result. By increasing the streaming frequency of streamers with over 10,000 followers, the program drove a 12.7% increase in the number of average daily active streamers hosting live sessions with over 10,000 followers. for the reinforcing virtual cycle follow-up growth and transaction performance in Q4. Through coordinated operations with agencies and leading KLL organizations, we expanded our KLL supply to further empower KLL as we advanced our platform-endorsed product offerings, which are trusted by both merchants and KLLs, building on this foundation, our KLL blockbuster initiative focused on high-demand product categories, highlighting our platform's strong order aggregation capabilities and driving greater KLL participation and distribution, The penetration of KOL within our distribution pool continued to improve, with a number of active KOLs more than doubling year-over-year, supported by our platform-endorsed product offerings. Mid-tier to small and medium-sized KOLs were able to overcome product selection challenges, and with platform traffic support, their two meaningful leaps in operational scale. In Q4, our Omni domain operations ecosystem, including pen-shop-based e-commerce and shorter videos, continued to demonstrate steady and resilient development. In Q4, the contribution of pen-shelf-based e-commerce GMV to total e-commerce GMV remained broadly stable quarter-over-quarter. We continue to expand our supply scale, driving sustained year-over-year and quarter-over-quarter increases in average daily active merchants for pen-shelf-based e-commerce. Superlinks and the official channel for platform-recommended products continue to strengthen its role as a core operational tool for shelf-based offerings, which achieved rapid growth during the quarter. In Q4, Superlinks penetration rate in shelf-based e-commerce product cards reached 19.1%. We also encourage merchants to expand all e-domain operations by leveraging our marketing hosting tools. We guided merchants in content-based scenarios toward shelf-based operations, significantly increasing the penetration rate of active merchants using our marketing hosting tools quarter-to-quarter. During Q4, we further advanced our short-video e-commerce content supply, prioritizing refined merchant-centric operations by continuously averaging the synergy between short videos and live streaming. We enriched our high-quality supply and optimized the file efficiency. These efforts were a significant growth in short-of-video e-commerce GMV, which continued to outpace overall e-commerce GMV growth. In Q4-25, we deepened AI integration across e-commerce scenarios, delivering tangible efficiency gains for merchants while supporting their growth. The broader rollout of Orange Rack, OneSearch, and other large-language model technologies across e-commerce scenarios continue to generate incremental value. Powered by an e-commerce knowledge graph and leveraging large models of world knowledge and reasoning capabilities, We strengthened our foundational understanding of products, buildings, and users. This enabled a more accurate long-term user interest in modeling, improved recommendation diversity, and drove high revisit and repurchase behavior. E-commerce content and generation capabilities have also advanced during the fourth quarter. Features such as live streaming highlights and AI-assisted content creation further strengthened emergence across narrow operating capabilities, propelling step-change growth in both content output and GMB. To improve operating efficiency, we launched an AI-powered order analysis feature in Q4, helping merchants identify abnormal orders more effectively, reduce pre-shipment and refund rates. Next, regarding our live streaming business. In Q4, live streaming revenue was 9.7 billion RMB. We remain focused on fostering a healthy live streaming ecosystem during the quarter, orienting toward high-quality, value-driven content and reinforcing the platform's community-centric core. For live streaming supply, we continue to intensify professional operations of our core competitive categories, including group live streaming and multi-host live streaming, while strengthening coordinated development across multiple categories. This enriched our live streaming content operations portfolio and drove sustained improvements on the supply side, better serving users diversified preferences. Our grand stage of deepening integration between online and offline scenarios, supporting the incubation of distinctive streamers on our platform, while increasing user engagement, On the product front, powered by CleanAI's video generation capabilities, our AI universe gives users with customizable special effects, enhanced interactive feature experience, dynamic motion rendering, and visual aesthetics. As of the end of the fourth quarter, the number of cumulative AI universe gift creations exceeded 1 million. In addition, we expanded the application of AI capabilities in our live streaming rooms, empowering streamers with AI interaction assistance, and AI digital avatar solutions to improve the streamer service efficiency. In Q4, our live streaming plus model extended the boundaries of the live streaming ecosystem while also unlocking additional commercial value. Through refined operations, our IDO housing and acquired hire business delivered both quality improvements and official gains. In Q4, the average monthly number of IDO housing paying clients increased by over 40% year-over-year. Finally, our overseas business progress. In Q4, we remain firmly committed to our high-value growth strategy, supporting a virtuous business cycle across our overseas business. Despite a complex market dynamics, we achieved a steady growth in overseas business. On the traffic front, while improving customer acquisition efficiency and optimizing our user growth structure, we strengthened the user mindshare for the Kuaishou community by expanding the supply of content. With the distinctive Kuaishou characteristics further broadening our core user base, Brazil, our key market for business development and maintaining stable DAUs and time spent per DAU. For online marketing services, we captured an industry opportunity to expand brand presence in Brazil. We're in our client base across diverse industries. In addition, we upgraded our products and solutions and actively explored the new content-driven marketing scenarios, including short reviews to improve client performance visibility and unlock a new growth momentum of supporting our client's long-term development. Our e-commerce business in Brazil achieved a steady year-over-year growth in GMB transaction scale and a little volume in Q4. Supported by AIGC-driven improvements in e-commerce content and quality and operational efficiency, and edited by more refined logistic cost of management, our overseas profitability improves significantly. Looking back over the past year, despite multiple challenges, we anchored our core AI-first strategy, leveraging our profound technological expertise, a thriving, diverse content ecosystem, and a continuously enhanced infrastructure and commercial footprint. We collaborated with ecosystem partners to drive system-added growth. Looking ahead, although challenges will intensify, we remain steadfastly guided by our user needs. We deeply cultivated the building of a warm, inclusive, and universally accessible digital community, while continuously deepening the seamless integration of AI technologies across our business hours. This empowers the merchants and marketing clients to effectively elevate their operational productivity. Staying true to our long-term vision will deliver superior user experiences, build a broader platform for our platformers, and create a more sustainable value for our shareholders, collectively unlocking new growth opportunities in the AI era. That concludes my prepared remarks. Next, our CFO, Ben, will review the company's financial update for the quarter and for year 2025. Thank you, Yixuan. Hello, everyone. Looking back to the past year, we significantly progressed our AI strategy and achieved remarkable results, leveraging our advanced AI capabilities. We strengthened AcquireShare's content and commercial ecosystems, delivering high-quality growth across both our operational and financial metrics. We continue to refine our user growth and retention strategies, resulting in an average user reaching 410 million for three years. At the same time, we deepened the application of the AI large model across multiple business scenarios, delivering superior experience for our users, creators, and business partners, while further improving our operational efficiency. For the full year of 2025, total revenue grew 12.5% a year over year to $142.8 billion. Adjusting net profits reached $20.6 billion, up 16.5% a year over year, with an adjusted net margin of 14.5%. Importantly, we achieved this growth while continuing to scale our investments in AI, making steady improvements to the group's overall profitability throughout the year. Now let's take a closer look at our Q4 financial performance. Our total revenue grew 11.8% a year over the year to $39.6 billion in Q4. The increase was mainly driven by growth across online marketing services, e-commerce, and coding AI. Online marketing services revenue increased 14.5% to $23.6 billion in Q4 from $20.6 billion in the same period last year. The growth was primarily driven by AI-powered upgrades to our online marketing product solutions, which improved conversion efficiency and drove higher spending from our marketing clients. Revenue from other services, including e-commerce and clean AI business, reached $6.3 billion in Q4, up 28% from $4.9 billion in the same period last year. The increase was mainly driven by growth in e-commerce and e-commerce GMV, which boosted our e-commerce commission income. And by the continued expansion of our Collinea business, by continuously refining Collinea's financial models and developing more innovative features, we have expanded this range of applications for professional creators and driven new breakthroughs and commercialization. In Q4, our live streaming revenue was 9.7 billion RMB. We're continuing fostering a rich, healthy live streaming ecosystem. At the same time, we refined operations across our core categories, providing users with more diverse, high-quality content, leveraging AI-empowered product innovation, We also drove greater user engagement through high-quality live streaming content. Cost of revenues increased 9.2% a year over year to $17.7 billion in Q4, accounting for 44.9% of the total revenue. The increase was mainly due to higher revenue sharing costs and related taxes in line with our revenue growth. In Q4, our gross profit rose 14.1% a year over year to $21.8 billion. Gross profit margin was 55.1%, up 1.1 percentage points year over year. Turning to expenses in Q4, selling and marketing expenses were $11.4 billion, compared with $11.3 billion in the same period last year. Selling and marketing expenses declined to 28.8% of total revenue, down from 32% in Q4 last year. Reflecting the stronger effectiveness of our sales and marketing, R&D expenses increased at 20.1% a year to $4.1 billion, accounting for 10.5% of total revenue. Increase was mainly due to higher employee benefit expenses, including share-based compensation expenses and increased investments in AI. Administrative expenses were 930 million RMB compared with 8.7 million in the same period last year. Administrative expenses accounted for 2.4% of total revenue, largely flat year-over-year. Group level net profit for Q4 was 5.2 billion RMB. Group level adjusted net profit rose 16.2% year-over-year to 5.5 billion, with an adjusted net margin of 13.8%. Our balance sheet remains robust cash and cash equivalents, time deposits and interest rates and restringed cash totals 104.9 billion RMB as of December 31st, 2025. Now the cash generated from operating activities in Q4 was 7.3 billion RMB. Additionally, we actively delivered on our commitment to shareholder returns based on the market conditions. As of December 31st, we had repurchased approximately 3.12 billion Hong Kong dollars or around 56.78 million shares representing about 1.32% of our total shares outstanding for 2025. Next, I'll provide a quick overview on financial performance for the three year. For the three year of 2025, our group's total revenue reached 142.8 billion RMB, up 12.5% a year over year. This includes online marketing services revenue of 81.5 billion RMB, which rose 12.5% a year over year, revenue from our online, our live streaming business increased by 5.5% a year over year to 39.1 billion, revenue from other services, including our e-commerce business, totaled 22.2 billion, an increase of 27.6%. Percentage of your growth profit margin extended by 0.4 percentage points a year to 55% in 2025 are just not a profit for the full year of 2025 was 20.6 million up 16.5% of you with an adjusted net margin of 14.5% looking ahead, we will continue to prioritize the user needs and remain committed to investing in Ai. Leveraging our leading AI capabilities will drive further innovation across Kuaishou's content and commercial ecosystems, maintaining our core competitive edge in the rapidly evolving market, and delivering high-quality and sustainable long-term growth for the company. Here concludes our preparatory remarks. Now we can open for Q&A.

If you would like to cancel your request, you can press the power key. We will now enter the Q&A session. The first question is from Lincoln Feng from Kaohsiung. The first question comes from Lincoln Kong from Goldman Sachs.

Please go ahead and ask your question. Thank you for accepting my question, Mr. Wang. Congratulations on the success of the document in the fourth quarter. My question is about KeLin AI. Recently, we have seen that the updates of various video production models have an acceleration trend, including the release of C-DANCE 2.0 recently. Thank you, management, for taking my question and congrats on the solid fourth quarter result. My question is about clean AI. So we have seen an accelerating pace for various video generation technologies models across the industry, including CDANCE 2.0, launching recently. So what's the impact for the overall industry and to clean itself? And for 2026, what are the strategy or plans for cleaning in terms of our model capability, product upgrade, as well as monetization?

Thank you. are open-minded in terms of input and output. The choice of technology and product is relatively high. There is also a lot of room for innovation. However, we still believe that video production is far from mature in terms of technology and product. Many participants can work together to improve the industry and better meet the needs of users. Recently, the 2.0 and other video model updates have had a positive impact on the industry. At the same time as reducing the normal user creation threshold, we have improved the penetration rate of AI video production in more application scenarios, making the industry cake bigger. CDONES 2.0 supports multi-mode input technical guidelines, which is consistent with the OOI model launched in December last year. It also confirms the simplicity of the model replacement around multi-mode. KOLIN's model and product capabilities are still at the top of the world. Thank you for the question.

As we mentioned before, large video generation models are highly complex Both the input and output modalities are open-ended, which allows for considerable flexibility in technical pathways and product strategies, leaving significant room for innovation. At this stage, we believe video generation technologies and products are still far from maturity, but expansion from diverse players in the ecosystem can help accelerate industry advancement and better meet user needs. Recent accelerated updates to large video generation models, including CDAMS 2.0 and others, have brought positive momentum to the industry. While lowering the threshold for everyday users to create content, they also have increased the penetration of AI video generation across a wider range of application scenarios, effectively expanding the overall market CDANCE 2.0 adopts a multimodal input architecture which aligns with the CLING-01 model we released in December last year. Underscoring our visionary early positioning in modal iterations centered on multimodal capabilities, CLING-AI continues to maintain globally leading position in both modal and product capabilities. CLING-AI was ranked among the top video generation models by artificialanalysis.ai with exceptional benchmark scores. Regarding characteristic consistency and controllability, physical realism, and stability in complex scenarios, the Cling AI 3.0 model series demonstrates stronger performance, reinforcing Cling AI's differentiated advantages among professional creators and enterprise clients.

In recent years, Kling AI has been involved in many virtual scenes and special effects production. While ensuring high-quality output of commercial and film content, it has greatly reduced the cost of film production. This collaboration has verified the commercial value of Kling AI in top film production. It also shows that we are focused on the right direction of film production scenes. And from the income trend, Kling has maintained a very good month-to-month growth trend this year. The AIR of January this year also exceeded $300 million. We are very confident that Kling's income will achieve more than 100% of the same growth this year.

Kling AI played a key role in the production of virtual scenes and visual effects in the recent heat drama, Swords Into Plowshares, produced by film and TV. It delivered high quality and commercial grade content while significantly reducing production costs. The partnership is the primary example of Cling AI's commercial value in top-tier film and television production. It validates our focus on film and television production scenarios as the right strategic direction. In terms of revenue, Cling AI maintains a strong month-over-month growth throughout the year, reaching an annualized revenue run rate, or ARR, of over 300 million U.S. dollars in January. Based on what we are seeing now, we are confident that Klinge's revenue in 2026 will more than double.

关于Klinge的模型迭代, 过去一年,我们始终围绕统一原生多模态的方向演进。 我们在Klinge 2.0发布时首次提出多模态视觉语言, The concept of MVL is to express creativity through the combination of multi-modal information. In addition to the limit of pure text-to-text interaction, in December 2025, we launched the CLIN O1 model. We deepened the interaction structure of MVL applications, and realized the input of text, image, video, and other multi-modal videos. At the same time, we launched the CLIN 2.6 model, and realized the multi-modal output function of audio and video communication capabilities. In February this year, we launched the 3.0 model based on the all-in-one concept. In one model, we realized the input and output of multi-models. Next, we will consider expanding the number of models in the model to further improve the credibility of video production, such as the mode of action and expression. We will also work hard to solve the setting and consistency problems of complex scenarios. At the same time, at the product level, we will also steadily promote Agilent's ability to realize a full-circuit creation based on user needs to complete the synchronization and control of scene arrangement roles and scenes, as well as the automatic function of lighting and movement design.

Regarding Cling AI's model iteration, over the past year, we have consistently evolved in the unified native multimodal path When we launched Cling AI 2.0, we introduced the concept of multi-modal visual language, or MBL, which enables creative expression by combining multiple modalities and addresses the limitations of pure text interactions. With the release of the Cling 101 large model in December 2025, we advanced the MBL interaction architecture even more, enabling multi-modal inputs across text Around the same time, we launched our Cling 2.6 model for simultaneous audio and visual generation, multimodal output capabilities. In February this year, we launched the Cling AI 3.0 model series, developed under an all-in-one product framework. This series supports full multimodal input and output within a single model. Looking ahead, we plan to expand the modalities in our models to further enhance controllability and video generation, including modalities for motion and facial expressions. We will also focus on addressing the configuration and consistency challenges of complex scenarios. Meanwhile, on the product front, we will keep advancing our AI agent capabilities to enable fully automated end-to-end content creation. The goal is to empower our models to automatically plan storyboards based on user needs, ensure consistency across characters and scenarios, and simultaneously generate well-aligned audio and visuals, design lighting, visual tone, and camera movement.

总之,可灵AI将继续秉持, 让每个人都能用AI讲出好故事的愿景, 不断打磨模型和产品能力, 持续保持可灵在技术和产品商业化编写上的全球领先地位。

Overall, Cling AI remains committed to its vision of empowering everyone to craft capital-advanced stories with AI. We will continuously refine our model and product capabilities, sustaining Cling AI's global leadership in technology, product, and commercial monetization. Thank you.

Thank you, Aubrey. Your next question, please.

Shout-out to JB Morgan and Daniel Chen. The next question comes from Daniel Chen of JP Morgan. Please go ahead, Chen Tiwen.

So my question is related to the AI investment strategy. So besides the Multi-model and the video generation area which related to clean AI. What are the other segments that management thinks are worth more investment in the future? Thanks.

关于AI领域的投入方向,除了多模态视频生成领域, 我们还将在内容和商业生态场景的大模型研发应用进行持续投入, 例如生成式推荐大模型和多模态理解大模型等。 Thank you for your question.

Regarding the direction of our AI investments, beyond the multi-model video generation domains, we will continue to invest in the research and development and application of large models across our content, and commercial ecosystem scenarios, such as large generative recommendation models and large multi-model understanding models.

In the past few weeks, we have seen the promotion of large-scale models. We will also continue to explore in this direction. For example, in the online marketing recommendation system, we explore the deeper combination of the production model and the recycling system, from the optimization of the single request to the long-term value modeling. In terms of model capability, by combining LIM with more logical reasoning and world knowledge, we can break the data cycle problem in the traditional recommendation system. At the same time, we are building a set of new generation of recommended large-model recycling architecture that can be deployed in the new generation of native high-frequency.

In terms of the large generative recommendation models, over the past few quarters, we have seen significant potential for generative models in recommendation scenarios, and we will continue to explore this direction. In our online marketing recommendation system, we are exploring deeper integrations between generative models and our ranking architecture, shifting from single request optimization to long-term value modeling. In terms of model capabilities, by leveraging LLMs to introduce stronger logic, reasoning, inference, and broader world knowledge, we are attempting to break the data feedback loop problem found in traditional recognition systems. Concurrently, we are building a native, highly concurrent, and scalable next-generation ranking architecture for large recognition models. Through system design and foundational engineering upgrades, we aim to ensure that the expansion of computing power and parameter scale translate into performance improvement.

In terms of multi-modal analysis of large models, our self-proclaimed multi-modal machine-based large model can be used to quickly understand the content of the secondary power supply. In the core short screen and live broadcast scene, the large model is used for video analysis and user behavior reasoning to effectively trigger the improvement of the user market and standard indicators. In the future, we will promote AI capabilities to upgrade from single-way passive questions to long-term understanding and complex task processing, and further expand to core revenue scenarios such as online sales services and e-commerce. The development of multi-mode user-friendly smart assistants will drive greater commercial value.

In the direction of large multi-modal understanding models, our proprietary multi-modal foundational large-language QI empowers QI to its content understanding infrastructure. In core short video and live stream scenarios, QI performs video parsing and user behavior inference. effectively driving improvements in the user time span and retention metrics. Moving forward, we'll upgrade our AI capabilities from one-way passive Q&A to long-term contextual understanding and complex task processing, further expand the application to core monetization scenarios such as online marketing services and e-commerce, and develop practical intelligence assistance with multimodal interaction capabilities to drive greater commercial value.

This year, we will also explore the ability of agents in applications in other business scenarios. For example, in online marketing business scenarios, we will create multiple automated AI agents for e-commerce merchants, covering smart products, creative editing, and AI material production, smart bid and dynamic pricing, AI customer assistance, and all processes such as data analysis after bidding, to reduce the threshold of customer bid. to improve investment efficiency and cost stability. In addition, we will also explore sales agents in the search industry, to help customers improve the conversion efficiency of the search engine, and to reduce the cost of purchase. In the e-commerce scene, we will improve the search experience of users through the construction of search agents, and improve the search frequency of users. We will also explore the automatic calculation optimization of agents,

In 2026, we will also explore the application of AI agent capabilities across other various business scenarios. For example, in online marketing scenarios, we are developing an AI agent that delivers automated marketing placement for our e-commerce merchants. This covers the entire workflow from intelligent product selection, creative editing, and AI-generated materials, smart bidding, and dynamic pricing, AI customer support and post-placement data analysis, lowering the threshold for clients to place marketing materials, and improving placement performance and cost stability. Additionally, we'll also explore sales AI agents for lead-focused sectors, helping clients improve lead conversion efficiency and reduce customer acquisition costs. In e-commerce scenarios, we will improve the user search experience through the development of a search and recommendation agent, driving higher user search-based order volumes. We will also explore agent-based automated computing power optimization We will further share our progress on these fronts with you at appropriate time.

最后,我们还将推出新的算力中心建设。 算力是公司AI发展的核心根基与底层支撑。 满足公司研发迭代模型训练与推理需求, 公司将算力中心建设纳入规划以夯实AI发展算力 AIT team

Finally, we will also advance the construction of the new computing power centers. Computing power is the core foundation and underlying support of our AI development, meeting the company's demands for R&D iteration, model training, and inference enhancement. We have integrated the construction of computing centers into our strategic planning, aiming to solidify the computing foundation for AI development. By reserving expansion space to accommodate long-term needs, these centers will deeply support core tasks such as AI algorithm optimization and large model training, empowering our AI innovation with a robust computing foundation. In summary, we will continue to deeply cultivate the R&D of core AI technologies and their implementation across multiple scenarios, With firm computing investments and a deep AI talent pipeline, we will empower our content ecosystem and realize continuous growth in commercial value for our ecosystem partners. Thank you.

Thank you, operator.

Next question, please. The next question comes from Thomas Chong of Jefferies. Please go ahead.

Hi, good evening. Thanks management for taking my questions. On e-commerce, how should we think about the growth strategies in 2026? How should we think about the trend this year and the growth opportunities? Thank you.

Thank you for your question. In 2022, the big direction is still to return to the trend of fast-moving content marketing, and to play our role as the top of the content platform. From the point of view of normal strategy, we will focus on supply chain reform and continue to make good products. As we mentioned earlier, in 2025, the city government will launch the Chengfeng Project, As a targeted service plan for the leading brands, we will help the brands achieve rapid start-up and continuous growth in the fast-growing environment. In 2026, we will invest more resources in the supply chain, from the three-party traffic, product operation and service, in four dimensions, not only to surround the brands, but also to do a good job of increasing the industry range. We are determined to maintain the dynamic with 100 key industry ranges, and to join our e-commerce industry team, the small and medium-sized teams, and the rich business resources to do a good job of new business. With the growth of the e-commerce market and the current challenging e-commerce environment, the relationship between brands and platforms is reshaping. From the past simple industry positioning to the transformation of cooperative partners that grow together with merchants, we also hope to optimize the supply and demand of platforms at the same time to provide users with richer product choices. Of course, we must continue to do well in purchasing and infiltrating buyers. Thank you. Thank you. Thanks for the question.

Our broad focus for 2026 reminds on returning to the essence of Kuaishou's content-based e-commerce and on maximizing our strengths as a content platform. Our growth strategy spans across three areas. First, we'll focus on the supply side of reforms to continuously refine supply and consistently offer good products. As mentioned earlier, in Q4 2025, we launched the Voyage initiative to provide a targeted support for top-tier brands. It's designed to help them quickly achieve strong start and sustained growth within the equation ecosystem. In 2026, we will also invest more resources on the supply side, primarily across four areas, merchant traffic, product, operations, and services. Our focus will extend beyond brands to include merchants in the key industrial zones. We've already identified 100 priority industrial zones, and we are actively managing them. Meanwhile, we are working closely with our e-commerce industry team, small and medium-sized merchant team, and service providers to empower our new merchants. As the e-commerce market matures and macroeconomic dynamics remain challenging, the relationship between platforms and brands is being reshaped. Platforms are evolving from single transactional roles to collaborative partners that grow alongside merchants. As we empower merchants more effectively, we also plan to refine the platform's supply ecosystem and provide users with a wider variety of products. Second, we will continuously improve paying user acquisition and penetration. Currently, there's still significant growth potential in the number of e-commerce monthly average paying users. In 2026, we'll focus on exploring and better understanding users' interest in e-commerce content. We'll also optimize our traffic strategies and leverage effective subsidy mechanisms, and across scenario synergies to boost paying user conversion and scale growth within superior products. Third, we will further optimize resource integration. This quarter, we have seen some preliminary success in implementing traffic synergy. Moving forward, we aim to deepen integration between e-commerce and commercialization, enhance coupon synergy, improve subsidy efficiency, and optimize overall resource allocation and investment efficiency.

We believe that the natural evolution of content e-commerce will still drive its penetration rate to increase in the online marketing and sales market. In the past year, the growth of the industry of men's clothing, fresh food, and other industries has been fast. This year, these industries will also continue the growth trend of last year. According to the 26 years we just mentioned, it will increase in the supply chain, and at the same time, smart marketing tools will help to increase and reduce the cost. We believe that the inherent conversion advantages of content-based e-commerce will continue to drive its penetration in the online retail market.

Over the past year, categories such as men's sportswear and fresh food grew rapidly. We expected these verticals to maintain their growth momentum this year. As we just mentioned about ramping supply in 2026 and leveraging intelligent operational tools to help merchants reduce costs and improve efficiency while offering them a clear, more certain path for growth, we expect even more structural growth opportunities in content-based e-commerce. In 2026, we expect to quite show e-commerce to achieve steady, high-quality growth.

在高质量增长的前提下,我们会进一步夯实电商的电信能力,深化全站推广和智能化投放的技术能力。 We expect that the core increase in e-commerce sales revenue will mainly come from the first, broad scale, focusing on the key segment industry, broadening the supply chain of the industry, and expanding the e-commerce sales service. Specifically, it includes makeup, outdoor sports, freshness, plus makeup, and other industry categories. Second, increase efficiency, actively introduce brand customers, optimize customer structure, and do a good job of resource integration and unification. We also hope to improve the efficiency of goods and services in the goods and services area. On the basis of expanding the supply of full-fledged products, we will increase the number of business cards in the business card market. portfolio set.

Under the promise of high-quality growth, we will further strengthen our e-commerce monetization capabilities and deepen the foundational capabilities of our Omni platform marketing solution and smart placement product. We expect that the core incremental growth for e-commerce marketing service revenue will come from three areas. First, skill expansion. By focusing on key verticals and broadening industry supply, we aim to scale monetization through e-commerce marketing services in categories such as cosmetic sports and outdoors, fresh food, and home furnishings. Second, efficiency improvement. We will actively bring in more brand clients, optimize client composition, and integrate resources to drive aligned growth for both GMV and marketing spending. Women's apparels, And healthcare will be a particular focus where we can enhance monetization efficiency. Third, sector expansion. We will expand into sectors where we lag our competitors, such as maternal and children pad and consumer electronics, identifying clear opportunities and driving breakthroughs. In addition, we are also looking to continuously improve monetization efficiency in shelf-based e-commerce, By expanding omni-domain product supply, we can increase merchants' marketing budgets on product cards.

In short,

Guided by the e-commerce growth strategy and monetization roadmap we outlined for 2026, we will take a steady, disciplined approach. We'll focus on the right long-term initiatives while leveraging our content platform strengths to better meet the consumption needs of our users. Thank you.

Thank you.

Operator, next question, please. The next question comes from Felix Liu of UBS. Please go ahead. Please ask the question.

Good evening. Thank you for answering my question. My question is about online marketing. In the current red light and industry environment, in addition to e-commerce, what are the main opportunities for online marketing in the industry in 2026? How can we seize these opportunities?

Sorry, just let me finish the English translation.

In addition to e-commerce, what are the main advertisement industry growth opportunities in 2026, and how do we plan to capture these opportunities? Thank you.

Thank you for your question.

From a sectoral perspective, we believe the key growth opportunities this year will mainly come from three sectors, lifestyle service, comic style short plays, and AI applications.

The traditional search platform continues to migrate to the content platform. The content form of short videos and live broadcasts is easier to establish user trust, reduce decision thresholds, and improve conversion efficiency. At the same time, the social service industry is still in the process of continuous onlineization. For example, for businesses in the agricultural, education, and car industries, the online platform is gradually becoming an important marketing and customer channel. From a platform perspective, we will continue to upgrade products to help businesses communicate more efficiently with potential customers and improve the conversion rate through the optimization of Wenda capabilities. In addition, customers in the life service industry, mainly small and medium-sized businesses, have higher requirements for customer service and delivery capabilities. We use the AIGC sales material solution to help small and medium-sized businesses produce sales material at a low cost.

In the lifestyle service sector, we have seen a continued shift in user behavior from traditional search platforms towards content platforms. Short videos and live streaming formats are more effective at building user trust, reducing decision-making friction, and improving conversion efficiency. In addition, the lifestyle service sector continues to deepen its online penetration. For merchants in sectors like agriculture, materials, education, and automotive, online platforms are gradually becoming key channels for marketing and customer acquisition. At the platform level, we'll continue to upgrade our products to help merchants reach their potential customers more effectively, while enhancing our in-platform interaction capabilities to improve conversion rates. Moreover, clients in the lifestyle service sector are mostly small and medium-sized merchants that require strong customer service and operational support. Through our AIGC marketing material solutions, we help small and medium-sized merchants generate marketing materials at low cost. In addition, our AI-powered customer service solutions enable merchants to provide 24-7 online support.

In the content consumption industry, as AI technology significantly improves content production efficiency and reduces production costs, the new content form of Manjus is rapidly developing. Manjus, as a global leading video production model based on the mature short film ecosystem and global leading video production model, has two advantages. Through the integration of content and technology, Manjus has been built into a fully integrated Manjus ecosystem that covers tools, content, and publishing. In addition, we also launched a full-fledged Manjoo service plan, which covers all aspects of flow and flow, and continuously enriches the platform's Manjoo content supply, and drives the growth of Manjoo's online sales consumption. Since the second half of last year, Manjoo's sales service revenue has increased rapidly. In March this year, the daily consumption value has surpassed 15 million RMB.

In the content consumption sector, as AI technology significantly improves content production efficiency and lowers production costs, the emerging content format, comic-style short plays, is advancing rapidly. As a top-tier player, Kuaishou has the dual advantages of our mature short play ecosystem and the world-leading video generation model. By deeply integrating content and technology, we're building a comic-style short play ecosystem that spans the entire value chain from tools and content to distribution. Additionally, we introduced a full-scale comic-style short play support program covering computing power, traffic, and other resources. These continuously enrich the platform's comic-style short play content supply and boost online marketing spending in this category. Since the second half of last year, the total spending from online marketing services driven by Kuaishou's comic-style short plays has increased rapidly. In March this year, peak daily marketing spending exceeded 15 million RMB.

AI application industry is also one of the most important growth opportunities for us in 2026. With the development of AI technology, all types of AI applications continue to be limited, and the entire industry is still in a rapid stage of development. We expect that the relevant industry needs in 2026 will remain significantly increased. The AI application sector is also what we view as another key growth driver for 2026.

As AI technology continues to advance and new applications emerge, the industry remains in a rapid growth phase. We expect the demand in these relevant sectors to continue growing significantly in 2026. Against this backdrop, we will strengthen and refine operations for our clients, continuously optimize short and long-term retention metrics, helping clients maximize the user lifetime value, all of which will prompt AI application clients to increase both their marketing spending scale and commitment on our platform.

总之呢,在汪2026年,我们将通过产品升级、内容生态建设以及重点行业客户运营, 更好的挖掘生活服务、漫剧及AI应用等行业增量机会实现线上营销服务收入的稳健增长。

In summary, for 2026, harnessing our product upgrade, content ecosystem development, and refine our operations for our clients in priority sectors, we aim to better capture incremental growth opportunities in the lifestyle service sector, comic-style short plays, and AI applications, driving solid growth in our online marketing services revenue. Thank you.

Thank you, Albert. Last question, please.

The last question is from Liao Yuan from Citix. The next question comes from Yuan Liao from Citix. Please go ahead.

Good morning, Mr. Liao. Good morning, Mr. Liao. Good evening, Mr. Matthew. Thank you for accepting my question. In this conference, we have mentioned many times the construction of a computing center. My question is, Thanks management for taking my questions. You have repeatedly mentioned the construction of computing power centers. So my question is, could management share your plan scale of AI-related CapEx in 2026, and the key area of your investment. So how will these CapEx investment affect your overall profit margin? Thank you.

Thank you for your question. As Mr. Yi said, in the past year, we have fully deepened our AI strategy. Multi-modal video generation model enables AI to upgrade its technology, Thanks for the question.

As Yixiao said, over the past year, we have fully deepened our AI strategy. Our multi-model large video generation model, Cling AI, has achieved impressive results in technology advancement, product duration, and commercial monetization. At the same time, AI has delivered strong value in empowering our content and commercial ecosystems, reinforcing our commitment and confidence to continue investing in AI.

In 2026, we expect the total capital investment of the company to reach about 2.6 billion RMB, which is about 1.1 billion RMB in comparison with the increase in the year 2025. These investments include the calculation investment on the Krillin model and other basic models, as well as the storage, processing, as well as data computing center construction and project investment. The improvement of KBX is due to the increase in the number of users and revenue size of KBX. On the other hand, this year we will also carry out an important overlap upgrade of KBX. Accordingly, we also need to increase the reserves on training computing. In the future, as the model overlaps,

In 2026, we expected the group's total compacts to reach approximately 26 billion RMB, an increase of about 11 billion compared with 2025. This covers computing resources for clean AI's large models and other foundational models, as well as routine server procurements, such as offline data storage and processing, and investments in data and computing center infrastructure. The increase in CapEx for Cling AI's large models is partly due to higher inference computing needs from our expanding user base and revenue scale. It also takes into consideration of our major Cling AI model upgrades scheduled for the year, which require additional investment in training computing power. With the advancement of model iteration in the future, we will also flexibly allocate computing resources between inference and training to maximize the efficiency of computing resource utilization.

Here, I would like to emphasize that we are a company that takes great care of cash flow management and maintains an abundant cash reserve. In 2025, although we have invested about 1.5 billion RMB in CapEx, There are still nearly 1.2 billion RMB of free cash flow at the group level throughout the year. Although CapEx investment has increased in 2026, our goal is to continue to maintain the positive free cash flow at the group level throughout the year. We believe that every investment at the moment will transform efficiency into future energy mobilization. While adhering to long-term technical investment, we also uphold financial discipline and maintain abundant cash reserves. to support the growth of high-quality assets in the AI era.

I would also like to emphasize that we are highly focused on cash flow management and maintaining ample cash reserves. In 2025, despite approximately 15 billion in capex, the group delivered nearly 12 billion RMB in free cash inflow for the year. For 2026, even with increased capex, we aim to continue maintaining positive free cash flow at the group level for the three year. We believe that every investment today will efficiently translate into future profit drivers. As we stay focused on long-term technology investments, we will maintain disciplined financial management and ample cash reserves. Our robust balance sheet will empower the group's sustainable, high-quality growth in the AI era. Thank you.

Let's conclude the Q&A session. Thank you, Operator.

Thank you, Operator.

Thank you for your participation. If you have any questions, Thank you once again for joining us today.

If you have any further questions, please contact our capital market and IR team at any time. Thank you.