Kuaishou Technology B

Good day, ladies and gentlemen. Thank you for standing by. Welcome to the Kuaishou Technology fourth quarter and full year 2024 financial results conference call. Please note that English simultaneous interpretation will be provided for the management's prepared remarks. This English line will be in listen-only mode. I will now turn the call over to Mr. Matthew Zhao, VP of Capital Market and Investor Relations at Kuaishou Technology. Thank you, operator. Good evening and good morning to everyone. Welcome to our fourth quarter and full year 2024 financial results conference call. Joining us today are Mr. Cheng Yi Xiao, co-founder, chairman, and CEO. Mr. Jin Bing, chief financial officer. Before we start, please know 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 recorded by law. For all important information about this call, including forward-looking statements, please refer to the company's public information on the fourth quarter and three-year 2024 results announcement ended on December 31, 2024, 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 definition of non-FRS financial measures and reconciliation of AFRS to non-FRS financial results and the related risk factors, please refer to our fourth quarter and full year 2024 results announcement. For today's call, management will use Chinese as the main language. A third-party interpreter will provide simultaneous English interpretation in the prepared remarks session and a consecutive interpretation during the Q&A session. Please know that English and interpretation is for convenience purposes only. In the case of any discrepancy, management statements in the original language will prevail. Lastly, unless otherwise stated, all currency units mentioned are in RMB. Now, I'll turn the call over to Yixiao. Hello, everyone. Welcome to Kuaishou's fourth quarter and full year 2024 earnings conference call. Over the past year, guided by our technology, Improving user-centric business philosophy, we leveraged AI technology to empower content and business ecosystem. This created greater value for both our users and partners and drove our robust financial performance. In Q4, our average DAUs remain above $400 million, and our total revenue grew by 8.7% year-over-year to $35.4 billion RMB. Our adjusted net profit reached a new quarterly high in Q4 of $4.7 billion RMB. For the 2024 full year, our total revenue increased by 11.8% year-over-year to 126.9 billion RMB, and our adjusted net profit increased by 72.5% year-over-year to 17.7 billion RMB, with an adjusted net margin of 14%. These gains reflect our steady improvement in profitability. Today, it has become increasingly evident that with the ongoing advancements and breakthroughs in large models and application capabilities, AI is pushing the boundaries of video content creation, user experience, and the broader business ecosystem. As a leading short video platform and content community in China and globally, Kuaishou stands at the forefront of this critical intersection of AI technology and large video models. While we are driving transformative changes across the industry landscape, we firmly believe that AI is much more than an efficiency improvement tool. It is the core engine that will create greater value for our business ecosystem while continuing to drive the platform's traffic growth. Importantly, our work proactively engages in and advances broader industries to prevent a transition. Next, I'd like to share some of our thoughts on our go-forward AI strategy. As the predominant content consumption for global users, short videos have been deeply embedded in the fabric of users' daily lives in every industry. Users have developed an increasingly strong demand for return and more diverse top-notch content. With the advent of large video models, we can greatly lower the entry barrier to the video creation, particularly the cost barriers. Users can now generate premium videos with a simple prompt or picture. This positions us to unlock the massive creative potential for non-professional creators. We believe Culling is becoming the new infrastructure for video creation in the AI era. Right now, it's a powerful resource for PGC and PGC creators that help them create high-quality videos and images efficiently. Next, Culling AI will evolve into a visual storytelling tool for amateur creators. With cinema-grade video generation capabilities, it empowers people to craft compelling visual stories and become a gen-AI-based director, forming an AIGC-driven community for content consumption and interaction, while creating a large-scale application in commercial scenarios. Moving forward, by empowering through Cling, we hope to transform and upgrade existing business while creating a new track for AI-driven video content production. By doing so, we'll strengthen and elevate our position as a leading platform for short video production and consumption community. We unveiled Kling, a large feature generation model, last June. After its launch, it quickly gained significant traction with video creators and artists globally, attracting more than one million applicants in just one month. Encouraged by users' positive response, the Kling team swiftly improved the model, releasing new iterations approximately every two weeks. At the end of Q4 last year, we rolled out Kling AI 1.6. This latest version features enhanced responsiveness to text descriptions such as motion, temporal actions, and camera movements. It also significantly improves the visual quality, enhancing style consistency, color accuracy, lighting dynamics, and rendering detail, keeping us in the lead in terms of global technological advantages. Kling AI 1.5 Pro is currently one of the top two larger video models in the world. according to the well-known AI evaluation website, artificialanalysis.ai, ranking alongside Google View 2, far ahead of other domestic video generation large models. In Q4 of 2024, we officially launched the standalone Cling AI app, giving global users multiple ways to access it, including through both the app and the web. With the continuous improvements in Cling AI's functionality, its user base has grown at an accelerated pace, while Cling AI's commercialization has also gained momentum, thanks to access channels such as membership subscriptions for individual users, and API interface solutions for merchants. As of February, since the Cling AI's monetization, we reached an important milestone of more than 100 million RMB in cumulative revenue from Cling AI. We expect that Cling AI will achieve another leap in revenue in 2025. For the next stage, we'll continue optimizing and iterating AI Cling AI's technological capabilities through dedicated research and development efforts and a steadfast investment in resources to ensure Cling large model maintain its industry-leading position. As Cling AI continues to advance, we firmly believe AI technology and its applications are creating value for our partners on the platform, especially for our marketing clients and e-commerce merchants. As such, we will upgrade our AI algorithms and underlying recommendation technologies by closely following and learning to apply the technological progress of DeepSeq and other global foundation models. This will continually amplify the matching efficiency between users, content merchandise, and improve conversion, leading to a higher ROI for our marketing clients. We'll also gradually upgrade our magnet engine into a next-generation AI-powered intelligent commercial engine. Through a large multimodal model, we generated a more creative and user-friendly marketing material content for our marketing clients dramatically reducing their content production costs and freeing up more of their budget to focus on brand promotion and traffic conversion. Meanwhile, AI-generated virtual digital human live streaming solution will lower the marketing entry barrier and help clients reach breakthroughs in sales. In Q4, both average daily spending on AIGC short video marketing materials and virtual digital human live streaming solutions exceeded 30 million RMB, with a daily peak of 60 million RMB. Moreover, we have 45-hour model estimation capabilities by using large models to analyze product systematics and user behavior, as well as integrating model architectures and sequence representations. As such, we have boosted advertising performance of marketing materials, driving more marketing spending. AI recommendation technology also offers us an opportunity to redefine the content-based e-commerce paradigm. With our deep understanding of exceptional content and superior products, We are shifting from a model where people search for products to one where products intuitively understand and find the right people. For consumers, we have validated the sales funnel of our AI try-on tool in the KLR live streaming rule, making personalized AI shopping easier and more interactive. For small and medium-sized streamers and merchants, we launched the Smart Presentation tool, which makes intelligent product augmentations and generates sales pitches, among other smart features, strengthening live streamers and emergent e-commerce marketing capabilities. The era of AI is happening right now. As short of a video platform and content community, we are determined that AI is not a replacement for human creativity, innovation, or creation. We see it as a tool to help more people in these endeavors. AI will allow everyday users to break industry and culture barriers, giving them opportunity to become globally loved artists and entrepreneurs through equal access to creative video generation productivity tools. With our enduring commitment to inclusivity, we have always strived to create a community where everyday people can showcase their talents and share their lives. We recognize this transition requires time and patience. Over the next three-plus years, we plan to increase our capital expenditures on R&D investments, which may impact the speed of our margin expansion. That said, I'm convinced that of the seeds we plant today, we will shape the entire video production and consumption industry over the next decade and firmly establish Kuaishou as the inclusive digital home in the AI era. While we set our sights on the stars and look forward to the future, we remain grounded in making steady incremental progress. With that overview of our vision, I'd like to briefly review our key business highlights and developments in Q4 and for year 2024. First, user growth and ecosystem construction, Q4. Average DEUs on the Kuaishu app reached 401 million, and AMAs used to reach 736 million, increasing by 4.8% and 5% year-over-year respectively. The average daily time spent per DAU on the Questio app was 125.6 minutes. While total user time spent rose by 5.8% year-over-year, our refined user growth strategy led to consistent improvement in new users' engagement, interaction, and retention rates. We focused on optimizing features and improved the end-to-end viewing experience for users, increasing the resolution and ensuring smooth streaming. We have continuously enhanced the video sharing and communication experience while introducing various innovative features in private messaging, which drove the daily average private messaging penetration rate up by nearly 5 percentage points over year in Q4 among users with mutual followers. We also refined our common ranking strategy, resulting in year-over-year increase of over 40% in user time spent on common features in Q4. Content is cornerstone of our dynamic community. Understanding why users open our app guides our content strategy. In terms of traffic distribution, we ensure that the standout and high-quality content featuring distinctive EquatIO characters reaches the right audience through more recommendations and higher exposure, forcing deeper engagement and stronger connections. For example, content creator dubbing live streaming room emerged as a welcoming space for EquatIO users to share their stories and interact with one another, strengthening the heartfelt bonds between users and content creator. In a pan-knowledge vertical, on one changing opera theoretic truth offline tools and online live streaming sparked a widespread user engagement discussions the troops new year eve performance achieved over 140 million cumulative views across live streaming and short views this fostered a connection between regional users and creators both online and offline and strengthen the core user sickness while promoting china's traditional art and intent for cultural heritage six online marketing services in q4 revenue from online marketing services grew by 13. 3% year-over-year to 20.6 billion RMB. This marked the first time the segment's quarterly revenue exceeded 20 billion RMB. In addition, for 2024, its revenue grew by over 20% year-over-year. Our quarterly revenue growth was mainly driven by a year-over-year ECPM increase in the high single digits. We captured incremental opportunities, such as commercialized workplace in the online marketing services and enhanced our marketing service recommendation models through AI. This improved our ability to predict accurate outcomes and increase the conversion efficiency of marketing materials. Additionally, smart marketing solutions such as the U.S. Placement Solutions and Omni platform marketing solutions significantly enhance our marketing clients' advertising performance. In Q4, external marketing services continue to be the primary driver of our online marketing services. In particular, the content consumption industry, which included short plays, mini games, and novels, experienced faster growth. Notably, Marketing spending from commercialized to short plays to search the more than three-fold year-over-year in Q4. On the product front, we upgraded to the UAS placement of solutions, transitioning from rule-based to model-based decision-making. As a result, UAS-based marketing spending accounted for over 55% of total external marketing spending in Q4. Closed-loop marketing services continue to support merchants in leveraging high-quality traffic on question on boosted operating efficiencies. In Q4, total e-commerce marketing spending by merchants using our Omni platform marketing solutions and a smart hosting product contributed approximately 55% of total closed-loop marketing spending. By focusing on enhancing small and medium-sized merchants' willingness for marketing placement on our platform and improving advertising performance, we drove a year-over-year increase of 30% in these merchants' marketing spend. Third, our e-commerce business. Leveraging our advantages in content-based scenarios and pen-shelf-based e-commerce, our e-commerce GMV grew by 14.4% year-over-year to 462.1 billion RMB in Q4. The more abundant e-commerce offerings and enhanced synergy efficiency of our Omni platform traffic have enabled us to better meet the needs of our e-commerce users in Q4. The number of e-commerce monthly active paying users increased by 10% a year over year to 143 million, with an MAU penetration rate of 19.5%. We also launched targeted programs to acquire new users from South and China and enhance their activity while harnessing key promotional events, refining coupons, and specific strategies. During the 2011 sales promotion, we gained over 7 million net new users for the fortifying users loyalty for repeated purchases going forward we'll continue to uphold our user-centric strategy and partner with merchants and accounts to optimize the consumer shopping experience on the merchant side in q4 merchants continue to thrive in quite short e-commerce ecosystem the number of average monthly active merchants increasing by over 25 percent a year-over-year gmv small from small and medium-sized merchants largely grew year-over-year in q4 mainly driven by our three core policies, namely improving policies for new merchant recruitment, optimizing policies for existing merchants, and leveraging diversified scenarios. To encourage new merchants to use Kuaishou, we launched the Golden Bounty Initiative and Set Sail Initiative and provided code star traffic through targeted scenarios. These programs helped early-stage merchants increase traffic, reduce operating costs, and align incentives to their key growth cycles and transition points. We worked with ecosystem partners to accelerate new merchants' growth providing small and medium-sized merchants with refined methodologies for economy-based e-commerce and establishing growth paths for merchants in many domains and areas, including KOL, live streaming and distribution, and shopping malls. In our KOL business, we strengthen the platform's merchandise management capabilities through our Blockbuster initiative and broaden the KOL's product offerings during sales promotions. Meanwhile, we further energized our quantum-based scenarios through diverse activities and marketing tools, including KOL competition to motivate streamers. During the 111 sales promotion, more than 39 million users joined group buy for KOL followers and over 2,500 live streaming rooms achieving GMV exceeding 1 million army. In terms of diversified scenarios in Q4, short video e-commerce GMV grew by over 50% year-over-year, As important components to our content-based scenarios, both short video e-commerce and interaction between short videos and live streaming has been incremental in helping merchants in Kowalski expand their business. Additionally, PennShelter-based e-commerce, GMV, contributed 30% of our total e-commerce GMV in Q4. Its growth consistently outperformed overall GMV growth, mainly driven by strong supply and demand. In Q4, average daily Active merchants grew by over 50% and average daily paying users in our shopping mall grew by nearly 40% year-over-year. As our pen-shelfed e-commerce increasingly complement our current basis narrows, we have enhanced merchants' Omni domain operation efficiency by strengthening the platform's control over blockbuster products. Next, regarding our live streaming business. Q4 revenue from our live streaming business was $9.8 billion RMB with a year-over-year decline continued to narrow compared to the previous quarter. We are determined about building a healthy and sustainable live streaming ecosystem for the long term and achieving diversified growth propelled by high-quality content. By the end of Q4, the number of our partner talent agencies grew by over 30%. The number of talent agency-managed streamers increased by over 60%, both on a year-over-year basis. On the supply side, leading categories continue to create value, such as multi-host live streaming, group live streaming, and grand stage. In addition, by expanding user engagement on grand stage in rural towns, we accelerated our ability to discover and support local, small, and medium-sized streamers. In terms of gaming live streaming in Q4, we explored comprehensive collaborations with key games, including Game for Peace, Crossfire, in areas such as streamer growth, content crowd creation, and event promotion while deepening our expertise in fighting games and other niche verticals. In addition, as our live streaming plus services empowered traditional industries in Q4, the average daily number of users submitting resumes on QuietHire increased by over 100% year-over-year, number of matches grew by over 270% a year over year. For ideal housing, daily lead generation surged by over 260% compared over the same period last year. Finally, our overseas business and local services progress. Regarding our overseas business in Q4, we're deeply rooted in Brazil, where we'll continue investing in local content operations and brand marketing. We achieved breakthroughs in innovative user acquisition channels, increasing DA used by 9.3% of year-over-year in Brazil. Benefiting from optimized traffic distribution mechanism and cooperation with the top tier local IP resources, we have gradually built a rich, diversified content ecosystem with a steadily growing user activity. The average daily time spent per year-over-year and quarter-over-quarter exceeding 75 minutes Thanks to these improvements, our total overseas revenue maintained rapid growth of 52.9% a year over the year in Q4. Notably, online marketing revenue increased by 83.5% a year over the year. As a result of our effective control over costs and expenses, the operating loss from our overseas business narrowed by 57.2% a year per year. In addition, we have initially validated our e-commerce business model in Brazil, achieving consistent growth in order of volume with improved subsidy and operation efficiencies. These early success have unlocked the potential for healthy, sustainable development in the Brazilian market. Now, our local services. GMV for local service is more than double the year-over-year in Q4. We focus on city clusters with strong user bases and consistently focus on our user needs by further optimizing price consumption capability and comparison capabilities and snower applications. These efforts enhanced our compelling value for money consumption experience, driving a 52.4% year-over-year increase in average monthly paying users in Q4. We also worked on improving product quality and optimizing the user experience. which steadily increased the conversion efficiency. Monetization also improved, with revenue from local services growing by 2.6 times year-over-year in Q4. As we further optimized the infrastructure of our commercialization products, we also strengthened our partnerships with more high-quality local operations and leader-based merchants by leveraging our differentiated traffic resources, empowering merchants to achieve incremental growth. At the same time, in pursuing higher ROI for our healthy and sustainable growth in local services, we amplified our subsidy and operational efficiencies and continue narrowing our operating loss year-over-year in Q4. Looking back over the past year, despite the numerous challenges, we achieved a systemic growth alongside our ecosystem partners by relying on the strengths of our robust and thriving content ecosystem, ever-improving structure, and expanding commercial scenarios. At this new era of AI technology and folks, we remain committed to advancing our AI strategy and remain dedicated to our technology-driven, user-centric business philosophy, staying deeply attuned to users' needs, continually expanding our content offerings and fostering our AI content and business ecosystem built upon a trust-based community to create a long-term value for our partners and shareholders. My prepared remarks end here. Now I'll hand the call over to Mr. Jinbing to introduce the company's financial update for the full year 2024 and Q4 last year. Thank you, Yixiao, and hello, everyone. Looking back to the past year, we achieved new breakthroughs in our operating metrics and steady improvements in our financial performance despite uncertainties in the external environment. We maintained a strong momentum by leveraging our robust content and healthy and sustainable business ecosystem and consistently enhancing the synergy efficiency of our Omni platform traffic. In the second half of 2024, Our average DA use surpassed 100 million, marking a significant milestone in our user ecosystem. As an AI-driven tech company, we proactively explored and deepened the application of large AI models in various business hours, further elevating our content and business system while driving greater operating efficiency. For 2024, our total revenue reached 126.9 billion RMB, increasing 11.8% year-over-year. Our adjusted net profit jumped by 72.5% a year over year to 17.7 billion RMB with an adjusted net margin of 14%, reflecting rapid improvement in profitability. Now, let's take a closer look at our Q4 financial performance. In Q4, our total revenue grew by 8.7% a year over year to 35.4 billion RMB, mainly driven by growth in our online marketing services and e-commerce business. In Q4, online marketing services revenue increased by 13.3% to 20.6 billion RMB from 18.2 billion RMB in Q4 last year and accounted for 58.3% of total revenue. The quarterly revenue surpassed 20 billion RMB for the first time. contributing to our full-year online marketing services revenue growth of over 20% year-over-year. This growth was mainly driven by our continuous enhancements to algorithms and models, as well as our smart placement products. Both upgraded our content and user understanding using a technology for the improving recommendations and a conversion efficiency, which promoted more marketing spending. Revenue from other services, including e-commerce, reached 4.9 billion RMB in Q4, up 14.1% from 4.3 billion RMB in the same period last year. This increase was mainly driven by the growth in e-commerce GMV, which boosted e-commerce commission income. We further advanced our Omni domain operations to expand merchants and chaos business by leveraging our rich content ecosystem and continuously improving the synergy efficiency for our Omni domain traffic. Meanwhile, We further provided abundant product offering with more diverse marketing activities to constantly elevate a user shopping experience, fortifying our e-commerce user loyalty for repeat purchase. All these measures lead to increases in the number of e-commerce monthly active paying users and monthly active merchants. Q4, our live streaming revenue was 9.8 billion RMB, a decrease of 2% from 10 billion RMB in Q4 last year. With the year-over-year decline continue to narrow sequentially, we continue to foster a healthy and sustainable live streaming ecosystem by consistently developing diverse leading categories in live streaming, deepening our expertise in niche verticals, accelerating the ability to discover and support local small and medium-sized streamers. We effectively increase the number of streamers and talent agencies and their engagement levels. Cost of revenues went up by 6.5% of year-over-year in Q4 to 16.3 billion RMB and accounted for 46% of total revenue. This increase was mainly due to increased revenue sharing costs and a related tax in line with our revenue growth, depreciation of property and equipment, and right-of-use assets and amortization of intangible assets. In Q4, gross profit grew by 10.6% year-over-year to 19.1 billion RMB. Gross profit margin was 54%, an increase of 8.9 percentage points year-over-year. Moving to expenses, selling and the marketing expenses increased by 11% to 11.3 billion RMB, accounting for 32% of total revenue. The increase was mainly due to increased spending on online marketing services and e-commerce business promotions. R&D expenses were 3.5 billion RMB, rising by 4.7% year-over-year, accounting for 9.8% of total revenue, dropping from 10.1% in Q4 last year. Administrative expenses increased by 15.2% year-over-year to 866 million RMB. As a percent of total revenue, administrative expenses were flat with the same period of last year. The increase in R&D and administrative expenses was mainly due to higher employee benefit expenses, including related share-based compensation expenses Group level net profit for Q4 rose by 10% year-over-year to $4 billion RMB. Group level just net profit rose 7.8% year-over-year to $4.7 billion, reaching a new quarterly high with an adjusted net margin of 13.3% year-over-year. Our balance sheet remains robust with cash and cash equivalents, time deposits, restricted cash, and wealth management products totaling 92.8 billion RMB as of December 31st, 2024. Through enhanced monetization capabilities and efficient working capital management, we generated positive operating cash flow of 8.6 billion RMB in Q4. Additionally, we actively implemented our shareholder return program throughout 2024. Within year 2024, we had repurchased an aggregate of a promisely 5.6 for 6 billion Hong Kong dollars in shares, or around 123 million shares, which accounted for about 2.8% of our total shares outstanding at the beginning of 2024. Next, I'll provide a quick overview of our financial performance for 2024. For the full year 2024, our group's total revenue reached 126.9 billion RMB, up 11.8% year-over-year. This includes online marketing services revenue of 72.4%, billion RMB, which rose 20.1% a year-over-year. Revenue from our live streaming business decreased by 5.1% a year-over-year to 37.1 billion RMB, which is better than we expected at the start of the year. Revenue from other services, including our e-commerce business, totaled 17.4 billion RMB, an increase of 23.4% a year-over-year. Gross profit margin expanded by 4 percentage points year-over-year to 54.6% in 2024. Our just net profit for the year was $17.7 billion, increasing by a remarkable 72.5% year-over-year with an adjusted margin of 14%. Looking at how we'll continue to adhere to our technology-driven, user-centric business philosophy under this approach, we prioritize the user needs and actively promote the development of thriving content and business system. At the same time, we'll continue to invest in decisively by reinforcing the critical role of AI technology to maintain our competitive edge in the ever-changing market. These efforts will unearth future growth potentials and drive the long-term sustainable development for our business and financial performance. This concludes our prepared remarks. Operator, now please open the call for questions.

As a reminder, to ask questions, please press star 1 and wait for a name to be announced. The first question is from Felix Liu from UBS. The first question comes from the line of Felix Liu from UBS. Please go ahead.

谢谢管理层接受我的提问。 我的问题是关于AI。 公司的视频大模型可灵AI最近获得了各界的广泛关注, 也包括了三月初两会的新闻发布会。 Thank you, management, for taking my question. My question is on your AI progress. We noticed that the company's large video model claim has attracted widespread attention. It was even mentioned at the press conference of the MPC and CPPCC in earlier March. Commandment, please share the reason for Cling's success and your plans to maintain the model's leading edge. Thank you.

Cling's question. First of all, I would like to thank everyone for your attention and support for Cling AI. Cling AI is a project that started its internal optimization from the beginning of last year. Since the first open test was launched on June 6th last year, the product has experienced more than 20 versions in a row. Last year, on December 19th, we officially released the CLIN 1.6 model, which has significantly improved in terms of image and motion quality, as well as supporting standard and high-quality mode. In terms of tracking, the CLIN 1.6 model is better at describing text such as movement, continuous movement, and movement, and can better understand the content of the hand and generate elements that are not included. In terms of image beauty, the style is maintained better, the color is more beautiful, the lighting is more reasonable, and the details are more realistic. In terms of dynamic quality, the rationality of the movement has been improved, and the facial expressions of the characters are more natural. Especially the Kling 1.6 model, the overall effect of the 1.5 model improved by nearly 200%. We are confident that the comprehensive effect of Kling i's image video function is in the first place in the world.

Thank you for question. First of all, I'd like to thank you all for interest in support for clean AI clean AI started out as an internally integrated project early last year. It opened for testing on June six last year and quickly iterated more than 20 versions since its launch on December 19 2024, we officially roll out the Cling AI 1.6 model. This significantly improves semantic adherence, visual aesthetics, and emotion quality while supporting both standard and high-quality modes. In terms of semantic adherence, Cling AI 1.6 model has enhanced the responsiveness to text descriptions, with better motion, temporal actions, and camera movements, as well as better understanding of first frame content and can generate elements not included in it. This model also boosts better visual aesthetics with a stronger style consistency, more appealing colors, better lighting dynamics, and more vivid details. The Cling AI 1.6 model has improved motion quality, featuring smooth motion that makes expressions more natural. In particular, in our internal evaluation, the overall performance of the Cling AI 1.6 model improved image-to-video capabilities by almost 200% from the Cling AI 1.5 model. This gives us confidence that Cling AI's comprehensive image-to-video performance is currently number one in the world.

The reason why Cling AI can achieve and maintain a global leading position It is closely related to our multi-modal large-modal development, focused and firm overall strategy. Especially when it comes to organizational coordination, algorithmic investment, algorithmic improvement, and data accumulation, we have quickly established the industry advantage of CLEAN AI. In terms of organization, Quaishou has always been a company that pays great attention to the technical team. Our algorithmic and video understanding areas are also rich in talent, and the cooperation level of each team is very high. In terms of computing, we will focus more computing resources on the development of the video model. Concentrated resources focus on breakthroughs. In terms of computing improvement, the selection combination of video production models is more complex. It requires more accurate technical judgment, system-based selection, and a lot of computing innovation. We have been at the forefront of computing innovation and have continued to improve our ability to break through. Finally, in terms of data accumulation, we also formed a very obvious technical barrier in terms of private data processing and synthesis technology.

The reason why Cling AI has been able to achieve and maintain its global leadership is closely tied to our sharply focused strategy for researching and developing large multimodal models. In particular, we quickly established Cling AI's industry advantages through organizational collaboration, investment in computing power, algorithm improvement, and data accumulation. In organizational collaboration, we have always ascribed great value to our technical teams We have a deep talent pool for developing algorithms and video understanding with a high level of collaboration across our teams. In terms of computing power, we put more computing resources on developing large video models to achieve key breakthroughs. On the algorithm front, the high complexity of video generation model selection requires more precise technical judgment, semantic selection, and a large amount of algorithm innovations We have been at the forefront of algorithm innovation, solidifying our ability for further technological breakthroughs. Lastly, in data accumulation, we built a strong technical mode in video data processing and synthetic data technology.

关于未来如何继续保持模型的先进性, 我们将继续坚持聚焦策略, 并给予可领发展更多的支持。 We will continue to invest in the technical development of the KOLIN video model and through the talent gathering effect brought by the KOLIN brand, combine the computing power, algorithm and data, etc. We are confident to continue to maintain the video generation model at the world's leading technical advantage.

As for how we maintain our leading edge with our models going forward, we'll continue to use our focus strategy and provide even more support to Cling AI. We'll also continue to make R&D investments in Cling AI's large video model technology and leverage the Cling AI brand to attract top talent. By strengthening and iterating our efforts across areas in computing power, algorithms, and data, we're confident we can sustain our global technological leadership in video generation models. Thank you.

Thank you.

Next question, please, operator. The next question comes from Lincoln Kong from Goldman Sachs. Please go ahead.

Thank you, Mr. Lee and Matthew. Thank you, Benjamin, for taking my question. My question is also about AI. I think management mentioned in the preparation remarks companies' longer-term AI strategy. Could the management break this strategy down in terms of action plan and milestone targets over the next few years? Thank you.

This is your question. In early 2023, we launched a new AI strategy and set a new strategic goal. to ensure that快手 is at the forefront of possible AI breakthroughs. We believe that in the development and gradual growth of AI technology, the market will inevitably have changes in the cold and hot cycle. For快手, the best choice is to rely on AI to upgrade existing businesses, quickly form a development, investment and profit cycle, so that快手AI can have the ability to go through the market cycle. At the same time, we will continue to break through the breakthroughs in technology innovation by the large model team of the conscience mechanism, and maintain the leading advantage of large model technology. In the long term, we hope to transform the upgrade to the current advantage through CleanAI, while creating a new track for AI video production content, and further upgrade QuickShop as the leading position of short-range production and consumption community platforms.

Thank you for the question. We kicked off our new AI strategy at the beginning of 2023 and made clear strategic goals to ensure we can maintain our leadership amid potential AI breakthroughs. As AI technology develops and gradually matures, we believe the market is bound to see volatility with up and down cycles. For Kuaishou, the best option is to use AI to upgrade our existing business. and rapidly form a positive cycle between R&D investments and returns. This will be key to ensuring our AI products can truly be equipped to transcend market cycles. At the same time, we're focused on building a strong system that drives ongoing innovation in our large model team to make more innovative breakthroughs and keep our leading edge in large model technology. In the long term, with CleanAI, we aim to transform and upgrade our existing business while pioneering a new frontier in AI-generated video-quantity production. This will help us strengthen and grow our platform's leadership position for short video production and consumption community.

When it comes to this long-term strategy, we think we need to continue to work on the following aspects. First, continue to polish the underlying technical capabilities of large-scale models and multi-modal large-scale models, and continue to realize the innovation of algorithmic technology. In terms of large-scale model research and development, we started in the third quarter of last year to turn the center of the research into a fast M-OE model with hundreds of millions of parameters. While maintaining the comprehensive performance of the model, we have greatly reduced the cost of training and promotion of the model, and can better match the existing business scenarios. With this long-term strategy in mind, we think there are a few key areas where we need to continue making progress.

First, we'll continue to refine our large language models and large multimodal models underlying technological capabilities. staying in sync with industry development and driving continuous innovations in algorithm. In large language model R&D, we shifted our focus in Q3 of last year to the KUAI LLM MOE model, which has smaller parameters. MOE model helped us maintain our model's overall performance and better suited to our existing business scenarios, while significantly reduced training and inference costs. This year, since DeepSeq R1 large model open sourced, We are even more confident that we can closely track and learn from the most advanced foundation models in the market to consistently improve the performance of our foundation models.

其次,我們的計劃不斷升級以可靈為代表的AI三頁化並線模式, 並實現營收規模的快速增長, 尤其是力增可靈AI早日成為營收規模全球第一的四品生產AI應用, Whether it is to face the image and video tools of professional creators, or to face the visual narrative tools of ordinary creators, we all believe that the powerful technology and product capabilities of CleanAI can greatly reduce the threshold of creators, stimulate the creation, consumption, production, and interaction of AIGC content. At the same time, we will also use the existing content and commercial ecology of AI ability, such as Through the improvement of the content understanding and recommendation, we will bring about an increase in the size and activity of community users. We will also expand our AI capabilities to advertisers and e-commerce merchants on the platform, providing them with multiple capabilities including video marketing materials, digital live broadcasts, AI SE, and smart prizes, and increase business efficiency. We also look forward to these benefits for business ecological partners,

Second, we will continuously upgrade the AI monetization model represented by Cling to drive rapid revenue growth with the goal of making Cling AI the world's top-grossing video generation AI application. Whether it's a tool for professional creators to produce high-quality images and videos or a video storytelling tool for amateur creators, We believe Culling AI's powerful technology and product capability can significantly lower the entry barrier for content creators in driving the production, consumption, and interaction of the IGC content. At the same time, we are integrating AI across Quasar's content and business ecosystem. For instance, by improving content understanding and recommendations, we can expand our user base and increase engagement. We will also make AI capabilities more widely available to advertisers and e-commerce merchants on our platform, providing them with expert tools in various areas, including AI-generated video marketing material, digital human live streaming, AI try-ons, and smart presentation. These innovations will help our business ecosystem partners operate more efficiently and unlock new growth for us in the online marketing services and e-commerce business.

Finally, with the further advancement of AI large-scale technology, content creation thresholds and costs will also be drastically reduced, and AIGC content production quantity and quality will explode and grow. It is possible to form a new business model to open new sustainable growth channels for company development.

Ultimately, as we advance our AI models, the cost and complexity of content creation will drastically decrease, leading to a surge of AIGC-generated content in volume and quality. This could potentially lead to new business models opening up new sustainable growth channels for our company. Thank you.

Thank you, operator. Next question, please.

The next question is from Brian Gong from Citi. The next question is Kai from the line of Brian Gong from Citi. Please go ahead.

谢谢管理层接受我的提问。 我跟进一下可灵AI商业化这边的一个情况。 其实刚才易教总有提到一些。 能否请管理层更多分享一下可灵AI近期在商业化变现方面的进展以及2025年的商业化目标? 谢谢。 I will translate myself. Thanks, management, for taking my question. My question is regarding CleanAI's commercialization. Could management please share with us the recent progress in CleanAI's commercialization and the commercialization target for 2025? Thank you. Thank you for your question. Since CleanAI was launched, we have maintained our industry leading position through technology and product delivery. On the other hand, we are also exploring the commercialization of the parallel model in order to realize the direct circulation from the development investment to the commercialization of the parallel model as soon as possible. Currently, CLEAN AI has built a multi-layer parallel model of Western subscription plus closed-end API service plus customized scene resolution program to provide corresponding exclusive functional services to domestic and overseas users.

Thank you for your question. Since launching Cling AI, we have focused on two key areas, maintaining our industry leadership through continuous technology and product generations, and exploring monetization tools that will turn our R&D investments into revenue. Right now, Cling AI generates revenue through user subscriptions, API interface solutions for businesses, and customized scenario solutions, providing exclusive functionality and services for users in both domestic and international markets.

From the results of the commercialization transformation, CleanAI has been trying to start the commercialization transformation since the 4th quarter of last year. The monthly revenue scale has increased rapidly. Since the commercialization of CleanAI in February 2025, the total operating income of CleanAI has exceeded RMB1 billion. We understand that CleanAI is currently the first video production AI application with a domestic commercialization transformation scale. It is worth mentioning that the revenue of overseas Western users is also growing at a high level. At the same time, the global number of users has maintained a high flow rate of fees, and has further boosted the leading technology and products of CleanAI, and has become a mainstream video creation tool for global AI content creators and artists. As Western subscriptions continue to grow, The B-end service size of CLIN has also shown a rapid expansion. We have worked with thousands of domestic and foreign business clients in China, including Xiaomi, Leicester Lighting, AWS China, and Raypix, to demonstrate the excellent capabilities of CLIN AI as a video creation technology facility for all industries.

Since we began monetizing Cling AI in Q4 last year, its monthly revenue has grown at a rapid pace, substantially. From the beginning of its commercialization until February 2025, Cling AI's cumulative revenue has exceeded 100 million RMB. To our understanding, Cling AI is now China's number one view generation AI application by revenue, and revenue comes from international users grow rapidly with a strong retention rate in our global payings. script or base this further underscores a clean as global leadership in AI innovation where it's become a go-to video creation tool for AI content creators and artists worldwide beyond our growing user subscription our services for business are also quickly scaling we have partnered with thousands of domestic and overseas enterprises including xiaomi blue focus AWS China free pic and showcasing Cling AI's exceptional ability to empower video creation infrastructures across various industries.

We believe that as Cling AI's technology and product functions become more and more perfect, it can meet the needs of more and more scenarios and users. We will work with creators and corporate clients to grasp the key needs and explore more high-value scenario applications. to build a full network ecosystem of tools for the use of content production and content transformation. Next, we will also develop more active market and brand operations in domestic and overseas markets, and continue to expand the user circle and brand influence of CleanAI. We are also very confident that we will achieve a transversal growth in CleanAI revenue in 2025.

We believe that as CleanAI continues to evolve with the stronger technology and more advanced features, it will serve as an even wider range of individual users and business across more scenarios. Working closely with creators and business clients will identify key demands, explore more high-value scenario applications, and build a comprehensive ecosystem that connects AI to app production, content production, and monetization. While keeping our ROI in check, our next steps will be to scale up marketing and brand activities, both at home and abroad, expanding Cling AI's reach and influence among broader demographics. We're confident that Cling AI will have significant revenue growth in the year of 2025.

Thank you.

Next question, please, operator. The next question comes from Daniel Chen from J.P. Morgan. Please go ahead.

请提问。 好的,谢谢管理层接受我的问题。 我的问题也是关于AI的。 您在演讲中提到公司正在通过AI技术应用 赋能现有内容和商业生态, 能否展开介绍一下具体的应用进展和规划? 谢谢。 So thank you for the opportunity to ask questions. My question is also related to AI. In your speech, you mentioned that the company is harnessing advanced AI technology to empower its content and also business ecosystem. Could you please elaborate on the progress of the specific application and also your plans for them? Thank you.

谢谢你的问题。 正如我刚才提到的,通过AI技术赋能现有业务形成研发、投入和SOE的同样发展。 to make fast-tracking AI truly able to go through the cycle of the market. Since the launch of the new AI strategy, we have been working step by step to continuously achieve stage-based results, while continuously optimizing the comprehensive performance of the machine model, and continuously strengthening the negative energy of AI technology to the existing content and commercial ecology.

Thank you for the question. As I mentioned earlier, harnessing AI to empower our existing business and align with our R&D investment and returns will ensure our AI products can transcend market cycles. Since we first launched our new AI strategy, we have made steady progress and consistently hit key milestones, one after another, along the way. And by continuously enhancing the comprehensive performance of our foundation models, We have fortified AI's role in empowering our content and business ecosystem.

We have fortified AI's role in empowering our content and business ecosystem. can automatically learn about video content, live content, comment content, commercialized content, and user-oriented attributes, and perform higher-level reasoning, recommendation, and matching. In search scenarios, we also integrated DeepSeq R1's large-scale ability at the beginning of March this year to optimize the search effect and user experience, and further improve the activity of users, and explore the commercialization of search scenarios.

In content understanding and recommendations, we use large AI models to analyze short videos, live streaming comments, and user interests. The result is more accurate user recommendations that have increased user activity and time spent on our platform. Large models' higher-level reasoning allows for more advanced content analysis, smarter recommendations, and better matching capabilities, effectively replacing the need for extensive manual labeling. Instead, our models automatically learn from videos like streaming, comments, and commercial content, as well as user preferences. For search scenarios, we started internal testing in early March, integrating DeepSeq R1's large model capabilities to improve intelligent search accuracy and user experience. This has helped to drive higher user activity while tapping into monetization opportunities in our search scenarios.

In the three-dimensional scene, we expect that based on the big model, the understanding of the product and its characteristics and user expression, and using the model structure and sequence expression, we can continue to optimize the model prediction ability and bring about an increase in advertising investment effect, and thus drive the growth of marketing investment. For example, in terms of content production cost for marketing customers, We expect to help most customers reduce 60% to 70% of the production cost of this product marketing material, or even higher. At the same time, as virtual digital AI technology continues to mature, we can also provide the business owners with virtual live broadcasts to help the business owners break through in the field of e-commerce. In commercial applications, we expect AI models to keep improving on analyzing product semantics and user behavior.

By refining model architecture and sequence representations over time, We can improve model estimation capabilities, which will enhance ad performance, ultimately driving growth in marketing spend. For example, we expect our AI-powered tools will reduce most of our marketing clients' short-of-view production costs by 60% to 70% or more. Additionally, as AI digital human technology matures, we are creating new opportunities for merchants with digital human live streaming rooms, helping them achieve standout content-based e-commerce. In 2025, we expect AI marketing products such as AIGC short video marketing materials and virtual digital human live streaming solutions to continue to add incremental value growth to our online marketing services revenues.

Thank you. Last question, please, operator.

这一道问题由来自CITIC Securities的廖源。 The last question comes from the line of Yuan Liao from CITIC Securities. Please go ahead.

请提问。 感谢管理层接受我的提问。 我的问题其实是关于今年公司在AI方面的一个投入。 管理层能否分享一下今年的一个投入计划规模 Thank you for your question. As Yixiao said, AI strategy is very important for the company's future development.

Thank you for your question.

As Yixiao mentioned earlier, our AI strategy is crucial to the company's future development. We've already seen the tremendous value it brings to our content and business ecosystem. With the CleanAI's Money Sensation Accelerating, We're even more confident about making a firm and long-term commitment to investments in AI.

那么在AI相关的资本支出投入方面,我们预计2025年较2024年仍然会持续增长, 但增长规模相对可控,主要有以下几个方面的原因。 首先,我们在23年和24年已经进行了一定提前量的训练算力的采购和储备。 Our plan is to continue to purchase computing power in 2025. The core goal is to ensure the sufficient training and computing power of Kerlin to ensure the technical advancedness and leading position of the video model. In addition, this year's AI-related capital expenditure will have structural adjustments. We will cooperate with Kerlin's commercialization transformation progress, dynamic adjustment and increase the expenditure on computing power in Kerlin's push. Satisfy the market supply of computing power used in Kerlin's push. Therefore, In terms of AI-related CapEx, we plan to increase our 2025 spend in AI compared with 2024, but in a measured way.

We already pre-purchased and stored some training computing power in 2023 and 24, and we'll continue purchasing computing power in 2025. Our main goal is to ensure that Cling AI has the training computing power needed to keep its leading edge. Second, we are making structural adjustments to our AI-related CapEx. As Cling AI's commercialization grows, we will adjust and increase investment in inference computing, the computing power required for inference is sufficiently supplied in the market. That is why, while we plan to add more inference GPUs, the overall CapEx increase will be controlled. Finally, we will keep improving the utilization efficiency of our existing computing power and servers through engineering innovation and operational optimization, making sure we get the most value from our existing infrastructure.

In terms of AI-related spending, for example, the cost of relevant R&D personnel, we expect the same ratio to increase to a certain extent, mainly due to the attraction, retention, and team building of AI technology talents. Based on the above capital spending and spending, we expect this year's investment in AI to have a 1-2% impact on the cash adjustment net profit rate. Of course, we will also combine business progress to dynamically adjust the investment scale in AI. For example, the continuous promotion of KOLIN's transformation progress may further improve the investment in the push-down algorithm. But it is worth mentioning that KOLIN has already achieved a great improvement in the interest rate level in the push-down production video, so the investment in the push-down algorithm will have a relatively small impact on profits. From a long-term perspective, with the continuous release of AI for content and commercial ecology's negative value, we believe that AI's increased investment

For AI-related expenses like R&D personnel costs, we expect a moderate year-over-year increase as we look to attract, retain, and develop our talented AI team. Overall, we estimate that AI investments will impact our adjusted net margin by about 1% to 2% this year. Of course, we will dramatically adjust the scale of AI investments based on business progress. For example, if Cling AI's monetization keeps accelerating, we might increase investment in inference computing power. That said, since Cling AI has already reached break-even in terms of the gross profit margin in inference-generated videos, additional spending on inference will have a relatively small impact on our bottom line. In the long term, as AI continues to empower our economy and business ecosystem, we expect this to meaningfully drive higher revenue efficiency and profitability for the company over time. Thank you.

Thank you. That's the end of the Q&A session. Back to you, operator.

今天的问答环节到此结束。 现在有请Matthew做总结性发言。

Thank you once again for joining us today.

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