Cheetah Mobile Inc.

Good day and welcome to the Cheetah Mobile second quarter 2025 earnings conference call. All participants will be in listen-only mode. Should you need assistance, please signal a conference specialist by pressing the star key followed by zero. After today's presentation, there will be an opportunity to ask questions. To ask a question, you may press star then one on a touch-tone phone. And to withdraw your question, please press star then two. Please note this event is being recorded. I would now like to turn the conference over to Ms. Helen Zhu, Investor Relations of Cheetah Mobile. Please go ahead.

Thank you, all three of you. Welcome to Cheetah Mobile's second quarter 2025 earnings conference call. With us today are our Chairman and CEO, Mr. Fu Sheng, and our Director and CFO, Mr. Thomas Yuen. Following management's prepared remarks, we will conduct a Q&A session. Please note that the management script will be presented by an AI agent. Before we begin, I refer you to the safe harbor statement in our earnings release, which also applies to our conference call today. I will make four looking statements. At this time, I will now like to turn the conference call over to our chairman and CEO, Mr. Fushun. Please go ahead, Fushun.

Thank you everyone for joining us today. In the second quarter, we delivered our best results since Q1 2021. Revenue grew 58% year-over-year, driven by a 39% year-over-year increase in internet business and an 86% year-over-year increase in AI and other segments. Our operating loss decreased 86% year over year, while long gap operating loss was down 97% from last year, almost break even. In the first half of 2025, our revenue grew by 47% year over year. We believe we can maintain fast growth in the second half of 2025, driven by about 100% year over year revenue growth in our AI and other segments, along with a stable internet business. This shows our turnaround is working and gaining momentum. What is even more important is how we work today. We have made AI a core part of our process, working in an AI-native way. Our R&D teams are small and flexible, using AI every day to design, test, and build products, much like open-source developers. This helps us move faster and use fewer resources and shows AI allows one person to do what once took a whole team. We have been investing in AI since 2016. And at the intersection of AI and robotics today, we now have advantages and experience that are hard to replicate. For example, DeepRoll, our AI tool that turns video, audio, and documents into summaries and mind maps, runs with only three full-time employees. Our core internet business remains solid thanks to our shift from advertising to a subscription model, which has improved user engagement and retention. Today, subscriptions make up about 60% of our internet revenues. This healthy base gives us the room to invest in new AI products while staying financially disciplined. We are enhancing existing apps like Duba Antivirus wallpaper apps and PDF tools with AI agents. For example, in Duba Antivirus, we are testing a new AI feature that helps users fix PC issues, especially long tail problems they couldn't solve before. And early feedback is encouraging. While we are still in the launch and improvement phase for most AI utilities, we believe Chase has a natural advantage in utility applications. At the end of the day, the core value of AI utilities is to help people work more efficiently and productively. If we can deliver on that, we believe users will be willing to use our products. On the service robotics side, we made solid progress. Revenues from service robots continue to contribute to growth in the AI and other segments. In late July, we completed the acquisition of Ufactory Wound of the few robotic arm companies that is already profitable and earns most of its revenue overseas. Combining Ufactory's strengths with JADIS distribution network and 100 plus global partners give us a clear advantage to scale globally. Ufactory arms, already being used at scale in real-world scenarios, from assembly, picking, painting, and dispensing tasks in factories to grabbing beverages, making coffee and beers in commercial applications, strawberry harvesting in agricultural settings, and even in universities for robotic research. We now have a broad range of robots and are piloting wheeled robots with arms that can handle more physical tasks in more places. We believe the true breakthrough in robotics is not just in using the most advanced lab technology, but about finding technologies that match real-world use cases which can scale and generate earnings for the company. While the future of robotics is exciting, our years of experience tell us that real commercial adoption depends on delivering sustainable ROI that customers can clearly see. Our strategy is to stay optimistic, yet patient, Moving forward steadily, we will continue to identify scalable use cases and grow the business gradually. That said, we want to caution investors that it is not something that will reach mass deployment in the coming quarters. The service robotics market is still developing, but AI agents are making robots smarter and easier to use. Seems adding agent to the ask. Our next generation voice system powered by AI agents. Earlier this year, our voice-enabled robot revenue in China grew by about 100% in Q2, both driven by recurring demand from our existing channel partners, alongside expansion into new high-quality customers in healthcare, education, eldercare, and cultural institutions, such as the National Center for the Performing Arts. In addition, this group does not rely on one-off large orders, but comes from steady and repeat demand, especially in use, cases like poor guiding and reception, which shows it is sustainable. Few companies have both our global experience in consumer internet products and use of real-world robotics operations. This unique combination allows us to apply AI agent technology across both software and hardware, creating synergies that are hard to replicate, supporting our goal to become a leading service robot company in the coming years. Looking ahead, our core internet business remains healthy and profitable. We will keep investing in AI tools and robotics with discipline, and we are on track to reach profitability in the near term. Our strong cash position and zero debt give us the flexibility to grow while keeping our finances strong. The transformation is just getting started, but it is already producing results. We are building two growth engines, AI-powered utility apps, and AI robots that work together as synergistic forces, combining software and hardware to create a stronger moat, expand our market reach, and open new growth opportunities. At the same time, our solid internet business and strong cash reserves provide a stable base. With over seven years of our being in AI-focused strategy and a culture of innovation, we are confident about the road ahead.

Thank you, Fusha. Hello, everyone, and thank you for joining the call. Unless otherwise stated, all financial figures are presented in R&B. In the second quarter of 2025, we continue to make meaningful progress in narrowing our losses and improving probability as we remain focused on execution efficiency and financial discipline. In fact, on a non-GAAP basis, We almost reached the break-even point on the operating level in Q2. Let me walk you through the key financial results in the quarter. Total revenue reached RMB 295 million, up 58% year-over-year and 14% quarter-over-quarter, marking a strong acceleration. Cross-profit increased by 85% year-over-year at 19% quarter-over-quarter to RMB 212 million. Gross margin improved to 76%, up from 65% in the year-ago quarter and 73% in the previous quarter. Operating loss narrowed to RMB 11 million, an 86% year-over-year decrease and 58% quarter-over-quarter decrease. On a non-goal basis, our operating loss declined to 2 million, down 97% year-over-year and 86% quarter-over-quarter, not loss attributable to Cheetah Mobile. year-over-year, and 32% quarter-over-quarter to RMB 23 million. A long-dubbed net loss attributable to Chernobyl shareholders, narrowed by 87% year-over-year and 35% quarter-over-quarter to RMB 14 million. These probability improvements reflect our ongoing efforts to sharpen focus, improve efficiency, and optimize our cost structure. particularly as we have switched from early-stage experimentation to ROI, focused execution in our AI initiatives. In our AI robotics business, we have excited certain compute-intensive directions, such as creating our own foundation models, a strategic shift that significantly reduced infrastructure spend. At the same time, we have streamlined our R&D process by leveling AI tools, and we focus resources on AI utility applications that generate user value. For example, R&D expenses accounted for 24% of our AI and other segment revenue in the quarter, down from 39% in the year-ago quarter and 28% in the previous quarter. These efforts have materially improved the operating profit of our AI and other segments. where adjusted operating losses decreased 63% year over year and 32% quarter over quarter. Looking ahead, we remain confident in our ability to achieve profitability. With a clear and disciplined strategy, we see two key drivers for this path. First, our internet business continues to deliver steady profits and serves as a solid financial foundation In Q2, adjusted operating margin for this segment was 14% up from 12% in the year-ago quarter. Our transition from an ad-centric model to a user subscription-driven model is showing good momentum. We believe this momentum is sustainable, supported by loyal use cohorts and diverse site distribution channels. Second, in our A and other segments, We are building for long-term probability by growth in both our consumer-facing AI tools and enterprise-facing robotics. For our robotics business, we are prioritizing sellable use cases with clear user demand and engagement, emphasizing our core competence, denied powered voice interaction, including natural conversation capabilities similar to LLM-based agents, and real-best indoor mobility, which we believe offers the most reliable and cost-effective solution for scalable robot deployment. continuously improving our robots' intelligence and product experience through AI agents, and maintaining a lean and agile team structure. A recent milestone was our acquisition of U-Factor, one of the few profitable robotic arm companies globally. Your factory brings a proven track record of profitable growth, clear market position, and consistent value creation, fully aligned with our vision to scale differentiated robotic solutions over time. On the AI tools front, DeepVoo, an AI tool that summarizes video, audio, PDF, and other documents, leads to concise takeaways and mind maps. has shown encouraging early user adoption, validating product market fit. Our balance sheet remains strong. As of the 30th of June, 2025, we held $282 billion in cash and cash equivalents, and $110 million in long-term investments. We generated RMB $362 million in operating car flow during the quarter. This financial strength gives us the flexibility to continue investing in high-potential AI growth opportunities while maintaining capital discipline. We will also remain open to strategic M&A that can accelerate capability building in targeted verticals. To summarize, this was another quarter of measurable progress on our path to breakeven. We are encouraged by early signs of sustainable profitability supported by, one, a profitable and resilient internet business. Two, a disciplined ROI, focused eye strategy, and three, strong capital flexibility to invest in long-term growth. Thank you. We are now happy to take your questions.

Thank you. We will now begin the question and answer session. To ask a question, you may press star, then 1 on a touch-tone phone. If you are using a speakerphone, please pick up your handset before pressing the keys. If at any time your question has been addressed and you would like to withdraw your question, please press star, then 2. At this time, we will pause for a moment to assemble our roster. The first question today will come from Thomas Chong of Jefferies. Please go ahead.

晚上好,谢谢管理层接受我的提问。 我的问题是我们看到上半年互联网业务恢复明显, 收入增速保持在约40%, 并贡献了超过4700万元的经营利润。 How do you see this trend in the next three and a half years? What are the main driving factors of income and demand? If we don't consider the contribution of AI tools and products, how do you see the open-ended income level and demand in the Internet business?

Thank you.

Thank you Thomas. Let me answer your question. First of all, thank you for your attention to the company's Internet business. Regarding the continuity of the Internet business growth trend in the second half of the company, we also mentioned in the last quarter that the key to the growth of the company's Internet business in recent years is that we have turned from the commercial model based on advertising to the core of user payment. return to the value of the product. By insisting on the user and continuously polishing our product power, we achieved a steady growth in user size. At the same time, we also established some stable long-term partner relationships and new effective customer channels. We believe that with our continuously improved product power and high-quality user relationships, the growth momentum is stable and sustainable, rather than a short-term wave. In the future, growth will depend on two aspects. On the one hand, we will further expand our new partners and channels on the successful cooperation basis on market expansion. To expand the coverage of our users and the basis of income. On the other hand, we will use the technology to play an active role in the AI agent technology. Whether it's PC or mobile, the tools and products will continue to improve our product competitiveness and the user's desire to pay for it to create a new value space. But I also need to explain to you that we can see that we have achieved a relatively large growth in the Internet business from the second half of the year. So the income base has improved significantly in the second half of last year. So, although we expect the absolute size of revenue and the ability to profit will continue to maintain a relatively healthy level, but in the second half of the year, there should be a slowdown in the same rate increase. This is also a mathematical rule.

Regarding the specific growth of the Internet board, the speed and ratio

can be higher than the level of last year. But whether this goal can be achieved depends mainly on how many seasons in the future we are able to successfully expand our new channels, find more partners, and realize AI agent technology's most limited capabilities, or make some new products. We will work hard to advance in these directions, and strive to continue to create long-term value for shareholders. Regarding your question about profitability, we believe that the Internet of Things business, whether it's advertising or fee-to-use, is based on some of the industry's peers and our past performances. I think we still have room for improvement in terms of profitability, but it also needs to be explained that we actually, as I said earlier, we will continue to make some AI-based upgrades to the Internet of Things products. for example, to transform the original tools into a new one, or to introduce new tools. These investments are mainly for the future competitiveness. In the short term, it may also have an impact on the profitability. At the same time, the net profit of the Internet business itself will also be affected by the different types of products, the mode of consumption, or the fluctuation of seasonal factors. So we actually need to to seek innovation, future growth, and maintain the existing healthy profit-making system to find a balance. And then you asked about not considering the increase in the contribution of new love tools, which is a normal profit-making level. This is a very good question. I think our basic Internet business is relatively stable. The standardized history will basically remain in a stable area. The factors of growth mainly come from the transformation of our current business, the improvement of new channels, and the improvement of operating efficiency. Of course, it will also be affected by the various product combinations, channel changes, and other factors we just talked about. We may also invest in some new products. Based on these possible factors of influence, I think the profit level of the entire Internet business at the current level of this quarter, up and down a few points, one or two points, may be a normal range. In short, our Internet business is still full of confidence in the company's continuous contribution to the company's profits. It will continue to be a cash cow business of the company, giving us reliable profits while also helping us Thank you. Operator, please move to the next question.

Our next question will come from Vicky Wei of Citi. Please go ahead.

Thank you for accepting my question. I want to ask a question about our domestic mobile robot business. We have seen a 100% increase in revenue in the second quarter. At the same time, our company's machine currency business will start to contribute to the increase in revenue in July from the second half of the year. Can you please introduce the current robot product sales situation? Thank you. Okay, let me answer this question.

In the first half of this year, one of the highlights in the robot service sector is that . . . Because of the noise problem, some words may be wrong in the voice translation, but the whole model can still understand this content and give you a good answer. So this makes the user satisfaction and the actual effect of the user interaction with the robot have been greatly improved. I think the most important factor for growth is this. Of course, I have to say that voice-over robots are not a large category in the entire service robot industry. Although the growth rate is high, the total number of robots is not too large. Then, let me make a guess. Let me finish the voice-over first. What we are going to do in the second half of this year is actually to really promote the multi-language ability of voice-over robots overseas. Of course, overseas, as you all know, the laws and regulations on AI data It's more complicated. And there are also various languages. With the big model, we believe this process will be faster. Recently, two days ago, I myself went to the Singapore Google Cloud Collaborative Conference. We are also the only one in this conference to go to the main conference venue and their entire demo section to put the latest Gemlight Live model on Google. It works well with our voice-to-speech robots. It works well in multi-language and Q&A. So I think we are optimistic about the growth of voice-to-speech robots. Of course, another factor is that voice-to-speech robots are not a 2C product. They are a 2D product. So this is also a change from a pure 2C company to a 2B company in the past few years. We have gained some experience in how to get customers and how to manage sales. I personally think we have entered the door. So this also depends on the construction of our entire sales system. So in the second half of the year, we are still confident in the continuous growth of voice robots. Despite the impact of various environments, I think the voice-to-speech robot has become a product that can really land in today's service robots. It may be different from what everyone thinks. Today, you see, in a truly commercialized service robot, one is the voice-to-speech robot, one is the voice-to-speech robot, and the other is the voice-to-speech robot with arms. But what we can see today, what can really form a certain commercialization, is mainly地送 and voice interaction. And in fact, the technical complexity of地送 is not that complicated. So if you are familiar with this industry, you will know that the competition in the地送 industry in China is quite fierce. But if you really want to make voice interaction into a user experience in all kinds of situations, its technical complexity is a little bit softer. This may be more suitable for us to do it in the field of Internet business. So we chose this one at that time. And you asked about the problem of weight. We have completely completed the intersection in July. I always think that it is to serve robots or this. Today, there are two paths for GSM. One is represented by Tesla. It wants to be an all-powerful human-like machine that is fully human and one step ahead. The other path is represented by us and some competitors. It is also called the same industry. It is to find a scene and realize a scene, and then gradually polish the technology on it, and accumulate points one by one. Anyway, I won't comment on these two technical routes. But what we have chosen is, for example, to start with the transport of wheels and the transportation of clouds. Now we are going to acquire a core of Zhongwei, which is what I think of as a currency, which is mechanical currency. From the factory to the service scene, I think we have reached such an opportunity. Because Zhongwei, let me introduce it. It is not a pure industrial currency company, but it is called co-operative currency. Many application scenarios of co-operative currency are not completely in the factory. For example, if you go to their website, you will know. For example, there are those who pick strawberries. There are those who take a camera to shoot. There are also those who make coffee, and so on. Such scenes. Their scenes are richer. The scale is not as big as the industry. But every scene is growing. The scale of income in Zhongwei, we don't have an example, right? Taiwan Taiwan Taiwan and the impact of the brand and the users. Secondly, it is also today, in fact, the machine currency track is quite winding, but it is a machine currency company with a relatively high profit rate, and income and profit have continued to grow in the past three years. Today, from the first half of the year, from its overseas data, I think their growth is still going on. Of course, in the future, I also hope to put the robot wheels and coins together to help you do housework or do division. But what I want to say is, although you may have seen a lot of such videos, it has really become commercialized. There is still a long way to go. It's all details. It's really possible to complete this. It has a good input and output ratio, and it can be successfully commercialized. I think we still need some time to explore. Overall, in the robot, Thank you.

Thank you, operator. Please move to the next question. Thank you.

Our next question is from Nancy Liu of J.P. Morgan. Please go ahead.

Hello, thank you for accepting my question. My question is, because we see that the second quarter of the company is close to a profit-loss balance, are you confident in achieving overall profit in the third quarter? From the structure point of view, we have just explained in detail the profit and loss level of the Internet business in the future. If the robot business wants to achieve profit and loss balance or future profits, what is the biggest obstacle we see? How does the company prepare to respond? Thank you.

Okay, thank you. Let me answer this question. First of all, about whether the third quarter will be profitable. In the last quarter, we also talked about it. We actually pay great attention to the improvement of profitability within the company. And we also aim to achieve the overall profitability as one of our important goals. From the second quarter performance we just released, we should be able to see that we are currently in the second quarter. and the performance of the third generation is full of confidence. The Internet business has just been mentioned. We are also confident in continuing to contribute to the Internet business. It is a forward-looking business of the company. Regarding your question about the profit-loss balance path of the robot business, first of all, the robot business is certainly a key to our entire company's overall achievement. If you want to I think the biggest obstacle is whether we can continue to find and focus on the specific landing site that can be scaled. Scaling means the need for technology has to have a consensus, the market space has to be large enough, and the business model has to be repeatable. We know that the entire industry is still in a stage of exploration and exploration. To blindly pursue a technical priority and ignore commercial applications It's hard to achieve sustainable profits. Our approach is to focus on deep strategies. First, we focus on high potential scenarios. We don't do illegal investment. We focus on the areas that we think are most likely to achieve commercial returns. We do deep growth in these areas. Second, we strictly evaluate the ROI of each project. When we decide on any investment, we evaluate the return on investment of the project more thoroughly. Although AI and other businesses, including our native businesses, including the robot business, have other native businesses, the loss of the business line and the rapid growth of the business are partly due to the changes we made in the robot business. In addition, the acquisition of Zhongwei is also a good example. It can become one of the few companies that continues to make profits because it has found a very clear scene. It is also very specific in the field of education, R&D, and green industry. It is a very good developer and innovator. The product also has a very good user reputation and a high resale rate. Thank you.

Let me add something. You asked about the robot industry, because we started working on robots in 2017. At that time, AI and robots were called out very early on, and they were not industrial scenarios. I think the biggest challenge in this industry is a very common principle, which is called ROI, which means investment in parameters. Today, due to the support of this large model, due to the arrival of artificial intelligence, this industry has become very hot. But I want to say that no matter how hot the industry is, when it finally comes to commercialization, it will be the investment of digital currency. Can you be efficient in this position? And then this is called high reliability to complete low-cost completion tasks. This is its core point. For example, we've seen a lot of videos of people doing something, then once you show it and repeat the work, but without making mistakes, there is still a big gap. So I think today the robot can really make a profit by finding the most appropriate technology to complete it, rather than the most advanced technology to complete it. So I think the real way to achieve profit is to find this kind of scene that can be focused, and the investment of this technology is not very big, to achieve certain results. Thank you.

Alfredo, please move to the next question. Thank you.

Our next question today will come from Brenda Zhao of CICC. Please go ahead.

Good evening, Director Guan. Thank you for accepting my question. In fact, the company has mentioned that AI agents will be integrated into existing Internet products and service robots. Thank you. Let me answer this question.

First of all, we have integrated the AI function into it. I have already spent a lot of time talking about it. It is integrated into the robot. This kind of interaction experience and the satisfaction of interaction is clearly increasing. This year, because of the growth of the robot, we can clearly see that it can win in many scenarios. This is a very new data. That is to say, including the customer and agent purchase rate of our robot is increasing. On this point, Because as such a robot, its ability to interact with users is very sensitive. This is the first part. Then you said that in the Internet products, you have added the AI function, and the user behavior has changed a little. I have to say that the AI agent is very popular now, but everyone is also looking for and trying this stage. At present, it is not too much to see that there is a large amount of real-time income in the agent field in the industry. We are now also in a combination of some agents and, for example, some of our older products. For example, products like Jin Shan Du Ba, we have already started to do a combination, which is to be able to make this When you are adjusting the computer settings, you can just go to the computer and talk to it. You don't need to go back and forth to find the settings. And then we also talked about some meeting summaries and other products like that. With a small amount of data testing, we think the user payback rate is still good. But today, we haven't done a large-scale promotion or coverage in our product line. We are more, but, yes, some of the frequency and frequency of use, we can see one thing, that is, if your big model brings this kind of interaction, the effect of bringing it to the result is good, the use of the user, this frequency and this is called, Okay, thank you.

Our next question today will come from Gigi Xiao of GF Securities. Please go ahead.

Thank you for answering my question. Because now I see that the competition for some applications is getting more and more intense. I want to ask what is the protection box of our AI tools for the long term? Is it because we are involved in the workflow to collect data or is it through product habits to cultivate and polish to form a user-friendly system? In the future, what direction will the company focus on to create some barriers?

Thank you. Okay, let me answer. This question is quite broad. Because today's telecommunications movement has been from a pure 2C company in the past to 2C and 2B. Let me talk about it. Today, we still think that 2B... Actually, from my point of view today, sales and technology are equally important in the 2B market. On the one hand, we need to build up our entire sales network. We need to know where the needs of our customers are, so that we can quickly integrate them. On the other hand, as you just mentioned, they are involved in its workflow, and they are accumulating data on its privacy. Once such users use our product, in a sense, its accumulation is relatively natural, because you are following the company. In fact, I saw that Oracle has grown a lot yesterday. It is also in this area. We are advancing in this area. I think you may want to ask more about this kind of 2C AI tools. To be frank, I think today in such a fierce AI competition, no company can say that it has something called a long-term backup. I think it's more about the ability to quickly, according to the needs of the users, to launch the product. It's a dynamic backup. Because at this stage, you can quickly use AI to launch on the needs of the users. Then it's your first stage of backup. I think what we have done this year is very important. We have put a lot of effort into improving the organizational management system within the company. In terms of AI So I think at this stage, we should have a lot of opportunities in the growth of AI tools. In the long term, I think it's kind of like a product to satisfy the needs of users. When users think you're good at something, Then you use the most advanced model to do better integration and better understand the customer, it becomes a dynamic barrier. In this, I think the long-term protection comes from you and you continue to get the user, to grab the user's mind, and then a lot of the behavior of the user becomes your recognition of their need for these tools. Please use such a link. Today it seems that such a link is possible. I won't advertise it to others. Today, the most popular applications overseas, such as AI and agent applications, are all based on this kind of delivery. Although users have enough choice, they won't spend so much money to choose. When you can meet their needs, the possibility of their transfer is not that high. So this is a process of continuous dynamic delivery. This is my answer. Thank you.

Operator, please move to the next question. Thank you.

Our next question today will come from Chengru Li of GoWon Securities. Please go ahead.

Hello, everyone. Thank you for accepting my question. And I also congratulate GoWon on achieving this goal. . . . And the second one is about whether the company has any other new products that are being planned. And what are some of the logic behind the landing of these products in the future? They will play a role in the overall strategy of the company.

Yes, that's it. Thank you. Yes, the new model of Lunzi plus B is indeed in the market. um um But to be frank, we don't have a very clear timetable at the moment. I think today is the time when it's difficult for the robot to land. Because it's not something that comes from the needs of the R&D section at all, but it really depends on the needs section. Of course, I know that there are some, for example, in the warehouse, in the division scene, there are some wheels and walls. The plan has already come out. But what I want to say is, is it a quick solution that can grow rapidly in these scenarios? It depends on the real landing effect that the current technology can achieve. So it should be said that we maintain this, it should be said that we maintain this very high, this is called, high degree, close to high degree of attention. When the time is ripe, We definitely will. Because we have the ability of the base plate itself. We have launched restaurants, hotels, and factories that weigh more than 150 kilograms. We have this kind of technical product. And the coin of Zhongwei has a pretty good reputation overseas. We think we need to look at it from a specific perspective. Right now, we don't have a very clear timeline. But in the future, I will spend some time on these markets and scenarios to run some first steps. And since you asked if the company has any new product models in its plans, we do have new product models in our plans. We should be able to find some bigger scenarios that are suitable for robot technology. We have such a plan and we are also implementing it. The logic of landing is to be able to really, or to say, in a scene where the needs are verified, to really make the AI, the AI of the robot, to really land and use it well, so that the user's use experience, the user's input and output ratio can be increased by a level. Thank you. Thank you. Operator, please move to the next question.

The next question today will come from Zhang Huang of Everbright Securities. Please go ahead.

Vice President, President Ren, President Helen, good evening. Can you hear me? If you can, thank you for accepting my question. I am Huang Zheng, an analyst at Guangda Hiawai TMT. First of all, congratulations to the company for its bright performance. I have a few small questions here. The main focus is on AI promotion. Of course, Vice President may have a certain answer to this part just now. The company has also made significant improvements through this small team and AI tools. The next step is the focus of our company's promotion, which may be focused on some aspects.

There are some particularly detailed data that we may not be particularly convenient to disclose, because we think this is also one of our core competitiveness, that is, the core competitiveness that is being built. I think there are some that I can roughly talk about. In terms of research efficiency, we think the small team plus AI tools have significantly improved the research efficiency. And this research and development cost increase is very obvious. I can give a small example. We once held a press conference, and we made a recording summary tool, right? Such a tool in the past, if you are an experienced student, you may need a team that is not too small, such as a team of five to ten people, or a team of six people, to launch websites, apps, and even hardware operations. But this time, we used a small team. We probably used two colleagues who didn't have much experience before, and it worked very well. This is a very intuitive indicator. There is one more thing. As you all know, we were able to get close to the profit after we merged with R&D. Last year, after we acquired R&D, we had a loss of 200 million RMB. Now, we are almost at a loss of balance. both in the development of robots and AI, and in the development of AI Agents. On the one hand, of course, there is a growth in some businesses. On the other hand, it is actually an improvement in our development efficiency. In terms of the ratio of income, my idea is that we will not only increase our so-called optimization space by cutting development costs. We will, in the case of the cost of the development, or in the case of the overall situation, on the one hand, rely on better products to gain better revenue space, revenue scale. On the other hand, the efficiency of the development team can still be improved. When it comes to the specific landing, what are the scenarios and challenges? Yes, I have been talking about this for a while, because I often post some videos. I think our purpose in these videos is not mainly to be popular, but to provide products to the company. I've talked about it before. This time, AI has entered into... It's essentially a productivity tool. And productivity tools are used within the company. In fact, the biggest bottleneck is not the tool itself, but the people who use the tool. I've talked about it in there. Today, it seems that many startup companies are called AI native organizations. That is, their entire organization and everyone's basis for working is built on AI, instead of being supplemented by AI. . . . . . . . . . . This is the organization system we are building today. Thank you. Operator, please move to the next question.

Our next question today will come from Guangpeng Zhang of Sealand Securities.

Please go ahead. Thank you, Mr. Wang, for accepting my question. In the field of robots, each product has a different technical line and product form. Compared to Tong'an, what is the difference between the technical focus and profit model of the company in terms of the path and business model of robot business? How will we show our unique advantage in the competition? Thank you.

Okay, I've already answered a lot of these questions. Let me make a summary. The word robot is a very big word. When it comes to robots, maybe everyone's definition is not the same. So today there are also a lot of product models. I think the biggest difference between us is that we first consider the location of the robot. Is it really possible to achieve the investment and output currency that we just talked about? Is it really possible to achieve the investment and output currency that we just talked about? Is it really possible to achieve the investment and output currency that we just talked about? And then, whether we can continue to purchase in this scene, to produce a resumption, to constantly expand the scale, or to consider it from the beginning, including how high the cost of technical investment is in this early scene. This is what we really, really value today. We do not think that It's our core competitiveness to create a free model or a free technology. The investment and output ratio of the real scenario is our strategic path and business model. Of course, there may be such a company that through a large-scale investment has finally come up with a technical route that is completely downgraded. But based on our understanding of AI today, I think the difficulty and the risk are both very high. For a company like ours, this is not suitable for our development path. Our unique advantage as a robot is that, first of all, we have accumulated over the past years, right? Where is the core technical advantage? We believe that in addition to the ability of some hardware, . . . China China China . . . In the hardware industry, we have accumulated our own money. On the other hand, we have acquired companies such as Zhongwei to continuously enhance our hardware processing capabilities. I think the last thing that a robot can really show is that it has to find the hardware, software, and scene. Thank you.

Thank you. Operator, please move to the next question.

Our next question today will come from Yanpeng Zhao of Guaytai HITOM Security. Please go ahead.

Thank you for the opportunity to ask the question. First of all, congratulations to the company for achieving excellent performance. Because we remember that CFO introduced in the last quarter that the investment department of Lebao has a lot of layout in the industrial chain of the robot. And then this season, Leopold has also completed the acquisition of Zhongwei, which is ideal. At the same time, we also noticed that there are more than 100 million dollars in long-term investments on the company's asset portfolio. I would like to ask the management if there are similar assets like Zhongwei, and then we can further promote the company's development in robot business or AI tool business through the way of acquisition. Thank you.

Thank you for your question. I will answer the question about investment acquisition. First of all, as you said, we have a relatively high long-term investment amount on the market. In fact, part of it is the strategic layout of the AI ecosystem around us. Investing in these companies is mainly through early investment and these innovative companies to establish a deep relationship, strengthen mutual understanding, and even establish some initial business cooperation. At the same time, these investments also provide us with some very valuable latest industry dynamics and some movements in the AI industry. But as I said in the last video, From our point of view, we need to consider whether to buy a company or not. We need to consider several aspects, including the collaboration of technology and business, the improvement of strategic value, culture and values, as well as financial valuation. We need to carry out systematic evaluation of these aspects. To evaluate whether a purchase is successful or not, Whether this acquisition can successfully integrate the core capabilities of the landmark company and the team to make a successful strategic acquisition, we think the most important key is the high harmony of culture, vision, and people. In the case of Zhongwei, which we have just acquired, we were able to successfully complete the acquisition and we are very excited. It is also because in the cooperation with Zhongwei for many years, we are sure that the team of the two parties has a highly consistent gene. We both believe that robot technology should be used to solve the problems of the real world and the real scene to pursue the real mobile robot. It is our common mission. The consistency of this underlying concept allows us to complete the integration of business, cooperation, and team integration smoothly. So, in conclusion, although our long-term combination provides us with this very valuable strategic money, if there is a similar high-quality market mark like Zhongwei, then we also need to make the most effective decision for us to improve the overall share price value after a thorough and rigorous evaluation process. Thank you.

Thank you. Operator, please move to the next question.

Our next question today will come from Johanna Ma of CNBI. Please go ahead.

Good evening. Thank you for answering my question. I would like to ask about the cost of AI tools. We saw that you mentioned that the cost of AI tools' development and maintenance is constantly decreasing. Um.

Okay, let me answer this first. As I said, we haven't made a large-scale investment and testing yet, so we can't get a clear data yet. You just mentioned that because of the use of AI today, we may all know that the latest round of financing from S-COPIC overseas has reached more than $1,800 billion. Its annual revenue from the estimated $1 billion in the beginning of this year has now reached $9 billion. And there are not too many sales teams. What does this mean? I think today AI is the first in the whole tool penetration of productivity. The largest penetration comes from the programming market. So what it brings is that we write programs today. Before, a lot of people wrote code. AI can already complete it very well. So this brings us to the fact that when you are developing new tools, the cost of development, the speed of development, including the speed of launching new functions, are all greatly reduced. The cost of development is greatly reduced, and the speed is greatly improved. This is a place where the cost is significantly reduced. As you just said, the cost of a large-model product is indeed higher than that of an Internet product. Because the more users use this token, the higher the cost. But today we can see two or three trends. The first is that the cost of this token is going down very, very fast in the entire industry. I didn't calculate it in detail, but today it should fall by at least 90% more than GPT-3.54. And we can predict that this decline will continue. In the next two or three years, it will definitely continue. Because through the optimization of the structure, or the introduction of all kinds of chip competition, it will continue to decline. And then, the first feature is the cost. The cost of the call and call will continue to decline. The second is that today, The user's willingness to pay for AI products is very strong. It is very, very obvious. That is, it has its own business model. Users are willing to pay for it. This has been reflected in many products. The third thing I want to say is that today, including the way it charges, I think This industry is also constantly searching, right? From monthly fees to now, there are volume fees, etc. What do I want to say? I want to say that the commercial model of today's AI products will be a little different from that of the tools of the last era. So it will repeat the commercial model. Then this the user's habit of using or paying for it, and constantly adjust the cost. So I'm not worried about the cost of the so-called AI tool. If we look at it from the perspective of development, as long as you can have more users using you, the more tokens you use, the better for you. This is a bit like the earliest online video website at that time. At that time, loan costs were its biggest cost. At that time, this cost made many websites not profitable. But today, the loan cost has become very, very... Its cost is very low. Of course, in another perspective, the copyright cost is much higher. This is another matter. So what I want to say is that when we look at the commercial model of AI tools, we must look at it from the perspective of development, not from the perspective of the present. And then the overall ROI, as I just said, at least from the cost-benefit point of view, the user is quite obvious. I believe this balance line will arrive very soon in many areas. At least today, I think it has arrived in the edge area. That is, you will find that the money you spend on a large model is much cheaper than the money you spend on a programmer. Then this market will be blown up very quickly. Thank you. Operator, please move to the next question.

Our next question will come from Jackie Yan of Mizuho. Please go ahead.

Thank you for accepting my question. Recently, I heard a saying that it is difficult to commercialize AI tools in China using the user payment model. This may be different from the overseas market. Based on Liao Bao's actual experience in China, how do you see the possibility and time table of this commercialized path?

Thank you. First of all, I think what you said in the first half of the question is right. Today, domestic users and overseas users pay. But I have to say, it's not all overseas users. It must be the ability of overseas developed countries to pay. This is without doubt. But what I want to say today is that the Chinese users' desire to pay for good tools and products was not the same as a few years ago. So I think that the way of paying for users in China today is definitely possible. Because you can see it in our financial reports, the payment income of users of our Internet products should be 60%. . . . . . Today, users who use tools that are not AI have already paid more than half. This also proves that in China, users do not have to pay a fee, but whether you really provide them with enough value. In the past few years, we have also grasped the experience of users on tools. We have been paying attention to users for a long time. simply focus on investment and output, right? the number of users, the number of products produced, and whether the users are really satisfied with your product. The return rate after payment, the long-term circulation after use, and the continuous use of these aspects. Now it seems to have achieved a very good effect. So I think based on our domestic experience, I think the user's payment for good products is the same in China. At this point, we have... Alfredo, can you please check if we have any further questions?

Thank you.

Certainly. Again, if you would like to ask a question, please press star and then one to join the queue. At this time, I am not showing any additional questions in the question queue.

Okay, and then we can just end the call.

Thank you.

Okay, thank you, ladies and gentlemen.

Thank you so much for joining us. Yeah, thank you so much for joining our conference call today. So if you have any future questions, please just let us know. You can send us an email or just give us a call. Thank you so much.

The conference is now concluded. We do thank you for attending today's presentation. And you may now disconnect your lines and have a nice day.