Cheetah Mobile Inc.

Good day and welcome to the Tito Mobile Fourth Quarter 2024 Earnings Conference Call. All participants will be in listen-only mode. If 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 your telephone keypad. To withdraw your question, please press star then two. Please note today's event is being recorded. I would now like to turn the conference over to Helen Jingshu, Investor Relations for Cheetah Mobile. Please go ahead.

Thank you, Operator. Welcome to Cheetah Mobile's fourth quarter 2024 earnings conference call. With us today are our company's chairman and CEO, Mr. Fu Shun, and our company's director and CFO, Mr. Thomas James. Fully in management's prepared remarks, we'll conduct a Q&A session. Please note that the CEO 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 earnings conference call today, as we will make forward routine statements. At this time, I would like to turn the conference call over to our CEO and Chairman, Mr. Fusheng. Please go ahead, Fusheng.

Hello, everyone. Thank you for joining us today. In Q4, Our total revenue accelerated robustly growing 42% year over year and 23% quarter over quarter. We continue to make progress by consistently reducing non-GAAP operating losses quarter over quarter. And we expect our first quarters non-GAAP operating losses to further decrease both year over year and quarter over quarter. For the full year 2024, total revenue increased by 21% to RMB 807 million. and our non-gate net loss attributable to shareholders narrowed year over year. This performance highlights our commitment to high quality growth and positions us strongly for a continued business turnaround in 2025. Our growth is driven not only by our enterprise facing AI and robotic businesses, but also by our legacy internet business, which continues to deliver a strong year over year revenue growth and margin expansion. Our apps for PC and mobile users continue to address pain points effectively, delivering excellent product experience even in a mature market. We believe that this positive trajectory will continue into 2025. Benefiting from the momentum of LMS, JITA entered 2025 with tremendous opportunities for our robotic business. LMS are playing a key role in transforming the service robotics industry, spanning its total addressable market and laying the groundwork for service robots to evolve into general purpose physical AI. At J-Curve, we are taking a measured strategic approach to growing our robotics business, focusing on product types that can generate seasonal revenue and profits, aiming to become a top three service robot provider globally within three years. Our wheel-based robots are designed to serve real-world business environments, enabling a data flywheel that will further drive product innovation and LLM advancement. At the same time, we continue to enhance our general-purpose robotic capabilities by leveraging the most advanced ILLMs available in the market, such as DeepSeq, CHAT, GPT, and et cetera, alongside our self-developed LLMs capabilities such as voice interaction capability and indoor autonomous are a foundation to the integration of physical AI over the long term. Speaking of voice interaction capability, building upon our success in voice-enabled robots, we will be launching Agent OS in the coming days, being the first to make our robot agent. Agent OS is our next generation voice interaction system for service robots' robots. backed by our self-developed AI agent technology. It powers robots to reason, to autonomously verify information, and call additional tools to complete tasks. AgentOS will expand our total addressable market. We build tailor-made AI agents to help ourselves or our distributors sell robots, and to assist restaurant and supermarket owners in greeting customers and introduce promotions. You can also easily Customize your robot by uploading your company's data, making the Cheetah robot especially adaptable to your needs. For example, our reception robot can greet visitors and even help visitors find nearby coffee shops, all by using voice commands. One of our distributors in China has already used Agent OS to sell Cheetah robots. They uploaded product details, images, and videos into the system. And now the robots act as smart sales representatives, giving clear and detailed answers that sometimes surpass those of human sales staff. This easy-to-customize solution has led to higher sales and improved efficiency. We believe Asian OS will strengthen our lead in voice-enabled robots. This improved product experience not only positions us to attract new customers, but should also drive repeat purchases from our existing customer base. particularly in corporate receptions, exhibition centers, public sector departments, schools, and universities regarding our indoor autonomous driving capability. As I stated in the past, we are gradually transitioning to vision-based solutions for our service robots. Vision, we help our robots better understand and navigate real-world environments further influencing the data flying wheel for further product innovation and LLM advancement. In addition, we are continuing to leverage AI to enable smoother collaboration between robots, as well as between robots and humans. Specifically, we use multimodal models to allow robots to follow users, even in crowded spaces. And we use enhanced AI algorithms for route planning, ensuring that multiple robots can work seamlessly together in a factory. Our ongoing product innovations and enhancements, combined with a strong focus on high-quality after-sales service and Trader's established brand recognition, have consistently helped us capture market share from peers and drive sustained sales growth. Going forward, we'll continue expanding globally by building local sales teams and distribution channels Learjinchi is proving experience operating outside China. Following our success in South Korea and Japan, we have made solid progress in Europe. After about 12 months in Italy, we became the country's largest service robot provider. We are planning to open our European headquarters in Germany, aiming to solidify our presence in the region. It's important to note that the adoption of service robots in developed markets significantly lags behind China presenting substantial opportunities, particularly in restaurants and fulfillment centers, moving our AI application initiatives. AIDS, our AI-based data service platform, has been well received by enterprises, including some well-known companies. We also continue to leverage the most advanced LMS to rapidly find product market fit. While building app applications with a small focus team, For example, since DeepSeek became available, our team quickly used it to transform my WeChat official account into an AI agent, a virtual fusion to enter users' inquiries. We have seen a significant increase in user engagement and new followers as a result. Additionally, we have integrated DeepSeek to support our voice-enabled robots. allowing businesses to easily upgrade their robots and gain access to the deep-seek model through our devices. Our strategy for the AI application business is to build a product portfolio while focusing on building killer apps in selected areas. By leveraging the most advanced ALMS and our experience in developing apps, we are making limited investments with a smaller team to drive innovations. Overall, we achieved solid growth in 2024, but it is just the beginning of Cheetah's turnout. With animal spirit and the valuable lessons from operating as a public listed company serving users and business globally, I am confident in our team's ability to bring Cheetah to even greater success.

Thank you, Fuxiang. Hello, everyone on the call. Please note that unless stated otherwise, all money amounts are in RMB terms. In Q4, we once again accelerated year-over-year revenue growth and achieved quarter-over-quarter reductions in non-GAAP operating losses. Also, it represented the first quarter that we reduced our non-GAAP operating losses on a year-over-year basis since Q4 2022. Our strong performance in both top-line and bottom-line demonstrates Cheetah's continued business improvement since Q1 2024. In Q4, total revenue increased by 42% year-over-year and 23% quarter-over-quarter to $237 million. Revenues from our internet business increased by 49% year-over-year and 19% quarter-over-quarter, accounting for 68% of total revenue. This segment remains a strong cash cow, supporting our initiative in AI and robotics. Revenues from AI and others increased by 29% year-over-year and 33% quarter-over-quarter. Within this segment, our robotics business contributed about one-third of AI and other revenues, or about 10% of total revenues. Notably, in the quarter, revenue from our robotics business outgrew other businesses in this segment, including our advertising agency business and multi-cloud management services. On the profitability side, we continue to manage our costs and expenses while strategically investing in top R&D talent and leading sales professionals to drive our robotics business forward. As of December 31st, 2024, we had 935 employees and increased from 845 in 2023. About 40% of our employees are R&D talent and about 30% are in sales. From a business perspective, we continue to optimize our costs and expenses and focus our resources on AI and robotics. On a corporate level, our non-GAAP gross profit increased by 74% year over year, and 32% quarter over quarter, to 172 million in Q4. Non-GAAP gross margin increased to 73% in the quarter, from 59% in the same period last year, and 68% in the previous quarter. Non-GAAP operating loss reduced by about 18 million quarter-over-quarter and about 7 million year-over-year to 42 million. As we enter the anniversary of consolidating Beijing Orange Star and our investments in AI and robotic business, we expect our operating losses to continue reducing on both year-over-year and quarter-over-quarter basis. A segment operating margins for our internet business increased to 16 percent in q4 up from nine percent in the year ago quarter and ten percent in the previous quarter another highlight of the quarter is our strong cash flow generation we generated about us dollar 39 million in operating cash flow closing q4 with a net cash position of about us dollar 250 million additionally we hold about us dollar 100 million in long-term investments on our balance sheet we are confident in keeping our balance sheet strong in particular maintaining our strong net cash position going forward before opening the call for questions i want to emphasize our clear objective to drive revenue growth by achieving break-even and generating operating profits as soon as possible. Furthermore, we continue to carefully manage our net cash position as we have recognized that earnings and cash position are the two key metrics that we measure our programs both operationally and financially.

Everyone, for today's conference call, now we'll start the Q&A section. Everyone, for today's call, management will answer the question in Chinese. I think all the analysts on the line will answer the question in Chinese. And we'll have an AI agent who will translate the analyst's questions and also management's comments into English in another line. Please note the translation is for convenience purpose only in the case of any discrepancy or mentioned statement in Chinese as well. If you are unable to hear the Chinese translation, I think a transcript in English will be available on our IR website within seven working days. Operator, now we are ready to take questions. Thank you so much.

Thank you. We will now begin the question and answer session. To ask a question, please press star then 1 on your telephone keypad. If you are using a speakerphone, we ask that you 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 momentarily to assemble our roster. And today's first question comes from Thomas Chong at Jefferies.

Please go ahead.

SPEAKER 1 Our new year, 2025, because the Internet is changing faster, we don't have a specific data trend. But we think its growth should exceed the level of 2024. And here, you said new direction or something. I think we have an idea that the technology of AI has reached a stage of landing. We will use these new models of AI and the technology of agent to modify many of our original Internet products into AI. We also believe this will have a better support and acceleration effect for our entire Internet business. Okay, thank you.

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

Our next question comes from Ziping Zhao at ICBC International. Please go ahead.

Hello, Director. I am a PMT-14 operator in the supply industry. I would like to ask about the very strong tool product capability and promotion experience that Leibov has accumulated in the PCT and mobile network era. Thank you.

Okay, thank you. We think that today, as I have already said, the word agent is very popular. We also think that AI agents have reached a dead end on all lines. This is our judgment. AI agents within companies will definitely appear in large numbers and become companies within companies. And then, You asked about the co-development of the robot business. I think the robot business, in the end, it is the software's intelligent ability that determines whether the robot business can truly meet the expectations of the users, reach commercialization or large-scale commercialization. I think that in fact, in the past year or so, the efforts of AI Agents and the efforts of big models can actually be used very well in the robot field. Because in the end, their prices were all very similar. For example, in our speech, we also said that we will release an operating system called AgentOS, which is completely based on this time's large model and agent system, bringing the level of AI of this robot to a new level. This is not what we did before in the internal application agent of this enterprise, including helping our customers to do the agent skills. It's not easy to do a good job of understanding the capabilities of various models, including training models. Although everyone is talking about JDIF-Seq today, it can really be used well. In fact, there are still a lot of detailed technical points to be done. It should be said that this is the effort of the agent on our software and the experience of combining the robot, which is the operating platform of the robot of the agent OS that we are going to release.

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

Thank you. Our next question comes from GG Zhao with GF Securities. Please go ahead.

My question is about DeepSeq has driven the development of global large model industry this year. Do you think AI already possesses a condition to achieve scalability at lower cost, or do we still need to wait for some critical turning points? Thank you.

Okay, I think the appearance of DeepSeq is indeed very meaningful to the whole AI industry. But what I want to say is that the most important thing about DeepSeq is not only its low cost, but its performance in the push model is already at the top level of the industry. And this top level is completely open source and free of charge. So this makes it Every manufacturer can have a large model of autonomous intellectual property. So I have a point of view. I said in an interview not long ago that there may not be a so-called large model company in the future, or very few large model companies. And all companies will become application companies. How can users really solve their special practical problems will become a very key point this year. And because of this logical improvement of the reasoning model, we think this time has come. So we think that the entire AI industry, whether it is software applications or hardware applications, has reached the conditions that you asked about. Of course, it still needs some time. This time is more about the specific needs and engineering time, rather than the time of technical preconditions. Operator, please move to the next question. Thank you.

Thank you. Our next question comes from Yizu with Founder Securities. Please go ahead.

Okay, thank you for the opportunity to ask a question. We see that with the rapid development of large-scale technology, the market for wheelchair robots is rapidly expanding. My question is, which types of large factories do you think may enter the wheelchair robot track? Especially for companies like EV manufacturers, will they be able to integrate into this field through their own hardware manufacturing and intelligent technology? In this context, do you think that the accumulation and advantage of Leipzig can build a deep enough layer of protection and a prospect for the future market competition?

Thank you. Excuse me, are you talking about EV manufacturers? Yes. Okay, okay. At present, I think that in a fairly long period of time, maybe in two or three years, you may think that these big factories will not enter this field. Because, for example, compared to the field of electric vehicles, today's wheeled robots have a very good prospect, but compared to electric vehicles, it is still a too small market. And the competition in the electric vehicle market itself is also very intense. So I think within 2-3 years, there won't be any big companies entering this market. Because this market is in the early stages, there are still a lot of investments. Last year, we invested heavily in AI and robots. um um This is called open source or all kinds of this information or the technology is difficult to become a full-fledged layer or you This is really through your product to get the recognition of this user and then in the case of this user recognition, according to the ability of the user's demand for fast reception I think if we can catch this window in the next two or three years I think we can build a system of fast-tracking from sales to development. I think this is our real Hucheng River. And even when there was a big factory at that time, there were still a lot of details that were difficult to be copied all at once. Let me give you an example. Someone asked me, if the manufacturers of electric vehicles have no one to drive, will it be very fast to make robots again? In fact, I said that the problem here is also different. Why? Because the chip on the car may cost tens of thousands of dollars. The price of my entire robot is more than ten thousand dollars. I want to achieve in-house unmanned driving on a chip with less computing power than theirs. In fact, the technical difficulty here is not the visual framework, but whether you can use such a little computing power to really achieve reliable and independent indoor navigation. Operator, please move to the next question.

Thank you.

Thank you. And as a reminder, it is starter than one to ask a question. Our next question comes from Joanna Ma with CNBI. Please go ahead.

Thank you for accepting my question. I would like to ask you a question about modeling. With the increase in the number of models that are now high-quality and cost-effective, have you noticed that some medium- and small-scale companies have turned from modeling training to modeling training? Oh.

Yes, you just asked this question. I originally thought that small and medium-sized companies did not need to do model training. In fact, the big model is ultimately a basic facility. The ability of this basic facility can actually be used by everyone. In the future, my point of view is that this point of view is actually two years ago. I haven't changed much today. In the future, the big model training is a big factory. He will train his own big model because they will think that without their own big model, there may be no safety boundary, right? Otherwise, it will be stuck in the neck. Moreover, the open source model will continue to grow. And now it looks like, right? I used to think that the performance of the open source model would not belong to this I think... Now a lot of companies are going from model training to reasoning, which is correct. Because in the future, the computing power of each company will become a standard like manpower. You have a lot of work to do, which is actually to complete the computing power, instead of the manpower. I think the first thing that has changed is our development model. We have been using a lot of AI in our design and programming, but it's not exactly the same as before. And in fact, we are reducing our dependence on the so-called professional high-tech people. We also built a research center in Xi'an at the end of last year. In fact, with the help of AI, some young people in such two or three-tier cities can quickly have a strong problem-solving ability. I think there are a few things that have an impact on us. One is that we will turn AI Agents into the core of our team. Today, robots are a very widespread concept. There are man-made, clean, heavy-duty robots, and so on. But our advantage lies in being able to interact with users and go on a smart mission. This is the reason why we launched AgentOS. Secondly, in essence, these Internet applications are all AI-based. We are now working closely with our traditional software to meet users in a better way using AI. So I think for us now, in the future, our business, due to the combination of AI, we will have more confidence. Thank you.

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

Thank you. Our next question comes from Min Sun at UOB. Please go ahead.

Hello, thank you for giving me a chance to ask a question. I'm Min Sun from UOB. I would like to ask about the development of deep-seq V3 and RIE language models. How do you see the introduction of these models to the impact of AI application competition? In the next few weeks, how will Leopold use these technologies to promote the company's progress, especially to improve product competitiveness? Thank you.

. . . . . . . . . . and even the large model for fees are almost the same. So I think the impact of this competitive pattern is that in the visible period of time, the innovation of DeepSeq will continue, and then the competition for the performance of the entire large model will become more and more white-hot. In the end, there are only a few manufacturers left in the large model. Most companies will turn to AI applications. Another thing is that DeepSeq is a very important step forward for China's AI agent business. In the past, if you were doing AI business in China, there was a problem. You can't use those top models for applications. Now, using DeepSeq, the model experience is almost the same as the top model. So I think there will be positive positive factors for this wave of Chinese AI business. I have already talked about how to use this technology to promote the company's development in the next few seasons. I think we have actually spent several years in the field of robots. We paid special attention to the interaction point in the first place. We made this voice-to-speech robot, including to listen to your instructions to complete some actions. But frankly speaking, at that time, it was limited to the ceiling of technology in the entire industry. So our experience of voice-to-speech is very difficult to achieve in that kind of public, . . . . . . . . . . . . So, I think this is not only a high-level experience, but it will make this type of robot really start to be able to be used on a large scale. It's a bit like this AI chat software that can grow so many users in a blink of an eye. We've seen a lot of videos before, but most of them are staged. The real thing is that you can interact with people anytime, anywhere, and solve your problems. This kind of interaction experience is almost non-existent. But today, because of this big model, it allows people to understand. So whether it's noise cancellation, or a lot of people getting in the way, we've all had, I should say, a relatively good experience. So we feel that we can to make the introduction-type and interaction-type robots become a fast-growing track. And as I said, on traditional software, we will turn AI Agents into a basic technical framework, so that they can be more intelligent when it comes to solving user problems and meeting user needs. I think that's why we have more confidence in both robot and Internet business this year. Thank you.

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

Absolutely. Our next question comes from Ben Kaishal with CICC. Please go ahead.

Thank you, Guan Jiteng. I would like to ask about the rapid development of the AI industry, the opportunity for AI to develop in the future, and the opportunity for AI to develop in the business model. And compared to the previous quarter, has the company made any adjustments to the AI business strategy? Thank you.

Yes, you just asked this question, I think I have made it clear. I'd like to add that the robot industry is very hot right now, but it seems that everyone's focus is on humanoid robots. My point of view is the same. Humanoid robots are not only a technical difficulty in software technology, but also a complexity in hardware technology. So it's actually a long way to go when it comes to large-scale commercialization. Today, humanoid robots seem to have more scientific elements. Thank you very much. So at this point, we feel that due to the maturity of this large model technology, we feel that we have reached a point. So we think that this year, in fact, we can't say that there are any adjustments. We still follow the route of this robot. It's just that we think that in the language interaction, reception, explanation, including promotion, translation, such a robot market should be bigger than last year or a few years ago. We have some internal Of course.

Our next question comes from Chengru Li with Goyang Securities. Please go ahead.

Hello, Mr. Guan. I would like to ask, how does the data used in real-time scenarios respond to the acceleration of technology? And when the data wheel is used to promote the business of the telecommunications engineer to a wider range of industry applications, will play such a role.

Thank you. Yes, based on the question I just answered, I would like to say that today, in fact, what we really do before is two-way technology. There is another one that has been predicted, so I won't talk about it. One of the two aspects is the language interaction aspect. Let me give you an example of this interaction aspect. In fact, our general environment noise in many public places may be 60 to 70 times. Thank you. Thank you. Thank you. This is an example of our own voice recognition model in this environment. It is much better than some models that you may have heard of. The recognition rate will be much better. This is actually a part of data delivery. And now that we're on AgentOS, it will respond to more complex language interactions, visual interactions, as well as multi-person interactions. In this response, there will still be some bad cases. So when we push it to the market, I believe that with the continuous data recovery, we will definitely solve this problem better, so that we are getting closer and closer to the real person in this experience, and we can know more and more whether you are talking to him or you are chatting with the person next to you. Is the intention of saying these words really to let the robot do something or to talk to the person next to you? . . . . . . . . . . Maybe some of you know, for example, in the past, when we had a restaurant, we had to put a QR code on the ceiling. If there were no such hard indicators, it would be easy to get lost, right? It would even be easy to hit people or something. But now, with the progress of vision, the number of data is constantly increasing. We have now cancelled this top-end thing called putting a QR code. And as the number of data continues to increase, we will also strengthen the self-discipline in space. Thank you. Operator, please move to the next question.

Thank you. And if you'd like to ask a question, please press star then 1. Our next question comes from Ben Shering Wu with TF Global. Please go ahead.

Thank you, Director Chen. My question is, I've recently seen that whether it's Yuanbao or Doubao, some of these AI assistants, including Copilot and ASO, have a very fast development. I just want to ask you, um um um

Will the appearance of AI agents have a major impact on the application ecosystem? I think the appearance of AI agents will definitely have a major impact on the existing application technology route. Will it have a major impact on the existing ecosystem? I think this also depends on the reactions of these ecological plants. The faster the response, the more likely it is to be blocked. The slower the response, the more likely there will be changes. I think this is a bit similar to when mobile Internet came out in the past. Can traditional Internet companies get passports on mobile Internet? Or can they move forward on mobile Internet? We also see that different factories have different There are different, I should say, later developments. But I think this may not be destined, but it depends on everyone's investment in AI, and the speed of action in productization on AI, including short videos like Douyin later, right? Those video websites must have AI Agents This should be said to have formed a wide consensus, whether it is the entrepreneur or the factory. So how does this change? To be honest, it's hard for me to predict. Because sometimes it depends on the work of each company. But the second question you asked is, will it replace many traditional apps now? I think this technology will definitely replace traditional apps. When users are used to solving problems in one sentence, they will no longer use traditional click-through methods, because natural language is not needed to learn. In the past, when we used any kind of app, we had to learn how to use the app interface. This is why the elderly couldn't use the health code at that time, right? They didn't know how to drive. And when this natural interaction became mainstream, that is, you don't need to learn anything. The machine will adapt to you. This is what I said before. It's called the first time the machine turns around the person, not the person turns around the machine. So in this context, the AI agent technology will definitely replace many traditional ones. It depends on whether the applications of these apps are fast enough for AI-based technology, right? It doesn't give new entrepreneurs or new players enough time. I think this is also very important. Anyway, I think the existing applications are indeed suitable to use AI-based technology to do a little bit more. This experience will definitely be different. Then ask about the super applications of micro-systems. I can't answer this because it depends on the future of WeChat. Will it be open in this regard? Will it be the first big thing to do? Will it continue to launch such experiences? In fact, I personally think that if this wave of AI agents is too slow, even some previous ecosystems may encounter problems if it is too slow. Let me give you another example. Today, if everyone is used to using these AI apps, you will find that your dependence on traditional search is greatly reduced. At least for me personally, I have not used this search for a long time. Basically, I use our first secret tower or some other apps. It will help you sort out the answers you want directly. So I don't think Thank you. I'll turn your please motion to the next question.

Thank you.

Of course. And once again, if you do have a question, please press star then 1. Our next question comes from Yunpeng Diao with Guita Junon. Please go ahead.

谢谢管理层接受的提问。 能否请管理层回顾一下猎豹2024年在机器人业务取得的一些成就, 以及2025年在这方面是如何规划,以及有什么样的目标? 谢谢。

I know. Our achievement in 2024... In fact, we have exposed ourselves in the field of AI and others, right? Yes. We... Yes, we think that in 2024, it is a success. We have opened up some overseas markets. We now think that the demand for service robots overseas is very obvious. Of course, I have to say that the overseas market is not the same as the one that used to do APG. It needs a cycle for channel construction, for agents, and for salespeople. We think that in the past 24 years, the entire robot industry has seen a real growth in demand from the market, including in Europe. This is a very slow-changing market. Today, due to the use of people, they can clearly see this demand. In 2025, I think our goal, as I just mentioned, maybe one of our major goals this year is to be able to achieve a differentiated competition with many robot manufacturers. I don't think we will be in a single so-called low-end market. Because the domestic market needs to be more flexible, right? In terms of price war, especially in the domestic market, price war is very fierce. What we want to achieve this year is to make a smart and interactive strategy as the core of such a development strategy. To really take this market, I think our product performance and product experience in this market are all leading. And this market, it should be said, following the big model, can officially open this year. Thank you, Vice President. Very clear. Thank you.

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

Thank you. And once again, if you do have a question, please press star then 1. Our next question comes from Yanlong Chen with CITIC Securities. Please go ahead.

Hello, Mr. Guan. I'm Yanlong Chen, a researcher at CITIC Securities. I just heard you talk about our future differential strategy. Because in the past, I understood that the market thought that such as in hotels and restaurants, as well as some exhibition areas. In fact, the entire market space is relatively limited. From the point of view of innovation, how do you think we at Liebbao are going to achieve some breakthroughs in these areas to open up the market space? And do you think that with this kind of AI drive, we can get away from this breakthrough? What is the landing pace behind it?

就是这个问题想请教您一下,谢谢。 好的,好的,那个就是我赞同你前半段的结论,不是特别赞同原因,就是轮式机器人的这个空间有限并不是因为它是轮式机器人,而是因为之前的机器人的智能化水平有限。 In the past, when you saw a lot of restaurants and hotels using it, it actually required a lot of deployment work. And its tasks were very monotonous. It had no way to... It had to complete a more complex solution-level problem before it could really be used. So its entire intelligent level was not enough to limit it. So it could only complete very fixed tasks, such as point-to-point landing tasks. Right? So I think the biggest problem of the humanoid robot before, the biggest problem of the humanoid robot before, basically, including the entire robot industry, we are not talking about the robot industry, but the industrial robot industry, because the industrial robot industry has been built for decades, its automation level has been enough to satisfy the industrial production line. This is because the intelligent level is not enough. The intelligent level is not enough to make it able to complete the task is a very limited point. But this year we added this, The agent OS is a new multi-mode interaction capability and its task planning capability. I think its intelligence level can go up to a very high level. After going up to the level, in fact, the work it can do will become like this. You can say it can't be said to be like a human being. This is ridiculous. That's close to a certain extent. It's close to the ability of a human being. Look, I'll give you a very detailed example so you can understand. For example, in the past, a restaurant operator could only go from table No. 1 to table No. 1. The waiter had to take training, he had to order, he had to do this. And it was just such a thing. He didn't know what the customer's reaction was. But today, with our system, actually, first of all, to go to table No. 1, to send this over, this kind of thing, the waiter doesn't have to take training. Just like what people say, just send it to where it is. Second, it can also carry out tasks that former waiters could do. For example, when you're about to close a table, you can talk to the other tables. You don't need to write any code to accomplish this. It can be done automatically through Aging. Say, now there are fewer people, you can go to the door and say a few words. This can also be done. In the past, for non-IT businesses like restaurants, you had to configure all kinds of things, In fact, the hidden cost is very high. So, of course, this is just a small example, because this is a press conference. We will also have some videos about these efforts. If you are interested, you can go and have a look. So what I want to say is that, in essence, it is not the wheel itself, but the intelligent level is not enough. So the biggest investment we have made this year, or since last year, is to improve the intelligent level of the robot. As long as the level of intelligence is sufficient, it can be applied in many places, and it can greatly increase its space. Of course, in the long term, I think, although I don't see the human form, but I think that with the addition of this mechanical arm and physical space, to achieve some of these tasks, I think this is also the next step in the expansion of the machine, which looks like it will soon be a feasible route. All right, thank you.

Okay, thank you. Operator, please move to the next question.

That does conclude our question and answer session, ma'am, so I'll turn it back over for any closing remarks.

Okay. So thank you so much for joining our earnings conference call today. If you have any further questions, please let us know. You can either give us a call or send us an email for your questions. Thank you so much.

Thank you. Thank you. This concludes today's conference call. We thank you all for attending today's presentation. You may now disconnect your lines and have a wonderful day.