Cheetah Mobile (CMCM) Q4 2025 earnings call transcript

Thank you, Operator. Welcome to Cheetah Mobile's fourth quarter 2025 earnings conference call. With us today, our company's chairman and CEO, Mr. Fusheng, and our company's director and CFO, Mr. Thomas Hien. Following management's prepared remarks, we will conduct the Q&A section. Please note that the management's prepared remarks are presented by AI agents. Before we begin, I refer you to the safe harbor statement in our earnings release, which also is applied to our conference call today. As we will make for word of his name.

At this time, I would now like to turn the conference call over to our chairman and CEO, Mr. Fuzong. Please go ahead, Fuzong. Good evening, everyone. Thank you for joining us. In 2025, we finished stabilizing the business and built a stronger foundation for Cheetah Mobile. During the year, our total revenue grew 43% year over year, driven by continued growth in both our Internet business and AI and other segments. In the fourth quarter, AI and others already accounted for half of total revenues, reflecting the increasing contribution of our new growth initiatives. More importantly, we achieved four-year non-GAAP operating profitability, our first time in six years. Our internet business remained resilient in 2025, generating approximately RMB 460,000 in adjusted operating profit every working day.

This consistent operating cash flow forms the financial backbone of the company and allows us to invest in robotics and the AI in a disciplined and sustainable way. Our second highlight is robotics, which is emerging as a key structural growth. For a four-year, robotics revenue grew approximately 31%. In the full quarter alone, robotics revenue reached about 60 million RMB, up 94% year-over-year and 43% quarter-over-quarter. Our voice robots in China achieved 100% year-over-year growth for three consecutive quarters, accounting for high single digits of the fourth quarter's total revenues. This progress is driven by our strategic focus on core strength in voice robotics and the integration of AI agent technology to enhance product experience. We are now seeing our voice robots become a must-have solution in resections, guided tours, retail environments, hospitals, and service halls as they deliver proven, measurable values.

We recently introduced a new version of our voice robots, which comes with building skills like guiding, patrolling, and advertising, enabling end customers to start using them right away. Our robotic car business, mainly serving overseas markets, is making up high single digits of the first quarter's total revenues. We focus on long-term demands from research institutions and the R&D teams that value openness and customization. This customer base is tricky and repeatable, supporting long-term demand. And building on our proven indoor autonomous mobility technologies, we are introducing a smart wheelchair targeting developed regions such as Western Europe and North America. This product is positioned as a premium solution for users who value safety, independence, and confidence in daily mobility.

We are seeing a clear shift in demand as users increasingly value safety, assistance, and intelligent features in mobility products, While scalable solutions in the market remain limited, by applying our experience in service robots, we are able to meaningfully improve the user experience. During my own reasons recovery, I personally used our smart wheelchair and felt a clear improvement in safety and convenience. Importantly, we can deliver these benefits without significant increasing the cost compared to traditional high-end electric wheelchairs, making this a more practical and accessible product for users. We have entered into framework agreements with established mobility brands who will manage branding, distribution, and after-sale services. Initial shipments are expected to begin in the second quarter of 2026, representing an early stage commercial validation of this product category.

Across the industry, more companies are starting to test and deploy service robots. We believe the next one to two years will be a validation phase where ROI and reliability will matter most. You don't need a robot that looks like a human. You need a robot that works every day, delivers measurable value, and is easy to operate at scale. This is exactly where our current products are positioned. Our internet business remains strong, generating steady cash flow, which allows us to invest in AI in a disciplined and sustainable way. For more than a decade, we have built utility applications serving hundreds of millions of users. This product DNA shapes how we approach AI. Rather than competing in model development, we focus on turning AI capabilities into practical tools that help users complete real tasks.

During the Chinese New Year, I spent a lot of effort experimenting with an AI agent system built on the open call framework, Starting from a single agent that could barely complete basic tasks, the system evolved into a multi-agent team capable of running tasks continuously. In one scenario, the system generated personalized New Year messages for more than 600 colleagues and managed the entire sending workflow automatically. What we see emerging is not simply a new AI tool, but a new way to organize digital work. AI agents can automate entire workflows from information gathering to processing and distribution, significantly improving productivity. Building on these learnings, we introduced EasyClaw, based on OpenClaw, an open source agent framework for both domestic and overseas markets.

EGclaw is our AI co-worker platform that helps users create and deploy task-oriented AI agents capable of executing real-world tasks autonomously. At this stage, we focus on execution capability rather than skill. We are already seeing a continued increase in user engagement as reflected in the rapid growth of our total token usage. We are building EasyClo into an agentic operating system that changes how users interact with software and machines. By integrating EasyClo into our PC products, we are improving user experience and driving higher conversion and ARPU.

In robotics, EasyClo allows users to program and customize robots using natural language, lowering customization barriers. This helps us deploy faster, reduce costs, and scale more easily, making our products more competitive. Some investors may ask how we compete with our training foundation models. We believe the real advantage in the agent's error lies not in the model itself, but in the systems built on top of it, including task orchestration, tool usage, and cost management. By leveraging open ecosystems and leading APIs, our products can evolve as models continue to improve.

Finally, our global DNA remains a core competitive advantage. We continue to expand both our AI tools and robotics businesses internationally with a disciplined approach. Looking ahead to 2026, we do not provide specific financial guidance, but we see continued structural improvements. We believe our robotics business will maintain strong growth momentum as commercial validation deepens and become a more important part of our revenue niche. At the same time, AI-enavored products will gradually enhance engagement and monetization efficiency across our software ecosystem. We will increasingly apply InCernalate to accelerate development, aiming to further improve operational efficiency. As we grow, we will continue improving transparency and disclosure. Credibility is built to data, and our focus remains clear. Execute discipline, and as results, compound over time.

Gita is entering its next phase of development, combining digital co-workers through AI agents and physical co-workers through service robots, supported by real operating cash flow and disciplined financial management. We are building a foundation for our next stage of growth. Thank you.

Thank you for showing up, everyone, and thank you for tuning in. Next, otherwise stated, all financial figures are presented in RMB. 2025 marks a year of meaningful operational recovery and improved financial history for Cheetah Mobile. During the year, we continue improving operating discipline and cost structure across the company. We concentrated resources on commercially validated use cases in robotic products and practical AI applications, while leveraging open source ecosystems and third-party models to improve warranty efficiency and optimize infrastructure costs. This approach allows us to accelerate iteration without significantly increasing fixed costs for the full year 2025.

Total revenue grew approximately 48% year over year. to 1,150 million. Although we recorded a GAAP operating loss of $179 million for the year, this represented a substantial improvement compared with operating loss of $437 million in 2024. On a non-GAP basis, operating profits reached $14 million compared with a non-GAP operating loss of $232 million in a prior year, reflecting improved operating leverage. What ended the year was $215 million, United States dollars, cash and cash equipment on hand.

Turning to our segment performance, our internet business continues to serve as a stable cash-generating platform for the company in 2025. Revenue from the internet business increased 19% year-over-year to $615 million. In addition, we observed that many users subscribe for periods longer than 12 months, reflecting the recovery nature of our utility applications and the and strengthening revenue visibility. In terms of profitability, the Internet business generated approximately RMB $115 million in adjusted operating profit in 2025, maintaining healthy margins and strong operating cash flows. As Gysha mentioned earlier, the Internet of Business generates roughly RMB, $460,000 in adjusted operating profit per working day, which provides predictable cash flow to support strategic investments and new initiatives.

Looking ahead, we expect the Internet of Business to remain stable and profitable while continuing to provide financial flexibility for the company to invest in long-term growth opportunities. Turning to our I and other segments, revenue from this segment increased 85% year-over-year to $535 million in 2025. As a result, this segment accounted for 46.5% of total revenue compared with 35.9% in 2024, reflecting the growing contribution from our emerging businesses. Within this segment, the robotics business continues to scale since the second half of 2025, making up 27% of this segment's revenue and 13% of tour revenue in 2025. Robotics revenue increased 31% in 2025, driven by deployments of voice robots in China and continued demand for robotic arms in overseas markets.

Other businesses, namely overseas advertising agency service and multi-cloud management platform within this segment, also contributed significantly to revenue growth, benefiting from increasing overseas expansion by Chinese enterprises. At the same time, we continue to improve operating efficiency to more selective investment and disciplined cost control for the full year. Adjusted operating loss from the AI and other segments reduced by 42% year-over-year to $274 million as we continue scaling the business while maintaining disciplined investments, turning briefly to the first quarter performance. Total revenue reached 309 million RMV representing a 30 year increase and a 7 quarter increase While the Internet revenue declined slightly year in the fourth quarter it increased quarter as we continued shifting toward a subscription business call In addition, user subscription revenue within the Internet segment increased 32% year-over-year and 16% quarter-over-quarter, as we chose to focus on the subscription business model, which supports a healthier product and user experience.

Revenue from the AI and other segments reached 153 million IRB, accounting for nearly half of total revenue in a quarter. With this segment, robotic revenues increased by 94% year-on-year and 43% quarter-over-quarter to about 19% of the fourth quarter's total revenue. Other than that, our revenues from overseas advertising agency service and multi-cloud management platform also contributed to this segment year-over-year growth. On a non-GAAP basis, the company generated operating profits of $15 million in the fourth quarter, compared to $42 million operating losses in the same period last year. We believe the improvements we achieved in 2025 reflected structural improvements in both our cost structure and revenue mix.

Looking ahead, our priorities remain clear. Disciplined growth, continued improvement in operating efficiency, balanced and disciplined capital allocation. With stronger financial discipline, greater strategic focus, and increasing contribution from our emerging businesses, we believe the company is entering a more stable and predictable operating phase. Thank you.

Okay. Thank you. So much for your question. Good question. Let me answer it. I think for AI agents in 2025, we have started to make a play out. We have always believed that not only large model companies, but also application companies have opportunities in this way. So in 2025, we launched this Vitality LLM, which is a desktop widget.

After OpenCloud launched, it has some advantages in architecture design. So we also quickly learned from it and launched our own cloud. We are one of the first domestic companies to launch open cloud-related products. I think the concept of open cloud has gained popularity in China after the spring festival. Although we started at the end of last year and have launched such products, we now have our own user base. This user base has shown that we have a lot of different products.

Strong user scenario demand. Which workflows are you asking about? It's not easy to go into detail, but today I think in the easy cloud scenario, we have targeted user demand to develop some skills, which are becoming more than competitive advantage. Frankly speaking, I think it's not a barrier now, but we definitely have our own unique features in this regard.

And as big companies are accelerating to enter this field, I think AIGC is a very broad field. Essentially, we're constructing the entire software and APP industry. When big companies enter, they will gain recognition from a wide range of users. Our differentiated strategy is to combine Cheetah Mobile's existing AIGC R&D capabilities of top tools and its emphasis on user experience details, as well as features like multi-cloud aggregation and many high-quality partnerships with cloud vendors, both domestic and international and our global expansion, namely our global advertising business where we've accumulated many to be user groups. Therefore, our core differentiated advantages are reflected in true vertical industry solutions, which is eagle eye cloud that we hope to offer to users.

Not just a tool that can connect to LLM, but also an agent that can run on users desktops. But also truly targeting vertical user scenarios, especially 2B user scenarios, and present an easy-to-use solution, so we are constantly advancing in productization, although it's not yet March. We have already done a lot of surveys and area exploration in this regard. In the medium and long term, we and the big companies, because we have repeatedly discussed internally, we think our advantage over the big companies is that We should be able to better target the needs of the vertical scenario to make products that are closer to users. In the medium and long term, we've positioned UFOO Cloud on AI agents.

I agree with the judgment of Peter, the founder of OpenCloud, that 80% of apps will disappear in the future. OpenCloud will become a very important carrier for the agent. I will now turn the call over to the agent. Whether it's called OpenCloud or EagleEye, or recently, Apple has also launched their mobile version of CloudCloud. This kind of product form should be a very core one. So now Cheetah is all centered around EagleEye, whether it's our 2C user group, 2B customer group, or our robot business. We are at EagleEye landing. So it will be a very important strategic product.

It's not like what some social media says that receives the opportunity well, but after years of tool development, we have a better judgment of the industry, including as we started in advance, and we will stick to it for a long time. Thank you.

Thank you for your question. I think the evolution path of this industry pattern may be very cool for everyone in this industry today.

After all, the developments need of AI and the degree of fierce competition are unprecedented. However, I still have a few judgment. First, I think the first question today, which layer will be more urgent today? Where will the economic value today be concentrated? I think for quite a long time, maybe in three to five years, whether it is public cloud, computing platform, or basic model vendor, or the one closer to the user side, there will be a place. This is in the near future. For the next three years, it will definitely exist. Because today you can see that the competition for models is also very fierce, and the models also show this diversity. Like the top models in the United States, which have a high call rate, but now due to the strong demand, the call rate of many domestic models has also risen very quickly, and they are actually suitable for different tasks.

Therefore, the competition model for different tasks will continue for a long time. And in this process, which application can bring better experience to users and entry, it will also form more cooperative relationships with the model. In fact, I think in this layer, I have a judgment. That is, I think in the future application later, it will be difficult for any one company to completely control the monopolistic entry.

For a rather long period of time, there will be a situation where different industries and user groups have different product positioning. So, as I mentioned in the previous question, our main target users today will be more to be, even in the overseas customer group, which is not a tech-leading company, but a traditional company far from tech, will be a key focus for us. We believe that in this user group, we have the opportunity to take the lead.

For example, the impact on the industry you asked today, IAM or traffic entry attributes are easy to occupy a core position, but it is more from the perspective of a wider user. As I just said, today you can see the user layering of products like this is very obvious. The difference between users who pay less and more, the difference in their demand attributes is huge. Although it is a product, the overall user demand attributes and its usage characteristics are very different. So I think this kind of product for the general public, in terms of 2G user demand, it is not easy to meet in a considerable long stage, because we have to take into account both implementation and permissions, as well as user cultivation and education, and token costs.

So we still think that if we fight this battle well, we can occupy some vertical users like us. And there is one more thing about users.

This product, in one word, is lobster, which is not used immediately, but needs to be nurtured. When it accumulates enough usage habits on your platform, it is usually not easy to migrate, This is why today, although large language models are very powerful, application companies will still have many opportunities because they are closer to the user side and match the user habits more. So overall, we believe that easy cloud is the best opportunity we have seen in the software field over the years. Of course, we will adhere to our independent development path, but embedding it in the super platform ecosystem in the form of functional modules does not rule out that we have some capabilities, including some core skill capabilities and some platforms to cooperate.

But I don't think this kind of product has a so-called universality. In fact, if users use it well, the cost of migration is very high, and they will not move it just because of habit.

Yes, this is similar to a product called Manist, which developed rapidly at that time. And when many users were imitating it, it continued to grow, right? Later it was also acquired by Meta. Since we are from the tool and application background, we are very sensitive to products with vertical attributes that can meet user needs and continue to develop. I'm done. We can talk about the following.

Your question is quite profound. I think for Asian products, I really particularly agree with the point that to judge the real value of users of an Asian product today, it lies in its usage volume. The data of this terminal usage volume should be our core measurement indicator. We have also seen that since the spring festival, especially after I did some promotion about my own use of open password, our token usage has definitely increased a lot. I don't know if it can be disclosed. I mean, the token usage today, since last year, we just started to enter the field of AIG.

So we didn't have much expectation and concept of the call volume at that time. So we thought that a volume of 10, indeed, with the help of AI, Many positions that used to require many years of work experience can now be done by young people with the same or even more efficient output. In terms of the overall cost structure, as we have mentioned before, we have opened up branches in Xi'an, Chengdu at the end of Ladders here, which has significantly reduced our road costs. We are now looking at the cost of the road. Our road efficiency has also been greatly improved. The launch of EZCloud today and the launch of the wheelchair project in the robotics field are based on the fact that we have not increased our road investment.

In other words, while our legacy products continue to grow, We still have the strength to invest in new areas. I think this is all thanks to the internal efficiency improvement brought by AI. This year our specific improvement directions are to strengthen the in-depth application of Eagle Life. And we have already done department level, executive level, executive level, and department leader level. I will now turn the call over to the Cheetah Mobile.

I think... This wave is a very obvious feature of AIG's product, that is, users come with their own revenue, which means that you may also use our product now, and we charge based on the number of calls. We have cooperated with various model manufacturers. Whether it is these semantic models or overseas ones like Google and Amazon, because we have this cloud aggregation platform, we have relatively in-depth cooperation with them. So we have both domestic and international versions. And currently, this version is based on what we call point consumption. In fact, it is the call to each company's token.

So today is also a natural model in terms of commercialization. So U3 Cloud itself has revenue. Although today the model is provided by the large model manufacturer, in terms of model invocation, we are still in the initial stage. We will not consider profit, but use the token called by the user, and then replace it with a point model, but we did not use a subsidy model, or give too many tokens for you to use, and then grab users, because we think that when users really use it, its token consumption is very large. And this kind of subsidy model is not suitable for this kind of product. We still need to find users and serve them well and provide them with better product capabilities.

In this way, when it improves its efficiency, it is very willing to pay for it. So I think EZCloud's commercialization is an actual model. There is no need to consider too much.

I think everyone thinks of the word. Maybe you have a good system that is, but I think without a breakthrough in a single product, there is no ecosystem. In the process of going global today, we are indeed considering making EasyLink a good product. As long as it is done well enough, it will generate a multi-product ecosystem. Because EG-Link is not a simple product, it is more of a technical framework.

When this technical framework or product can be accepted by users, the entire ecosystem of that computer will be aligned with this product.

So today, regardless of which business or traditional tool for jobs, Including the team of our previous desktop software, whether it's the so-called EZCloud for top or our EZCloud for top or our advertising team going global or our cloud team, we are now integrating all our services on EZCloud because its performance for users is different. Different demands require different skills or different demands require different agents. This is our idea today. As long as we can truly customize EZ Cloud for different users or personalize it, and it will form an ecosystem in the future.

So if we are at this stage, we are actually making a breakthrough in software products, but we will build different agents and skills on this single point. So, these different agents and skills are for both overseas advertising clients and overseas clown clients. They can all be good served within such a product system. This was indeed very difficult before because the service systems of cloud services and overseas advertising and the product system of the CND users are different. To be honest, this was a problem that troubled us a lot in the past because it was too scattered. But now it can be unified into a product system. We just need to install different skills and agents.

So this is our overall idea. Another point you mentioned is good. I think in the process of going global, we have some unique advantages. Whether it's a client in the cloud, Amazon cloud, including Oracle cloud, our top partners, and MetaAd, are all first-year agents, and also our past team experience. So our EDK has both domestic and overseas versions. We hope to deepen all our services within such a product.

This is a good question. Hi, I'm Thomas, the CFO of the company. Let me answer first and see if Mr. Poo has any additions. First, let's talk about the first question. Let's review the revenue growth in 2025. I think this shows that in the past year, our company's overall strategic focus on business structure optimization have yielded results. Specifically, as we mentioned in our presentation, on the one hand, we have been continuously promoting the subscription-based transformation in our internet business. This has improved the stability and visibility of our revenue. On the other hand, in AI and other sectors, including the robot business and the total service business overseas, almost all our business plans achieved rapid growth in 2025.

The different business plans together drove the overall revenue growth. We think from the current business structure, our current business combination is relatively healthy, and the contribution of new business is continuously improving, laying the foundation for our long-term development.

Regarding the main source of growth in robot revenue in 2026, General Manager Blue mentioned this in the previous question, including voice robots and wheelchairs.

Let's summarize the business lines of robots. Firstly, the business of robotic farms. In fact, after the consolidation in 2025, its operations have been very stable. Meanwhile, the demands from industrial scenarios, research institutions, and R&D teams are continuous with a high recursive rate. It is expected that the robotic-owned business will maintain a steady contribution in 2026 with a small increase. Secondly, our voice service robots in China have entered a stage of large scale deployment and have been landed in various scenarios, including reception, explanation, elderly care, and hospital. Additionally, we will soon launch a new product called Xiaming Live, which combines some Asian technology and is a new voice service robot that can be used out of the box.

And we expect to continue to grow in the domestic voice service robot market. The third point is just mentioned by Histafood, including what we mentioned in the manuscript. Our newly launched Smart Wheelchair product will also achieve commercial landing this year. Actually, this product without generating any revenue in the past. Now we have already cooperated with mature overseas mobile travel brands, and have a personal cooperation in Reva. And it is expected to become a new source of incremental levity for the robot management, starting from the second quarter. The current framework cooperation agreement has already established the foundation of 30-scale and the third batch of shipments are mainly concentrated overseas. Yes, regarding the voice robot in the overseas market, I learned that last year we mainly focused on the voice robot scenarios in China.

And after we think the scenarios in China are mature, we are actually doing some overseas markets, especially some attempts in high-end We hope to extend more clients in overseas markets this year, bringing new revenue growth. That's my answer. Let's see if President Kuhl has any additions.

Let me add something. I feel that I have particularly systematically talked about So why do you want to make intelligent wheelchairs?

Relatively after so many years of robots, we have carefully considered and investigated the market. We have made a strategic decision, which is a strategic level product, because the growth of voice robots, I think it grew well last year, and it will also grow this year. But what is the difficulty of this kind of product The difficulty lies in changing the users existing usage process Actually today actually today change the application of robots in China Whether it is enthusiasm or acceptance, it's at the forefront of the world. So this is the current situation. I always think that this partner is called Zunin Start. One, you have to design a new product form, where everyone has to figure out what the robot is in their mind.

This is hard to figure out, or our education doesn't think about it. Everyone has a different idea. So, on the one hand, you need to define the product form, and on the other hand, you need to guide users in the task workflow to put money into it. So we can see that the truly fast-grown ones are robots, right? Either it's a traditional robotic arm that replaces labor, which has been developed for a certain period of time, or another sometimes, which is like some particularly simple machinery, such as sleeping robots. So when we made robots in the past, it was quite difficult to get the market to accept a new product, and it changed its entire working mode.

So this is what we later repeatedly thought about. What kind of product can not only make good use of robot technology, but also be quickly accepted by users. The reason why we chose the smart wheelchair market is that we think the smart wheelchair is actually a robot for educating people. The Cheetah Mobile, the Cheetah Mobile, the Cheetah Mobile, the Cheetah Mobile, the Cheetah Mobile, the Cheetah Mobile, the Cheetah Mobile, the Cheetah Mobile. So its entire technical chain, the entire technical chain of the wheelchairs, whether it's the chassis, motor, three electric systems, and the intelligence part we want to do, including navigation and obstacle avoidance, are actually very close to our entire technical path and the usability, which is very high.

The second is that its market size is quite large, much larger than the current robot market size. The global smart wheelchair, or electric wheelchair market, should be over 2 billion US dollars, and it is growing every year. Especially the smart wheelchair market is growing in the fastest. According to third party reports, it can grow by more than 10% globally, which is actually a fast growing market globally. So in this market, it can't be said that this industry has a long history. It has a history of over 100 years, but it is relatively backward in terms of intelligence. Therefore, we believe this product can very well combine our technological paths and our understanding of the technology of wheeled robots, including the user experience that we use to excel at.

And it should have a relatively stable and certain market size. I won't go into further detail. In fact, our understanding is that transforming the delivery robot into one secondary passenger is why we ventured into the smart wheelchair market after over a year of refinement, as you might know. Many startups are now securing VC funding to develop similar products, and some such companies have already emerged. However, I want to say that previously, we had tens of thousands of robots, including those for export, which moved in various scenarios. We have accumulated substantial data about scenarios, as well as various rates of scenarios. Therefore, we can use a highly cost-effective solution to greatly enhance the intelligence of the entire wheelchair.

This is inseparable from the accumulation over the years. Therefore, I am quite confident in the growth of this product. Of course, as I just mentioned, its production cycle and overall certification will only start shipping in the middle of the year. Now, samples have already started to be provided to our suppliers and distributors, and the first batch will soon be offline, but the real sales and launch will start from June. For example, we only have half a year for sale, and this is the first model, so I think in the long run by next year the growth of this product will be a good start. Okay, these are my additions to this, because whether it's for these additions, I just heard someone speak more clearly.

Thank you, Vicky, for your question. I am Calvin, and I will answer it. First of all, entering 2026, one of our important internal management goals. We still to take the management discipline and input-output efficiency as core internal management goals. Our overall strategy is to advance business development while maintaining a reasonable profit pace. We will not forego profit growth to chase an opportunity and make large-scale advanced investments without discipline. But in terms of expectations for 2026, I need to make it clear that since the profitability of some of our 2B business depends to some extent on cooperation with various cloud vendors and advertising platforms and adjustments in their policy, which is one of the external variables we have to consider, it is also a factor we will focus on when formulating the 2026 annual plan from the perspective of internal management in the past.

We have developed a relatively mature method for the incubation and investment of new businesses for different products, which is different from what we said earlier. We will not make large-scale investments first and then see the results, but continuously verify our business logic and input-output ratio while investing. And the product music lab is actually promoted in strict accordance with this framework.

As Mr. Pugia said, we will not adopt the method of subsidizing computing power to obtain users. From the business design of the Uzi cloud itself, it is a product paid by users. It comes with a revenue model from the very beginning. From the overall business structure, we rely on a dual-engine model. With the Internet business continuing to serve as a stable source of profits and cash flow for this company through refined operations of the Internet business and continuous improvements through AI enhancement in terms of AI and other new business directions, which we have repeatedly mentioned before, we will focus on the directions that have already shown the potential of business closeness or have undergone preliminary verification through technological optimization, including using AI to enhance our R&D efficiency.

We will continuously improve our overall input output efficiency from the perspective of capital allocation. The company will tend to use the profits and cash flow generated by existing businesses to support the development of new businesses. It will not make large-scale proactive investments to drive expansion. Overall, we will still adhere to our historically effective, prudent, and disciplined approach while continuously exploring long-term opportunities in AI to maintain a healthy operational foundation.

Thank you for your question. Let me answer your question. Indeed, judging from our performance in the fourth quarter, the profitability of the Internet business has indeed improved significantly, but more importantly, we are actually seeing that with the continuous promotion of the user payment model under this background, the Internet business has been continuously optimized in terms of revenue visibility, business sustainability, and cost structure, which is also a core reason for the improvement of the operating profit margin of the entire Internet business. From a business positioning perspective, as I mentioned earlier, the Internet business will continue to serve as the company's stable source of profits and cash flow, providing a solid foundation for the overall operation to support the development of new business.

From a valuation perspective, we have always hoped to guide everyone to use a segmented approach to understand the company's value structure. And in our understanding, for Internet businesses with stable revenue and profitability and clear and mature business, PE, as a profit multiple, is usually used for valuation.

Our own Internet business actually achieved a non-GAAP operating profit of about 110 million yuan in 2025, with a year-on-year increase of 83%. We believe that this part of the value can be further recognized by the market.

Regarding how to further promote the value release of the Internet business and other sectors, We always maintain an open attitude, but we still need to emphasize our core judgment, which has never changed. Our business itself, regardless of a specific business, its continuous development is the premise of value. Only when the fundamentals of the business continue to improve can the value have long-term support. So from this perspective, we also hope that the market will not only see the Internet business as a stabilizer, But also, with the development of new AI agent products, we also hope everyone will pay attention to the company's long-term opportunities in the AI direction, like the new product EC2CL, which is still in its early stages, but actually represents the direction for the company to explore in the next stage of growth.

So to sum up, looking at it as a whole, we will maintain the steady development of the business and continuously improve the transparency of our disclosures in all aspects to guide the market to have a more comprehensive understanding of our company's long-term value. Thank you.

Let me add a few words. Okay. I think everyone is very concerned about our problem. Whether it can be sustained. I think Cheetah Mobile has been developing for so many years, it has encountered some problems in the development process. Now we mainly have two businesses, the internet business and the top business, and then there is the robot business. Our overall idea is like this. As Atali just mentioned, after years of adjustments, we are sure to achieve stable growth and profit in the internet business, We are also in the software business. We are very pragmatic and we really grow together with users. We have also answered the question of commercialization. We will not need subsidies to move our software business forward.

Instead, we will truly build a software business into a basic profit-making business. This is the first point. Second, some policy factors may affect some other income. In fact, this is not our main concern. We are more concerned about the machinery business, although we have been working on it for many years, including launching new products this time. We still hope to truly seize this opportunity. We will also turn the final of the project into a company company and put in the company and the company. Because today's AI pilot is empowering robots faster and faster, we definitely don't want to get up early and miss the market. Therefore, when formulating strategies for the entire year of 2020, we don't necessarily make profitability of our core assessment plan.

On the one hand, the software business is definitely our very important foundation. In terms of the robotics business, we will also, based on market conditions, consider it if the opportunity is good.

As for our investments, including the answer to the question about the way of profitability of robots, I think the scale of the robotics business should be given to Skylarge and continuous growth should also be prioritized. Especially when AI is still hot, I think businesses like robotics have always reached a critical point. So this is our overall idea. After all, Cheetah Mobile still has over $200 million in cash on its books today, yet I remember over $200 million, so we will focus more on long-term business development. In other words, we must seize the opportunity in this API way.

Thank you so much for joining our conference call today. If you have any further questions, please do not hesitate to let us know. Thank you so much.