| Jing Zhu | executive |
| Sheng Fu | executive |
| Thomas Jintao Ren | executive |
Good day, and welcome to the Cheetah Mobile First Quarter 2025 Earnings Conference Call. [Operator Instructions] Please note, this event is being recorded.
I would now like to turn the conference over to Ms. Helen Jing Zhu, Investor Relations of Cheetah Mobile. Please go ahead.
Thank you, operator. Welcome to Cheetah Mobile's fourth (sic) [ first ] quarter 2025 earnings conference call. With us today are our company's Chairman and CEO, Mr. Fu Sheng; and our company's Director and CFO, Mr. Thomas Ren.
Following management's prepared remarks, we will conduct a Q&A session. Please note that the management's prepared remarks 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 as we will make forward-looking statements.
At this time, I would now like to turn the conference call over to our Chairman and CEO, Mr. Fu Sheng. Please go ahead, Fu Sheng.
Good day, everyone. Thank you for joining Cheetah's Q1 2025 Earnings Call. I am Fu Sheng, the CEO of Cheetah. We started 2025 with a clear plan to strengthen our position in both our long-standing and new business areas. Q1 2025 marked a strong start to the year, and I'm happy to share some great news about how we are doing.
First, our revenue grew significantly, and we made solid progress in cutting losses. In Q1, our total revenue went up 36%, compared to last year and 9% compared to last quarter.
Our Internet business did especially well with a 46% increase in revenue year-over-year.
Our AI and Recovery segment grew 23% year-over-year and accelerated to 30% quarter-over-quarter.
Just as important, our loss dropped sharply while still investing in AI and robotics and we believe this positive momentum will continue.
Second, AI Agents are becoming a real game changer that smarter AI models keep improving. They can now go beyond chatting. They can handle real tasks and solve real problems with our strong background in building and launching new products. We believe Cheetah is well positioned to take advantage of this big shift.
We are actively applying agent technology to upgrade our consumer products and power our innovation pipeline. These smart enhancements are making our products more efficient, user-friendly and align with the expectations of the new AI era.
For example, we launched [ M AI ] and AI tool app that turns videos, audio, PDF and other documents into concise summaries and mind maps, making knowledge easier to digest and act on. [ Duba ], a strong example of how we are planning AI agents to create practical data use tools that improve productivity.
Third, AI has always been at the center of our AI strategy.
We are investing even more in R&D and using AI agents to upgrade our consumer products and robotics.
One of our biggest step forward is AgentOS, our next-generation voice system for service robots. AgentOS is designed to be a successful digital purpose, AI brain that can handle everyday tasks and further strengthen our leadership in voice-enabled robots.
Since the transaction, we have been working with our distributors to test AgentOS. Customers, say AgentOS makes the interaction with the robots feel much smarter. They understand the conversation, notices what people are wearing and can use tools like maps. They also like that it doesn't get confused.
If you pause, say something wrong or switch between languages.
We are already working with schools, therapy centers, libraries and museums to bring AgentOS into their daily routines.
Our goal is to create industry-specific apps on top of AgentOS that's smart, helpful and personalized. These apps can greet people, give presentation, help care the elderly and offer companionship. They will use tools and keep running over time, which will help us grow our market share and move closer to general AI that can handle many tasks.
While we plan to offer AgentOS for 3, first, to enhance our robot performance. We see strong potential for future subscription-based business model. In the coming months, we will add 8 agents to our existing apps, including our flagship, Duba Anti-virus, and introduce new AI tools to help users work more efficiently in the LLM era.
At the same time, our legacy Internet business remained strong. It continues to deliver steady revenue and profit and gives us a natural entry point for our new AI experiences.
Overall, the strength of our legacy business gives us the resources we need to push forward with our AI plan, while being financially responsible to wrap up Q1 2025 with a strong quarter. We grew our revenue, reduced our losses and took steps in our AI journey. We believe agentic AI is driving the Chinese LLM industry into a new phase, shifting from infrastructure development to application-driven innovation.
This change benefit companies like us both with a proven track record of turning cutting-edge technologies into real world products across the PC, mobile and now AI eras.
Our ability to productize innovation is what truly sets Cheetah apart in this new phase of AI applications. We remain focused on building AI especially utility-focused AI tools and robotics that not only understand people, but also help them get things done. Thank you.
Thank you, Fu Sheng. Hello, everyone, on the call. Unless otherwise stated, all financial figures are presented in RMB. Q1 2025 marked another quarter of meaningful loss reduction and improved efficiency. Building on the momentum from 2024, our Q1 results reflect our key us focus on disciplined execution, operational efficiency and strategic investments in AI.
Let me walk you through the key numbers. In Q1, total revenue reached RMB 259 million, up 36% year-over-year and 9% quarter-over-quarter. Gross profit increased by 67% year-over-year and 10% quarter-over-quarter to RMB 190 million. Gross margin was 73.2%, up from 59.2% a year ago and 72.9% in the previous quarter.
Non-GAAP gross profit was RMB 190 million, an increase of 67% year-over-year and 10% sequentially. Non-GAAP gross margin improved to 73.2%, up from 59.6% a year ago, and [ 72.7% ] in the prior quarter.
We also made meaningful progress in reducing loss. Operating loss was RMB 27 million, reduced from RMB 81 million in the year ago quarter and RMB 207 million in previous quarter. Non-GAAP operating loss narrowed to RMB 14 million, down from RMB 66 million in the year ago quarter and RMB 42 million in the previous quarter.
Net loss attributable to Cheetah Mobile's shareholders of RMB 33 million, reduced from RMB 80 million in the year ago quarter and RMB 367 million in the previous quarter. Non-GAAP net loss attributable to Cheetah Mobile's shareholders [ increased ] RMB 21 million, a significant improvement from RMB 66 million in Q1 2024 of RMB 202 million in Q4 last year.
By segment, our Internet business continues to provide solid cash flow and profitability. Operating margin nearly doubled year-over-year to 15.5%, driven by improved mitigation and a leaner cost structure. Losses from our AI and other segment narrowed to RMB 46 million, compared to RMB 82 million a year ago and RMB 228 million in the previous quarter. This reflects ongoing efforts to strike the right balance between investment and efficiency.
We remain focused on scalable, monetizable use cases.
We also see real improvements in operational efficiency. AI [indiscernible] coding is now part of our daily workflow, improving efficiency and helping our team to scale faster.
On the robotics side, we have prioritized use cases that can be deployed at scale and address real customer needs.
We also leverage open source models like [ VIE ] models to enhance hardware performance, including robotic arms.
Following the OrionStar acquisition in late 2023, we have continued to consolidate teams and optimize operations.
As of March 31, 2025, our total head count was approximately 815, down from 860 to a year ago. Despite continued cost and expense control.
We also launched new products and made our service robots agenetic.
Looking ahead, we expect further margin expansion and continue loss reduction. At the same time, we will continue to invest in AI, but in a disciplined and focused way.
Our balance sheet remains strong.
As of March 31, 2025, we have cash and cash equivalents of approximate USD 234 million, long-term investments of about USD 112 million.
Looking ahead, our goal is clear, reach breakeven while maintaining a healthy cash position.
We will continue to invest in AI, but in a disciplined and focused way, ensuring every dollar spent support sustainable long-term value creation.
Thank you.
We are now happy to take your questions.
[Operator Instructions]
We noticed that you mentioned 2 directions of AI in this financial report.
On the one hand, tool-based AI products.
On the other hand, service robots. From the perspective of strategic resource investment and revenue contribution in the next 3 years, will Cheetah's future development focus more on building an AI tool matrix and focusing on improving AI efficiency on the C-side, or will more resources be invested in robots? How do you balance the differences between these 2 directions in terms of technical challenges, commercialization rhythm and long-term moat?
Well, I think this is a very good question.
In fact, after all these years, Cheetah Mobile has been focusing on 2 major businesses, AI tool software on the C-side and robots.
Regarding the commercialization efficiency and risks on the B-side, I actually don't think these 2 are contradictory because essentially, for all products, software capabilities are what matter in the end. Take Apple for example. Apple is known for its strong software and its hardware manufacturing is also very good. But ultimately, what users care about is the software experience.
So I think the AI tool matrix and robots today have a short-term and long-term relationship, respectively. That is to say, with the Internet business, we can achieve rapid development. Especially now that programming technology has matured, we believe that the AI tool matrix will develop rapidly. This includes the transformation of some traditional software from the past such as Kingsoft Antivirus and others, which can rejuvenate them.
So I think in the short term this year, the area where we can see rapid development is definitely the AI tool matrix.
However, the robot itself is a hardware entity that carries AI, or you can think of it as a hardware entity that carries AI tool.
So in the long run, I think the robot is, after all, a long-term development direction.
Regarding the technical challenges you mentioned earlier, I think the cutting-edge technologies of these 2 are actually quite similar, which is the final productization of AI technology in enterprises.
Of course, robots are more inclined to the long chain of hardware, while AI tools tend to be more short, flat and fast.
In terms of the commercialization rhythm.
I think the efficiency improvement of AI tools will be faster, which is obvious in the industry. And the development of robots is a long-term task that requires continuous improvement. Surely, the moat of robots is deeper because it involves hardware and business models.
As for this wave of the AI tool matrix, it depends on whether we can really change users' minds in some vertical fields. But overall, to put it simply, this year, the AI tool matrix is the area where we can generate benefits quickly.
We've noticed that the Robotics division is making the construction of a data factory, a key strategic investment aiming to accumulate a vast amount of high-quality data from the physical world for model training.
However, Cheetah has already amassed a large amount of scenario data during actual deployment. Could you share the company's thoughts on data asset construction and self evolution? Do you consider providing data externally or forming a B2B service in the future?
This is a very comprehensive question. In the robotics industry, especially when it comes to service robots or the currently popular humanoid robot, there are numerous challenges. A crucial point is that it's difficult for us to convert the data related to human labor into robotic data. This is quite different from autonomous driving where the data from human driving is already machine accessible data. Indeed, we've seen many in the industry, including some start-ups working on the construction of data factories.
However, as of today, in the robotics industry, the conversion of data factories into actual productization and commercialization is still in a very early stage. In the foreseeable 3 years, I won't say 5 years because AI is evolving so rapidly, it won't be possible to turn it into a truly commercial product.
So regarding the data we've accumulated up to now, we can't claim that it has significantly contributed to the company's productization, but we are conducting some exploratory research at the forefront. On one hand, I'm very optimistic about the long-term prospects of the robotics industry.
On the other hand, I'm extremely cautious at the moment.
We've been investing in this industry for 7 or 8 years and have poured over RMB 1 billion into R&D. We started large-scale R&D in this area very early. From a technical paper to certain technical direction, and finally, to actual scene-based applications, there's still a long way to go.
Moreover, there will be various changes in the industry landscape, including the impact of open source technologies. In short, for Cheetah, the construction of a robotics data factory is not our priority at present. We'll keep an eye on it, but won't invest blindly.
As for whether we'll provide data externally or offer B2B services in the future, we have no such plans for now. Because in my opinion, it's practical application is still a long way off. I've been in Silicon Valley recently and talked to many people there, including those from relevant startups. There's basically a consensus that this matter is still in a very early stage.
Currently, everyone is still exploring how to build this data factory, whether it's through human remote control or using some videos for data collection. It's not like the situation with ChatGPT, where a clear path has emerged, and we just need to follow it to turn it into a product. I don't think we've reached that stage yet.
So this question is too premature for us and we haven't considered providing external data services.
Just now, the management mentioned that the company is leveraging open-source BLA models to drive the intelligent evolution of robots.
Given the increasingly mature open source ecosystem, how does the company balance the use of open source models and the self-developed approach in actual deployments, especially in terms of inference efficiency, security and controllability and cost structure? How does the company allocate technologies and resources?
In addition, from a medium- to long-term perspective, does the company believe that Cheetah's moat in the robot business should be built on model capabilities or scenario data assets? Is it possible to consider building a long tail advantage through a data loop?
These are really professional questions.
Regarding your first question on how to balance the use of open source model and the self-developed approach, I think most companies are already quite clear about this.
For the vast majority of companies, they don't make a strict distinction. If open source models were better, of course, they'll use open source ones because for private deployment, open source models are no different from self-developed ones and they can save a lot of resources and cost. There's no need to reinvent the wheel. Even Tencent [indiscernible] and DeepSeek and Baidu also uses relevant open source resources.
So there aren't many companies that are so insistent on self-development. Maybe companies like Google, OpenAI, et cetera, might be, but for a company like ours, we definitely use open source models.
As long as there are suitable open source models, we won't self-develop. There's no need to repeat the work. I've repeatedly emphasized the power of the open source community in my short video programs over the past 2 or 3 years. In the AI industry, open source is extremely powerful. It's very difficult for a single company to compete with the combined efforts of so many peaks worldwide. We've been clear about this for a long time and have been acting accordingly.
For example, in our AI-based operating [indiscernible] Regarding the 3 aspects of inference efficiency, security and controllability and cost structure, many people today only consider the model [ sequencing ] when talking about various VLA models or other models, but rarely mention efficiency or application scenarios.
For example, if you ask it a question and it takes a long time to respond, you can tolerate it when you're sitting in front of a computer, especially when it's writing an article. But if we are using it for robot interaction and it takes 5 seconds to reply after you say something, the person might have already left.
So among these 3 dimensions of inference efficiency, security and controllability and cost structure, we definitely prioritize inference efficiencies. When used in commercial scenarios, the response speed is of utmost importance. There's a lot of fine-tuning to be done here.
Of course, security and controllability are also crucial and are a basic requirement, but I believe inference efficiency comes first.
Cost structure is relatively easier to deal with as it's constantly decreasing.
If you look at any AI facing the market company, if it's a real market-oriented company, it will definitely attach great importance to efficiency combined with the application scenarios. In our case, efficiency is definitely the top priority.
So in terms of technology and resource allocation, we invest a lot in improving the reaction speed of our robots.
If you actually run some models on a robot, the latency might be unacceptable to ordinary users.
Regarding the medium- to long-term perspective on whether Cheetah's moat in the robot business should be built on model capabilities or scenario data assets, I think there's no doubt that it should be built on scenario data assets. When it comes to so-called model capabilities, I have some reservations. What exactly does model training ability needs? As I said before, in the context of China and the U.S., model training ability is not a problem for many companies. There are a lot of open source papers and models available. Buying a GPU and setting up the training environment is not an extremely difficult task, but very few companies in the world can actually modify the underlying logical structure of a model.
So the current industry trend is becoming increasingly clear.
Most companies' advantages lie in scenario data assets rather than model capabilities. This is also why Tencent strongly promotes Yuanbao, which has DeepSeek integrated because only scenario data can feedback into the model. When the time is right, developing a model that suits your own scenario data is the way for most companies to survive.
Take the autonomous driving field as an example. Initially, Google's model capabilities were very strong. But today, Tesla's autopilot experience is definitely better than Google's self-driving system. That's because Tesla built its advantage on a large number of deployed terminals and real-world scenarios.
So our strategy is very clear. We won't invest blindly in self-developed model. Instead, we'll focus on implementation, pay more attention to interacting with our users and thus build our own scenario data. Then this data can, in turn, enhance the capabilities of even an open source model.
In terms of computing power, algorithms and data, without a doubt, we believe data is the core determinant. This is what we truly value.
What considerations does the company have regarding the commercialization path of AI tool applications. Will it consider the user subscription system? Or will it launch enterprise SaaS product or explore directions such as QC licensing? Against the backdrop of the current shift of AI applications from proof-of-concept to actual commercial value, how does Cheetah plan its commercialization path?
I think a very obvious characteristic of AI tools today is that there is an inevitable question whether users are willing to pay for these AI tools because essentially, this wave of AI tools are productivity tools.
So basically, the business models that have emerged globally for this wave are all about subscriptions, whether it's the model of OpenAI or the model for PPT-related software or the model for programming software like GitHub Copilot, they are all subscription-based.
And the subscription model is constantly evolving into different tiers.
If you use more, you pay at a higher level. Essentially, when it helps users improve their efficiency, users are willing to pay.
I think this is also where AI is different from the previous wave of the Internet. This time, the business model is very simple, clear, and has a high user acceptance.
For example, we developed a small product, [ BBL. ] Now users are actively asking how to pay for it, and some have already paid.
So for us, the user subscription model is not a consideration. It's a clear-cut choice. Maybe because Cheetah Mobile faced some setbacks in the globalization of tools in the past, we've converted many of our tools to the subscription model in the past few years. Even Kingsoft Antivirus is like this.
Today, user payment is the mainstream, not advertising. In the past few years, in the Chinese software market, although many people may not be fully aware for our own experience, subscription-based payment has become the mainstream for Chinese tool software. Paying for the effect makes us focus more on polishing the user experience rather than on negotiating advertising deals.
I think this is a very important reason why our Internet business has been growing continuously in recent quarters. We've made user-centered payment our core business model.
As to whether to launch enterprise-based products, we've actually been trying.
Regarding enterprise-oriented products, in response to your question about how we plan our commercial path during the transformation of AI applications from concept to actual commercial value, I think our commercial path is becoming relatively clear.
In the Internet business, we have products like Kingsoft Antivirus, some of our businesses in Japan and other software businesses. What we're doing is using agent technology to transform software like Kingsoft Antivirus and our overseas software from simply delivering functions to delivering results. This is very important for us.
The second thing is that because we've been developing tools for so many years, and I believe the concept of agents is similar to that of tools. In the past, due to immature technology, we could only provide users with a list of functions. But now with agents, we can provide users with results. This allows us to utilize our entire R&D efficiency.
So the second part is that we'll have a lot of new launches in the innovative content within the Internet sector. And because of the efficiency improvement brought by AI tools, our R&D costs for these innovative applications will be much lower than in the past.
So one aspect is to use AI agents to transform traditional Internet businesses.
The second aspect is that during this era of the big explosion of agents, we'll also make many innovative attempts. These attempts won't be made with large-scale investments, but rather in a start-up like small team-based way.
In the robotics area, our main focus is on our agent-based OS. This is what differentiates us from our competitors. We won't focus on humanoid robots or robots with overly complex mechanical structures, such as lawn mowing robots. What we're good at is robot interaction, companionship and task planning-based logic with software as the core and hardware as the foundation.
I think in 1 or 2 more quarters, this path will become even clearer.
What progress has the company's robots made this quarter? Could you please share some specific cases of actual implementation. From an industry perspective, what significant changes have taken place in the robot industry this quarter? And how has Cheetah Mobile perceived and responded to these changes?
Let me talk about the industry first. I've been not only in China, but also recently traveled a lot in the U.S., meeting many entrepreneurs.
Here are my views on the industry. We've always believed that humanoid robots are still a long way from commercialization.
By commercialization, I mean the kind that can form repeat purchases and become a productive force, not the commercialization in the form of exhibitions, rentals or for educational purposes.
Although these forms exist and are currently at a certain scale, the idea of humanoid robots being used on production lines, I think, is still a long way off. In my opinion, it will take more than 5 years to achieve real commercial implementation. That's my view at the industry level.
Besides the hype around humanoid robots, I've noticed that there's a rise of robots for various specialized scenarios, including those from startups. These robots are designed for very specific tasks and don't necessarily look human like. This is a clear change in the industry.
Now let me talk about our own progress.
I think our progress can be summarized in the following aspects.
First, we've clearly sorted out our development ideas for robots.
As I mentioned just now in the field of robots, what we're best at is not complex hardware mechanical structures.
Companies like Yaskawa are indeed very strong in that aspect, and I admit it. What we focus on is the integrated interaction experience of perception and action. That is we ensure that the wheel movement of the robot from point A to point B is stable and reliable. This has been verified by our customers in Japan, South Korea and Europe over the past 2 years.
Also, we aim to give full play to the real-time interaction ability such as in scenarios like reception, explanation, promotion and delivery.
With the support of large language models, I believe this kind of scenario can fully thrive.
Regarding specific cases of actual implementation. In many corporate exhibition halls and urban service centers in Beijing, for example, our robots have started to be used as tour guides. I even shot a short video yesterday.
With the support of large language models, the robot interactionability, their understanding of what you say, and their task planning ability have been significantly improved compared to before.
I think the era when the robot can become an excellent tour guide has arrived.
Moreover, with large language models, it naturally becomes a translator. In the past, we had a lot of headache dealing with multilingual capabilities like English, Chinese and Japanese.
Before, for [indiscernible] robot, we didn't even dare to turn on the voice function because we had to go through a series of complex operations by connecting to Google's services to enable a foreign language ability, and the effect was not good. But now, we started to launch multilingual interactive robots overseas.
If you really ask me what our advantages are in making robots.
I think our biggest advantage doesn't come from our so-called technological documentation or the years of technological accumulation. Instead, it's that we now have hundreds of domestic and overseas agents. Through these commercial channels, we can quickly obtain users' opinions on things, and we know how to quickly find our customers to actually test run our products and really bring them to the market.
Due to these advantages, we can perceive that, for example, starting this year, with the improvement of large language model capabilities, the market for explanation related services is clearly changing. This year, we're focusing on language version first and the multilingual capabilities for overseas markets are some progress and may be a bit slower.
But in Q2 this year, the repurchase rate and the degree of completion of our indicators for domestic interaction-based scenarios are the best in recent years. I can clearly feel that the voice interaction-based market is emerging. What's our next step? Yes, we may launch some special robot products related to companionship, targeting scenarios like elderly care.
You can wait and see.
Regarding [indiscernible] share further customer feedback, this includes user stickiness, customer satisfaction and whether there have been customized deployment or active inquiries.
Additionally, how does the company internally evaluate the commercialization rhythm of AgentOS?
These are very detailed and crucial points. No matter how grand the concept is, ultimately, it comes down to whether users are willing to pay for it.
So far, we've conducted some user satisfaction surveys. Generally, users have reported that when it comes to real-life conversations, especially in noisy and crowded environment, the responsiveness has significantly improved compared to the previous generation. I don't have the specific satisfaction data at hand. Maybe we'll release some articles about it in the future.
Regarding customized deployments, it's like what Henry Ford said.
If you ask customers what they want, they won't ask for a car, but a faster horse. In the past, due to the limitations of previous ASR, automatic speech recognition technology, which involves converting speech-to-text, and then processing it with NLP, natural language processing, it couldn't meet users' requirements.
As a result, users thought these products were useless. We all know that when people bought smart speakers in the past, they could only use them to play songs, and the speakers would become unresponsive with a bit more complex instructions.
However, after the emergence of GPT, people realized that it could understand such complex text, which triggered the development of various applications.
So with our AgentOS, through multiple sensors such as vision sensors, microphones and even some radars, its ability to understand user intentions has improved significantly. I believe this is what can truly open up the market for users. We've already received some requests for customized deployments and inquiries, but we won't disclose the specific details for now. We started with domestic operations.
First, we're training our agents and providing authorization and training for the secondary development platform, so that they can develop their own applications on it.
As for evaluating the commercialization rhythm of AgentOS, we mainly focus on the sales progress of our voice interaction-based robot. It's currently Q2, and we're looking at whether we can achieve our goals in Q3. This is a very critical point in evaluating the commercialization rhythm.
Overall, at this stage, the key is whether we can integrate user needs with our products more efficiently, enabling our products like our robot tour guides to be ready for service at any time, and our robot salespersons to perform well. If we can achieve this, I think it will mark the beginning of rapid commercial development. I'm quite confident about this. I believe the basic framework has been established.
Cheetah currently holds over $200 million in cash. I'm wondering if the company is considering making acquisitions to further address the shortcomings in the AI application chain.
Thank you, Mr. [indiscernible] Chang Lu, for your question. We appreciate your attention to our cash reserve scale and the focus on our strategic investment directions.
Indeed, having over $200 million in cash provides us with considerable strategic flexibility. In recent years, Cheetah's investment department has been closely monitoring and actively evaluating areas related to artificial intelligence, including AI large model, vertical AI applications and the upstream and downstream of the robot technology industry.
We believe that external cooperation or integration is an important way to accelerate the construction of our capabilities and popularize key chains, and it is also crucial for promoting Cheetah's long-term competitiveness in the AI field.
Regarding the acquisition strategy specifically, our core considerations mainly lie in 2 aspects. One is the alignment with Cheetah's strategy, and the other is the potential to enhance the overall value creation for Cheetah's shareholders.
For potential target companies, we generally conduct a systematic evaluation from the following aspects.
Firstly, the synergy between their technology and business and our company.
Secondly, the strategic value they can bring to us.
Thirdly, the compatibility of their team with Cheetah's culture and values. And finally, the fourth aspect is the financial valuation and its rationality.
If a potential target fully meets our standards and both parties can highly agree on strategic operations, we will consider acquisition as a major strategic option to accelerate the construction of our capabilities in key links of the AI or robot industry chain.
Of course, during the evaluation process, they're quite flexible and maintain an open attitude. Depending on different targets, development stages and cooperation depth requirements, we may also participate in the construction of the entire ecosystem through forms such as minority equity investments, strategic partnerships or joint ventures.
In summary, the core principle of how we use our cash reserve is to maximize the long-term value for our shareholders. In strategic key areas such as AI and robots, we will continue to actively seek and rigorously evaluate opportunities that can bring competitive advantages and value enhancement, including strategic acquisitions that meet our standards.
Will the company achieve overall breakeven in the second half of 2025? I would like to know further that on the revenue side, will future profitability rely more on the restorative growth of the Internet business or the new driving force of the AI business?
At the same time, we've noticed that the Internet business revenue has grown well in the past few quarters. What are the main driving factors behind this? Do these factors have sustainability? For the next few quarters, can the management give some directional judgments regarding the revenue growth rate and profit margin level of the Internet segment? In addition, since the AI business is currently in the investment stage, does the company have an internal plan for achieving breakeven at certain stages?
Okay.
Let [ Thomas ] answer this question.
Your questions focus on several aspects, including our breakeven situation, growth drivers and business outlook. I'll answer them separately.
Regarding the company's outlook for overall breakeven. In the second half of the year, achieving profitability in the second half is a major internal goal for us, but we do face some challenges.
Whether we can reach this goal largely depends on the progress of our core businesses, especially the speed of business development as well as the overall market environment.
Of course, our management and team will go all out.
We will also update our expectations to the market in a timely manner according to the progress.
Regarding the future drivers of profitability and the analysis of the Internet business, I think the main drivers for the company's future profitability will surely come from the driving force of our AI and other businesses.
The Internet business is an important foundation for us, and it is expected to maintain stable growth.
Some of the driving factors for the growth of the Internet business in recent years, as the Vice President mentioned earlier, is that we have completely transformed from the traditional advertising model to the user payment model over the past few years.
By returning to the value of the product after years of refinement, by adhering to the user-first concept, we have enhanced our product strength, which has brought stable user growth as well as stable long-term partners and customer acquisition channels. I believe these driving factors are sustainable.
The subsequent growth of the Internet segment mainly depends on whether we can expand more partners based on the existing channels and partnerships. Also, as the Vice President mentioned, we will use [ AI ] technology to upgrade our traditional PC and mobile end tool products to enhance the competitiveness of our products.
Regarding the growth rate and profit margin level of the Internet segment, generally, we don't usually make specific forecasts in the short-term. But in the short-term, the growth rate and profit margin improvement mainly depends on what I mentioned earlier, that is the expansion of new partners or channels in the next few quarters as well as the development and implementation effect of new AI-related features that empower our traditional tools.
We will actively promote these aspects.
Finally, regarding the investment and planning of AI, as mentioned before, our business focus is on how to refine our products for scenarios that are more in demand by users and have more commercial prospects. The AI business, especially the robot business is the core growth engine for our future. Currently, it is still in a crucial strategic investment stage.
We have set clear phase goals internally.
Our key task is to concentrate our R&D resources and strive to roll out products suitable for various user usage scenarios aiming to achieve the goals as soon as possible.
The losses of the AI and other business segments significantly narrowed in Q1. What were the main areas where investment was scaled back? Does this mean that early exploratory projects have been phased out and their transition towards an ROI-oriented approach has begun? Against the backdrop of the current shift of AI investment from proof-of-concept to actual commercial value, how has Cheetah adjusted its investment strategy?
Let me explain a bit. The significant narrowing of losses in the AI and other business segments isn't just due to one factor. On one hand, some of our explorations have indeed reached a certain stage.
For example, we've realized that large-scale model training doesn't hold much significance for a company of our size.
So we saved a significant amount of computing costs by no longer starting from pretraining.
Although we're still doing things like fine-tuning after pretraining, we've stopped the pretraining process for 2 models, one with 141 parameters and a medium large mode model. Once our team had grasped the entire technical chain and its capabilities, we ceased this training. We believe that in the future, there won't be many model providers.
Only a very few companies will succeed with models. Maybe OpenAI is one of them. But in the future, most companies will be application-based rather than model focus. The key is to do a good job in applications. Whoever can excel in applications has the potential to become a giant. And perhaps in the future, after having successful applications, one can then consider modifying the model.
For now, the focus is on applications.
On the other hand, our R&D has become more efficient. This is an important reason for the significant reduction in losses.
As for what you mentioned about some exploratory projects being phased out, for instance, we had some projects related to large-scale systems for the geospatial domain. But later, we found they weren't suitable for us, so we made quick and decisive adjustments.
In fact, almost the entire company is now transitioning towards an ROI-oriented approach, whether it's the Internet business, the advertising business or some new products we're developing for robots.
As I mentioned in previous answers regarding robots, simply emphasizing technological advancement doesn't carry much meaning. The ones who can find a scenario and scale up first has the opportunity to win in this round of competition rather than relying solely on a few advanced technical points. Technical points don't confer a first-mover advantage. The real advantage lies in the scenarios, the users attracted and the resulting growth. Technical points themselves don't have an ecological edge.
So we have indeed shifted towards an ROI-oriented model.
However, it doesn't mean that all exploratory projects will be eliminated. We still need many of them.
We believe that this wave of AI agents represent a major upheaval in rewriting applications. No company can clearly predict how it will unfold. At this time, we need to not only transform our existing businesses like the Internet and robot businesses with AI technology, but also carry out exploratory projects. Nevertheless, the ROI orientation for these exploratory projects is very clear. We'll strongly emphasize it.
As I also mentioned, since the user subscription payment model is a very clear-cut business model, it's easy to understand. In short, this is the overall operating concept of Cheetah. That is use AI to transform old products and improve the energy efficiency ratio.
For robots, also use AI to enhance capabilities such as interaction and task planning rather than competing in terms of mechanical structure complexity. We'll enhance the productization of the overall AI experience across all lines.
For AI agents, we'll continue to adopt a startup-like approach with small teams making rapid attempts, but with ROI as the assessment criteria.
So in the context of the transformation from concept validation to commercial value, our investment strategy, both for internal and external investments, highly valued practical implementation. Whether it's an internally incubated project or an externally eyed project, the key point for us is whether it can generate real-world commercial returns from the market. Thank you for your questions.
In the current context, with the capabilities of large models at home and abroad are gradually converging, what are Cheetah's competitive advantages in the AI application layer? And how does it guard against the risks of being replicated or marginalized by platform-based products in the future?
Yes. Actually, this is something I've been thinking about a lot recently and have also witnessed in practice.
Given the driving force of large model capabilities, an important point is that if you can create a product, you won't be easily marginalized by platform.
First, many of today's so-called platforms are built on past experience points. At this time, the products made with agents bring a brand-new experience to users.
Just to answer your question, I won't go into details about what an agent is and the specific differences it brings to enterprises. But it's very clear.
Let me give you some examples. Take the search field. Google has been in this business for so many years. Today, products like ChatGPT, you can also consider OpenAI's ChatGPT as a kind of trend, have led to a gradual decline in some of Google's vertical search traffic. It's the same in China.
I even made a special video called, The Death of Search Engines. Besides search in the programming field, products like GitHub Copilot are emerging. When big programming tools like Visual Studio were very powerful, they were all in one product. But GitHub Copilot has developed extremely rapidly and offers a completely different user experience. Today, I know many companies are using such products across the board.
What I'm trying to say is that if we start from the user's perspective instead of the competitive perspective, we'll find that the experience by building products with agents is difficult to achieve with traditional software technology. At this time, you can create a new perception for users, which has nothing to do with the past.
So today, it's actually the platform that should be worried because various agents-based products might, in some aspects, really have the potential to disrupt the platform. This is the first point.
Second, whether it's due to the convergence of large model capabilities or the rise of open source model capabilities isn't a decisive trump card in large model operations currently.
For example, when building your product experience, you can choose to buy services, use models like GPT through API.
Your enterprise can stand on the comprehensive capabilities of all models.
So the user experience can be very good, and it won't be affected by the fact that, for example, the current financial situation leads to all major platforms opening up various APs.
Take the example of our product names in the specific area.
You can try it. Its ability to summarize meetings is, I can say with full responsibility, far better than [indiscernible]. This is the result of a [indiscernible] by a small team of ours. The user experience is so good that users are willing to pay for it. It's much better than some companies that have raised a lot of money.
If you give them APDS and ask them to summarize it, using our product name will definitely give you a better result.
Moreover, the capabilities of large models are constantly improving, which is a great advantage. It's not like traditional coding with fixed logic. Once someone comes up with better code, you'll be surpassed. But for agents, a true agent relies less on a specific model and more on the capabilities of large models themselves to make judgments.
As long as the large model capabilities improve, the capabilities of your agent-based products will also improve, and users can feel this improvement. They won't be easily replaced by a new model because the product can simply choose a better model.
Just as I mentioned, this is why agents are so promising.
First, they allow users to form a new perception.
Second, since the underlying capabilities are provided by large models, it's like using electricity provided by a power plant. The more stable the power supply, the better your electrical appliances will perform. When we were kids, the power supply was unstable, and we felt that electrical appliances were not easy to use. But when the power supply is stable, electrical appliances can be sold better.
So the second point is that the new product model based on agents actually benefits from the advancement of large model capabilities.
For a company like Tencent, although they've developed their own large model, recently, I saw they're still focusing on integrating data with the platform. But as platforms manage to do all these things, small companies might really be in trouble.
However, I want to say that during this era of rapid technological change, platforms usually move relatively slowly. At this time, if you can quickly seize the opportunity brought by this technological change and acquire enough users, you can gradually form a growth flywheel that platforms can't replace, especially platforms from the previous era, and it won't be easily marginalized.
I can give another example.
Although it has nothing to do with the agent platform for now, but I think it will soon, a product like Kingsoft Antivirus has a history of 20 years. By seizing one opportunity after another in the Internet era from the initial feature to the subsequent intermediate feature and now to current feature, it has continuously grown. Many of its contemporaries like competitor names have long since disappeared, but it still thrived today.
I think in the era of agents, we might be able to seize this opportunity to provide better services in terms of users' computer using convenience, and this is not something that a platform can simply replace.
Ladies and gentlemen, the conference has now concluded. Thank you for attending today's presentation.
You may now disconnect your lines.