Kingsoft Cloud Holdings - Earnings Call - Q4 2024

Transcript

Operator (participant)

Good day, and thank you for standing by. Welcome to Kingsoft Cloud's fourth quarter and full year 2024 earnings conference call. At this time, all participants are in a listen-only mode. After the speaker's presentation, there will be a question-and-answer session. To ask a question during the session, you will need to press star one and one on your telephone. You will then hear an automated message advising your hand is raised. To withdraw your question, please press star one and one again. Please be advised that today's conference is being recorded. I would now like to hand the conference over to your speaker today, Nicole Shan, IR Director of Kingsoft Cloud. Please go ahead.

Nicole Shan (IR Director)

Thank you, Operator. Hello everyone, and thank you for joining us today. Kingsoft Cloud's first quarter and fiscal year 2024 earnings release was distributed earlier today and is available on our IR website at ir.ksyun.com, as well as on GlobeNewswire services. On the call today from Kingsoft Cloud, we have our Vice Chairman and CEO, Mr. Zou, and the CFO, Mr. Henry He. Mr. Zou will review our business strategies, operations, and company highlights, followed by Mr. He, who will discuss the financials and the guidance. You will be available to answer your questions during the Q&A session that follows. There will be consecutive interpretations. Our interpretations are for your convenience and reference purposes only. In case of any discrepancy, a management statement in the original language will prevail.

Before we begin, I'd like to remind you that this conference call contains forward-looking statements within the meaning of Section 21E of the Securities Exchange Act of 1934, as amended and as defined in the U.S. Private Securities Litigation Reform Act of 1995. These forward-looking statements are based upon management's current expectations and current market and operating conditions, and relate to events that involve known and unknown risks, uncertainties, and other factors, which are difficult to predict, and many of which are beyond the company's control, which may cause the company's actual results, performance, or achievements to differ majorly from those in the forward-looking statements. Further information regarding these and other risks, uncertainties, or factors are included in the company's filings with the U.S. SEC.

The company does not undertake any obligation to update any forward-looking statement as a result of new information, future events, or otherwise, except as required under applicable law. Finally, please note that unless otherwise stated, all financial figures mentioned during this conference call are denominated in RMB. It's now my pleasure to introduce our Vice Chairman and CEO, Mr. Zou. Please go ahead.

大家好，欢迎参加金山云2024年第四季度及全年业绩电话会。我是金山云的CEO周涛。本季度，我们的高质量可持续发展战略继续结束硕果，调整后经营利润首次扭亏为盈，一系列财务指标创下历史新高。首先，我们的收入增速再上阶梯。在Q3同比增长16%的基础上，本季度实现收入人民币22.3亿元，同比增速翻番，达到了30%，领跑行业。其中，公有云同比增长34%，行业云同比增长23%，实现双轮驱动均衡发展。其次，我们的盈利水平加速提升。本季度，我们实现了增收又增利，公司一举扭转了历史上长期经营亏损的局面，向着净利润盈利目标迈进。调整后经营利润首次实现盈利，调整后经营利润率为1.1%，相较去年同期的亏损10.9%提升了12个百分点。调整后毛利额达人民币4.3亿元，创历史新高，同比增长63%。调整后毛利率达19.2%，调整后EBITDA率也创新高，实现了16%的高双位数，同比提升了17.7个百分点，环比提升了6.3%。这是我们提升收入质量、调整收入结构、落实降本增效、多重举措并行所取得的喜人成果。第三，我们继续以AI引领转型。本季度，AI业务账单收入达4.7亿元人民币，同比增长近500%，环比增长31%，占公有云收入比例继续提升至34%。第四，作为小米金山生态的唯一战略云平台，我们继续全面拥抱小米金山生态在AI时代的业务机会，得益于小米汽车、手机、IoT、智能制造和WTS AI等历史性机遇。本季度，小米和金山生态收入达4.93亿元，同比增长76%，占收入比例进一步提升至22%。

Tao Zou (CEO)

Hello everyone, thank you and welcome all for joining Kingsoft Cloud's fourth quarter and fiscal year 2024 earnings call. I am Zou Tao, CEO of Kingsoft Cloud. This quarter, our high-quality and sustainable development strategy continued to bear fruitful results. Our non-GAAP operating profits achieved profitability for the first time, and a series of our key financial performance measures reached all-time highs. First, our revenue growth rate has climbed to a new level. Building on the 16% year-over-year growth in Q3, we achieved total revenue of RMB 2.23 billion this quarter, doubling year-over-year growth rate to 30%, leading the industry's growth rate. Among this, public cloud revenue grew by 34% year-over-year, and enterprise cloud revenue grew by 23% year-over-year, achieving a balanced development with dual growth engines. Second, our profitability improvement accelerated.

This quarter, we achieved both revenue growth and profitability improvement, marking a turnaround from the company's long-standing history of operating losses and marching towards the goal of net profit breakeven. Non-GAAP operating profits turned positive for the first time, with non-GAAP operating margin reaching 1.1%, representing a 12 percentage point improvement compared to a loss of 10.9%. Non-GAAP gross profits reached a record high of RMB 427.7 million, up 63% year-over-year. Non-GAAP gross margin was 19.2%. Our non-GAAP EBITDA margin also reached a new high at 16%, representing a year-over-year increase of 17.7 percentage points and a sequential increase of 6.3 percentage points. These are gratifying results of our comprehensive efforts to improve revenue quality, optimize business structure, and implement multiple cost reduction and efficiency enhancement measures simultaneously. Third, we continue to lead our transformation with AI.

This quarter, our AI-related business achieved a gross billing of RMB 474 million, representing nearly 500% year-over-year growth and 31% quarter-over-quarter growth. Its contribution to public cloud revenue further increased to 34%. Fourth, as the sole strategic cloud platform of the Xiaomi Kingsoft ecosystem, we continue to fully embrace the opportunities brought by the AI era within the ecosystem. Thanks to the unique historical opportunities driven by Xiaomi EV, mobile phones, IoT devices, smart manufacturing, and WPS AI initiatives, this quarter, our revenues from Xiaomi Kingsoft ecosystem amounted to RMB 493 million, up 78% year-over-year, with its share of total revenues further increasing to 22%.

下面我向大家具体介绍2024年第四季度的业绩情况。公有云方面，本季度实现收入14.1亿元人民币，同比增长34%，得益于小米AI全面赋能人车加全生态、金山WTS AI等生态业务推进。作为小米金山生态的唯一战略云平台，AI时代所带来的云服务业务增长潜力空前庞大。2024年12月，股东大会批准了我们大幅提升来自小米金山收入的金额，2025年到27年三年合计113亿元，是23年的10倍。我们正在扎实地落实业务合作。本季度来自小米和金山的收入同比增长76%，预计2025年继续保持增长势头。同时，AI业务继续发力，本季度AI账单收入大幅增长至4.7亿元，环比增长31%，占公有云收入比例达34%，处于行业领先地位。生态内外部客户均有显著用量增长，在大语言模型、互联网AI应用产品等领域均有客户突破。在智算云时代，公司凭借深入的行业洞察、快速的研发投入和丰富的算力积累，在客户服务、技术能力、质量保障等方面收获了客户和行业专家的高度赞誉。更难能可贵的是，我们的市场竞争力也获得了头部友商的认可。智算云浪潮下，公司的市场份额、行业声誉均有显著提升，以跻身行业领先地位。

Now let me walk you through the key business highlights for the fourth quarter of 2024. In terms of public cloud services, revenues reached RMB 1.4 billion this quarter, representing a year-over-year increase of 34%, capitalizing on the opportunities presented by Xiaomi's human, car, home, smart ecosystem, and the WPS AI penetration. As the sole strategic cloud platform within the Xiaomi Kingsoft ecosystem, we are seeing strong growth potential in the AI era. In December 2024, the shareholders approved a significantly increased revenue contribution from Xiaomi and Kingsoft from 2025 to 2027 to a total of RMB 11.3 billion, representing 10 times the amount in 2023. This quarter, revenue from Xiaomi and Kingsoft increased by 76% year-over-year, and we are steadily advancing and executing our business cooperation as we expect the growth momentum to continue in 2025. Meanwhile, our AI-related business continues to gain momentum as well.

This quarter, the gross billing of AI-related business reached RMB 474 million, increasing by 31% sequentially and contributing 34% of public cloud services revenue, leading the industry. Both Xiaomi Kingsoft ecosystem and external AI customers increased their usage, and we made breakthroughs in customer acquisition, including large language model customers, and expansion in use cases, including AI applications for internet company customers. In the era of AI cloud computing, with in-depth industry insights, advanced R&D investment, and substantial computing power, we have been highly praised in terms of customer service, technology capabilities, and quality assurance by our customers and industry experts. Moreover, our competitive strengths have gained recognition by top tier in the industry. In this wave of intelligent computing cloud, the company's market share and industry reputation have significantly improved, securing a top position in the industry.

行业云方面，本季度实现收入8.2亿元，同比增长22.7%。公共服务领域，我们以模型能力、大数据、协作办公为底座。

Operator (participant)

Please stand by while your speakers reconnect. Please continue to stand by while your speakers reconnect. Speakers, you may proceed.

行业云方面，本季度收入8.2亿元人民币，同比增长22.7%。公共服务领域，我们以模型能力、大数据、协作办公为底座，推行标准化运营解决方案。本季度，我们完成了武汉光谷云平台投资建设。金山云将作为联合运营团队，全面负责该区域运维运营和信息化任务，进一步加强数据资源管理和开发应用，最终实现数字光谷的统一规划建设和管理。数字健康领域，2024年12月，宜昌市中兴人民医院系统建设完成阶段性验收，KDX数据中台投产上线。金山云采用云原生微服务架构，以统一数据平台解决了宜昌市中兴人民医院多院区业务协同和实施数据融合问题，能够支撑600家数据作业。我们还协同金山办公，联合研发出基于WTS的电子病历产品，以上线61个业务科室。

Tao Zou (CEO)

In terms of enterprise cloud services, revenue amounted to RMB 822 million, increasing by 22.7% year-over-year. In public services space, we promote solutions with standardized operations, leveraging a foundation of model capabilities, big data, and collaborative office tools. This quarter, we completed the construction of the Wuhan Optics Valley Cloud Platform, where, as a joint operating partner, we will be fully responsible for the regional informatization, operation, and maintenance, strengthening the management and application development of data resources so as to achieve the unified planning, construction, and management of Digital Optics Valley. In digital healthcare space, in December 2024, we completed a significant milestone of the system construction of Yichang Central Hospital, and the KDX, which stands for Kingsoft Cloud Data X Data Middle Platform, has been put into operation.

We adopted a cloud-native microservice architecture to solve the multi-campus business collaboration and real-time data integration difficulties of Yichang Central People's Hospital, with a unified data platform supporting more than 600 data operations. In addition, in collaboration with Kingsoft Office, we jointly developed an electronic medical record product based on WPS AI, which has supported 61 medical departments' daily operations of the hospital.

产品技术方面，我们继续秉持技术立业，聚焦于打造核心产品的一流客户体验。本季度，我们持续提升智算云的产品能力，推出第九代云服务器，能够高效支撑AI与机器学习、算法推理等高计算密集型场景，有效满足大规模数据库实施数据分析，并且增强了数据安全与加密处理。我们正式发布了自研的新流训训推平台，能够支持客户开发、训练和自定义镜像等需求。我们的智算网络实现了大模型训练、故障排查，从小时级降低到分钟级的重大突破，进一步保障服务稳定性。此外，我们的公有云、行业云场景已经实现了DeepSeek的适配使用。未来，我们将继续夯实我们的智算云产品技术，为客户提供稳定、快速的计算服务。

In terms of product and technology, we uphold the principle of building success based on technology and innovation, focusing on delivering best-in-class customer experience across our core product offerings. This quarter, we continued to enhance the product capabilities of our intelligent cloud computing services. We launched the ninth-generation cloud server, which efficiently supports high-performance computing use cases such as AI, machine learning, and inference. It also effectively meets the demands of large-scale databases, real-time data analysis, and has strengthened data security and encryption processing. We officially released our proprietary training and inference platform, namely Xinliu Platform, which supports customers' needs for development, training, and customizing images. Meanwhile, our intelligent computing network has significantly reduced the troubleshooting time for large model training from hours to minutes, ensuring service stability. We will continue to strengthen our intelligent cloud computing products and technologies to provide customers with stable and fast computing services.

In addition, both our public cloud and enterprise cloud solutions use cases have achieved compatibility with DeepSeek.

总体而言，在高质量可持续发展的战略引领下，我们的业务基本面得到了根本改善，进入了全新的健康发展周期。展望未来，我们将同步推进收入增长与利润提升，保持与小米金山生态深度合作，全面理解和探索新AI机遇，为我们的客户、股东、员工和其他利益方持续创造价值。接下来，有请CFO Harry为大家介绍第四季度财务业绩，谢谢。

In summary, led by the high-quality and sustainable development strategy, our fundamentals have substantially improved, and we have entered into a new era of healthy growth. Looking ahead, we aim to simultaneously expand our revenue scale and improve profitability, deepen cooperation with Xiaomi and Kingsoft ecosystem, comprehensively understand and explore AI opportunities, thereby continuously creating value for our customers, shareholders, employees, and other stakeholders. I will now pass the call to our CFO Henry to go over our financials for the fourth quarter 2024. Thank you.

Henry He (CFO)

谢谢周总。 Thank you, Mr. Zhou, and thank you all for joining the call today. We are very pleased to conclude 2024 with strong financial performance. I will now walk you through the financial results for the fourth quarter of 2024. I would like to highlight the following four areas of the progress. Regarding the performance of this quarter, first of all, we overturned from loss to profit in non-GAAP operating profit for the first time since our inception in 2012, demonstrating our strong execution in our high-quality and sustainable development strategy in the past two years. Second, our revenue has been growing for three consecutive quarters year-over-year, and this quarter, we outpaced all other listed public internet cloud companies and achieved a very good speed growth rate of 30% in total revenue, reaching RMB 2,232.1 million.

Third, our AI growth split increased by around 500% year-over-year to RMB 474 million, accounting for as high as 34% of the public cloud revenue. It has achieved a three-digit year-on-year growth for six consecutive quarters. Fourth, last December, our shareholders approved the revenue from connected parties of Xiaomi and Kingsoft Group for the next three years of RMB 11.3 billion, around 10 times over the revenue of 2023, providing solid support for the company's revenue and profit growth. We believe we are well on track to support the ecosystem's fast-growing demands and build up solid cloud infrastructures to support ecosystem AI development. Particularly in this quarter, we are very pleased to see that our revenue from Xiaomi and Kingsoft increased by 76% year-over-year, verifying the effectiveness of our ecosystem strategy. In terms of outlook, we have been firmly executing our AI strategy and the Xiaomi Kingsoft ecosystem strategy.

Going forward, we're expecting our improvements in both revenue scale as well as profitability. First of all, we believe the revenue growth of both public cloud and enterprise cloud will accelerate in 2025. Second, we expect our profit profile will continue to expand, and we expect to deliver a positive non-GAAP operating profit in the full year of 2025. Now, let me dive into the details of our financials. Total revenue of this quarter was RMB 2,232.1 million, reflecting a 29.6% year-over-year increase. Of this, revenue from public cloud services was RMB 1,409.8 million, up 34% from RMB 1,052 million in the same quarter last year. This growth was primarily driven by a significant increase in AI-related business, whose growth split reached RMB 474 million.

Revenues from enterprise cloud services reached CNY 822.3 million, up from CNY 670.3 million in the same quarter last year, primarily driven by increased demand in select verticals and growth in the Camelot IT services. Total cost of revenues was CNY 1,806.2 million, up 22.9% year-over-year, which was in line with our revenue expansion. IDC costs dropped by 2.6% year-over-year from CNY 740.4 million to CNY 721.5 million this quarter, reflecting the strategic scaling down of our CDN services and a better rack utilization. Depreciation and amortization costs increased from CNY 146.9 million in the same period last year to CNY 343.1 million this quarter, mainly due to the depreciation of newly acquired GPU servers. Solution development and service costs rose by 10.8% year-over-year from CNY 502.9 million to CNY 557 million, driven by expansion in Camelot personnel to support revenue growth.

Fulfillment costs and other costs were RMB 102.4 million and RMB 82.2 million this quarter, respectively. We have recorded a higher growth rate in our gross profit than the industry. Our adjusted gross profit for the quarter was RMB 427.7 million, up 63% year-over-year, with an adjusted gross margin of 19.2%. On the expense side, excluding share-based compensation and impairment of long-lived assets, our total adjusted operating expenses were RMB 446.4 million, a decrease of 9.8% year-over-year and 9.1% quarter-over-quarter, of which our adjusted R&D expenses were RMB 169.6 million, narrowed by 27% from last quarter. Adjusted selling and marketing expenses were RMB 107.8 million, down from RMB 110.6 million last quarter, representing 4.8% of total revenues. Adjusted G&A expenses were RMB 169.1 million, slightly increased compared with last quarter due to fluctuation of credit loss.

Our adjusted operating profit for the quarter turned into profit was CNY 24.4 million, compared with a negative CNY 187.6 million in the same period of last year. Our non-GAAP EBITDA profit was CNY 359.7 million, compared with negative CNY 27.7 million in the same quarter last year. Our non-GAAP EBITDA margin achieved 16.1%, compared with negative 1.6% in the same quarter last year. We believe that our strong commitment to AI cloud computing development, strategic adjustments of business structure, and our strict control over costs and expenses bear fruit, and we are all well on track to further improve our profitability as well as scalability. As of December 31, 2024, our cash and cash equivalents were a total of CNY 2,648.8 million, providing a strong liquidity position to support operations and AI investments.

Meanwhile, with healthy accounts receivable arrangement for AI business in this quarter, we have achieved a net inflow of operating cash flow reaching RMB 570.2 million. We are able to generate cash internally and organically and support part of our further capital expenditure. Looking ahead, we remain committed to the principle of high quality and sustainable development. We expect accelerated revenue growth on an annual basis and a higher and more stable profitability in the coming years. Thank you. This concludes our prepared remarks. Thanks for your attention, and we are now happy to take your questions. Please ask your question in both Mandarin, Chinese, and English if possible. Operator, please go ahead. Thank you.

Operator (participant)

Thank you. As a reminder to ask a question, you will need to press star one and one on your telephone and wait for your name to be announced. To withdraw your question, please press star one and one again. Our first question comes from the line of Xie Dan Shang from CICC. Please go ahead. Your line is open.

Xie Dan (Analyst)

Hey, Zhou Zong, Henry, Nico, Clark, Lai Shanhao. 首先，祝贺公司取得了非常亮眼的业绩。我们看到公司过去两年在这个 GPU 算力端的投入，也在持续地推动 AI 收入的一个环比增长。公司也在持续地受益于这个金山小米生态的需求释放。我这边有两个问题想请教一下。第一个是我们看到今年的这个 DeepSeek R1 模型的发布，其实是引发了市场非常热烈的讨论。想请管理层分享一下就是如何看待 AI 近期的这个产业趋势的变化，对于云计算行业供需结构以及市场格局的影响。那么今年云面临哪些机遇和挑战？第二个问题是想请管理层更新一下我们对于 2025 年的一个资本开支的计划。如果说在这个供给端我们面临更严格的限制的话，那公司有哪些应对措施？我快速地翻译一下我的问题。

Thank you, management, for taking my questions. My first question is regarding the industry outlook. Could management share your views on how the recent industry trends in AI are impacting the supply and demand structure, and also the market landscape of the cloud computing industry, and what opportunities and challenges does Kingsoft Cloud face? Secondly, could management provide an update on the 2025 capital expenditure plan, and what measures will the company take to address the possibly more stringent supply-side restrictions? Thank you.

Henry He (CFO)

model open-sourcing and in model capabilities based on reinforcement learning, or in multi-modal reinforcement learning model capabilities. This has, to some extent, also driven internal competition within the large language model industry. In terms of training, this is also somewhat helpful for us. That is all I will add. Okay. Allow me to quickly translate. Mr. Zhou Zong mentioned that DeepSeek by itself is still a general large language model. The reason that it has caused such widely received discussion is because it has proved that the technology company in China can also achieve such impact in the world, can also achieve such height in terms of technology and product, and therefore has helped the revaluation of Chinese technology companies. Secondly, as we all know, the fact of DeepSeek, the advantage of DeepSeek is that it greatly decreased the deployment cost.

This for Kingsoft Cloud is also a beneficial factor, as we have seen that this round of DeepSeek heat has also caused the Kingsoft Cloud share price to increase a lot. I have actually talked about this back in 2023, where I used to say that the advance of AI has essentially removed the glass ceiling of cloud business. Therefore, the release of DeepSeek R1 actually will accelerate the deployment of large language models in China. Therefore, more enterprises and more individuals will start to use large language models. It is a wider pushing kind of power for the advancement and development of AI and cloud industry in China. Kingsoft Cloud as an active and proactive participant of the market will definitely benefit from it.

In summary, I would say that this is essentially a broader sort of education to individuals and industries about the strategic value of large language models and of AI. Therefore, people are increasingly getting to know Kingsoft Cloud as well in that regard. Secondly, as our SVP Liu Tao mentioned, from a technology perspective, the advance of DeepSeek has achieved optimization and optimization in engineering methodologies, which enabled the lowering of costs and increasing the penetration of applications. We have been seeing a lot of potential customers, both from the government and enterprises, as well as from the internet space, have increasing application and adoption of AI and DeepSeek models. Secondly, the DeepSeek success has essentially increased the application of enhanced learning in the large language model and AI industry in China in general.

We are seeing increasing competition and increasing technological improvement in the AI space in China. That includes enhanced learning as well as multimodal use cases. Thank you.

Tao Zou (CEO)

Thank you, Sophie, and Mr. Henry. Happy to answer your second question regarding the capital expenditures. I understand it's a very important one for many of the investors and analysts. There are probably a few points I want to address here. First of all, before I go into the numbers, I want to mention our total investment into AI is comprised by two parts. First part is our OPEX, which is primarily spending to the data centers, which include the racks, rentals, electricity payments, and all the related expenses.

On this part, Kingsoft Cloud adopts a very efficient asset light model, which means that we do not have to spend a huge amount of money upfront to build the warehouse and the real estate properties ourselves. We have a very good partnership with long-term arrangements, for example, with telecom companies and other IDC companies, which can reduce our leverage and increase efficiency. It is the first part. On the second part, which is our capital expenditures, we buy efficiently, acquire the necessary computing powers, i.e., the servers and other network equipment. Basically, it is A plus B. I can share some numbers. In the past four quarters, we have already spent on the part A, which is OPEX on the data centers, around CNY 2.9 billion on the AI data center.

We spent on the part B of the capital expenditures around RMB 5-6 billion to acquire the necessary computing equipments and the network equipments. If you are putting those two numbers together, in addition to the research and the development investment as well, in the past probably four quarters, we spent around RMB 8-10 billion in total. Going forward, I think for the entire investment into the capital expenditures and also the OPEX going forward, I think the RMB 10 billion total AI investment for the full year 2025 is just a start.

My third point is, given the strong important support from our shareholders, including both Xiaomi and Kingsoft Group, as we have already disclosed and approved by the shareholders in December 2024, we have arranged the leasing and other off-the-balance sheet arrangements to support our ongoing and growing investment into AI on both OPEX and CAPEX. We also want to remind the audience that the existing cash on our balance sheet is not the limit of our capacity into AI investment because of the leasing and other alternative financing arrangements we arrange directly with our shareholders. I think we are in a good position with a strong demand. We basically can on par to increase the investment. I think given the revenue growth rate, we have already delivered above the industry average.

I think our spending will also catch in pace and maybe also accelerated and above industry average going forward, especially for 2025. Thank you.

Nicole Shan (IR Director)

Thanks, Sophie. Your next question, please.

Operator (participant)

Thank you. Please stand by. Our next question comes from the line of Brian Gong from Citi. Please go ahead. Your line is open.

Brian Gong (Analyst)

Hey, Zhou Zong, Henry, Clark, Nico, 各位领导，大家晚上好。首先恭喜非常好的业绩。我这边有两个问题。第一个问题是关于2025年的收入的增长的具体的一个预期，能不能拆分下背后的 driver，比如说我们今年 AI 收入我们觉得能增长有多少，然后像小米和金山体系的关联交易，刚才您提到的，这块能实际今年能贡献多少的收入。这是第一个问题。第二个问题就是想很快问一下，我们这个 margin 表现得非常好，想问一下我们觉得长期的利润率能够达到多少。我自己翻一下。

Thanks management for taking my question and congratulations on the solid results. Regarding the expectation for 2025 revenue growth, could you kind of share your thoughts and break down the drivers and how much growth we expect for AI-related revenue and how much revenue contribution from Xiaomi and Kingsoft Group? Our second question is regarding margin. Our margin performed quite well in the first quarter. I'd like to share your thoughts on our long-term profitability trend. Thank you.

Henry He (CFO)

Thank you, Brian. This is Henry. Happy to share some color. As I mentioned in my prepared remarks, I think a few things I just want to emphasize. First of all, in Q4 2024, the 30% top line on the YOY growth has exceeded even the top players in the industry. I think that demonstrates all the verticals, including both public cloud and enterprise cloud, and within the public cloud, AI and the non-AI are all growing at a very healthy rate. Without these fundamental support, it is very difficult to achieve this high growth rate. That is actually the first point I want to mention. The second part is, given we already completed our business strategy adjustments in the first half of 2024, we do have a very healthy split and a mix of the client structure.

As you may know, for example, the CDN business used to contribute much higher rates, and right now it is only probably 10-15% or even lower going forward in terms of how you are looking at that. I think our growth has been built on a very solid foundation. That is why in the prepared remarks, I also mentioned that the growth rate, the keywords for 2024, is acceleration on a full annual basis. AI is going to be a very important driver, but other verticals, as I mentioned, in different product lines will contribute healthy growth going forward as well. It is actually the first question I just want to share. The second part regarding the margin, while we do not have official guidance for the margin, as you can see, our gross margin has been improving very consistently for the past few quarters.

I also want to mention our operating cash flow in the past three quarters, starting from Q2 2024, has been positive as well. That is going to be a very important leading indicator for our margin expansion. Given the AI contribution of our entire margin profile, especially for the EBITDA and OP side, we think the EBITDA and the operating profit will grow at a more fast pace compared with the gross margin. We are happy to see that our EBITDA margin actually already exceeded the industry average and maybe is in the top tier of industry peers. I think the EBITDA margin improvements, which is also a leading indicator of the OP cash as well as the free cash flow, will drop down to the operating profit side. I think our intention for management is two things.

First of all, our intention is to keep a healthy, steady pace of margin expansion for all three key indicators from gross margin, EBITDA margin, and OP margin. The EBITDA margin and the OP margin will be accelerated and improved a little bit faster than the gross margin. The third point, given the cash flow is also important, and we are going to keep a very close eye on cash flow margin expansion, that will also give us a good momentum and support to recycle the cash to input the investment into the new incremental AI investment. I think these are probably the few points I just want to mention. Thank you, Brian.

Nicole Shan (IR Director)

Thank you. We're happy to hear more.

Operator (participant)

Thank you. We'll now move on to our next question. Our next question comes from the line of Linlin Yang from GF Securities. Please go ahead. Your line is open.

Linlin Yang (Analyst)

周总，何总，关礼堂，晚上好。我是广发的杨林林，然后我想请教两个问题。一个是关于今天刚才也提到，除了小米跟金山生态客户是我们今年非常重要的一个增量，那怎么看这个AI推理需求，然后以及这个AI推理这一块，大概比如说今年或者明年大概能占到多大的一个比例。那第二个问题是关于因为我看到我们传统的工业那一块，现在设备的减值或者折旧基本上已经折得差不多了，那这一部分未来它是否还会产生收入，那这个传统雇用营业务后续的一个业务增长怎么看。那我简单翻译一下。

My first question is about the AI inference. How do you think about the demand of AI inference and how much contribution will it be? And my second question is, after the equipment in the legacy public cloud business is fully depleted, will it continue to generate revenue? What is your outlook for this part? Thank you.

我先回答第一个问题。你刚才谈到除了小米金山自身的之外，对于整个推理的需求的是怎么看，这是我的观点。刚才谈到了上一个问题，谈到DeepSeek的时候，DeepSeek的时候我就谈过我的观点，这一波一定会加速这个大模型到千行百业。实际上从DeepSeek出来之后，我们自身从公有云也好，从行业也好，从政企云也好，从客户侧这边的，我就单从客户数量这个角度，其实还是有了非常大的增加，甚至也包括我们本身去跟一些，虽然不一定是我们，现在还不是我们客户去沟通的时候，这种意愿也强烈了很多。这是从客户数，但是从规模数，就是你像训练至少都是128，甚至256，甚至512集群，这种成规模的上量来看，我们目前还没有看到。所以综合在一起是什么呢？似乎爆发出这么一个趋势，但是从规模、从体量的单一客户这种，我觉得我们目前还是没有看到。但是对未来，我们还是有很大的信心，尤其是一些行业类客户。所以这是我基于目前的一个判断。当然你刚才谈到小米和金山本身的，说实话，那其实还是非常旺盛的，几乎所有的小米的生态的产品，包括它的汽车、手机等等，以及它的智能家电这些，未来都会是去接入。办公就更不用说了，我们也今天刚发完业绩，反正应该还没有披露具体的用户数，但其实上增长的也是挺快的。好吧。第二个问题，Henry来讲。第一个是不是要翻译一下？

Tao Zou (CEO)

Okay. Allow me to very quickly translate. For those demands of inference outside of the Xiaomi and Kingsoft ecosystem, what we have been seeing is, on one hand, as I was responding to the DeepSeek question earlier, the advent of DeepSeek obviously accelerated the application of large language models and AI to various sectors. We have been seeing a huge amount of increasing in customer numbers, as well as the strong desire and willingness for them to adopt and apply AI applications. However, on the other hand, if you look at the number of the amount of actual business, for example, for one single customer, we might be needing a cluster of 128 units or 256 units or even 512 units cluster. This kind of single customer demand we have not been seeing yet.

In summary, the trend of penetration of AI and DeepSeek is very clear for inference, but we think it still might take some time for that to be shown on the financial statements. Secondly, for your question about the inference demand within the Xiaomi and Kingsoft ecosystem, the answer is that it's going to be very strong. Essentially, all of the products and services across the Xiaomi ecosystem, including the cell phones and IoTs, including the electric vehicles, all of them are going to be supported by AI and large language models. On the Kingsoft side, needless to say, actually, the Kingsoft Office AI, the WPS AI user account has also been increasing quite strong.

Henry He (CFO)

Thank you, Yanli. Regarding the question of asset and depreciation, this is Henry. Happy to offer also some colors. I think, first of all, if our audience and investor do some analysis and a comparison, Amazon, AWS, including Google and Microsoft, are all using five years of DNA policy for the servers and computing powers. Kingsoft Cloud, we are conservative in terms of managing the financial policy. Right now, regardless of the server's type and equipment, we are using four years of D&A policy, which means that we basically, if we compare with five years, we actually overestimate certain costs. In turn, there is a potential to release certain profitability down the road. That's actually my first point.

My second point is, in the late of 2023 and early 2024, we already at that time forecast that the industry will have a major shift from the traditional IaaS to the AI cloud. That is why at that time we did a full round of internal review of the asset distribution and our internal resources, and we decided to really filter out certain assets from CPUs and the traditional assets. We did the asset impairments and some of the changes and provisions of asset valuation. Right now, all the assets on the balance sheets, including the RMB 5 billion capital expenditures required for the computing power in the past few quarters, which currently sitting on our balance sheets, are all healthy and the revenue and the profit of the generating assets, and it actually can match with our revenue incremental opportunities.

I think right now we do have a kind of healthy asset balance sheets on the left-hand side and also have a good potential to improve the shareholder return and the profitability and the cash on the return, especially OP cash on the return on the right-hand side. I think we are in a very good balance on the left to the right, and we can improve the foundation of our business model and efficiency going forward. Thank you.

Nicole Shan (IR Director)

Thank you. Very nice question.

Operator (participant)

Thank you. Our next question comes from the line of Wenting Yu from CLSA. Please go ahead. Your line is open.

Wenting Yu (Analyst)

cloud revenue? I'll translate the question. We've seen that prices for high-performance AI service have been falling in 2024, while inference services are relatively well supplied. How will this affect our pricing strategy for GPU cloud revenue and its impact on our AI cloud revenue and earnings? Thank you.

Henry He (CFO)

Thank you, Wenting. First of all, thanks for the very good question. As you know, on this quarter, we deliver probably, I say probably, I put a big disclaimer, but maybe true, the highest growth rate in the industry for all the public companies in the internet sector, which has cloud business. Number two, given the gross profit, we already reached probably also on the top tier across all leading players on this quarter. You may understand that we want to remain relatively conscious about our business secrecy and very sensitive information. We do not want to hurt investor interest because we share too much information on our success tips. On that backdrop, what I can share is I think we do have a higher margin compared with traditional CPU-centric ICE services. I think it is kind of first topic, first point.

The second point is, depends on the client demands, we do have a good mix of all kinds of different computing formats, which can fit into the client demands. We are charging them as the demand goes up. It is a supply demand, which actually favors Kingsoft Cloud as a very important supplier in this current market. Number three is the pricing on the GPU, as you know, it actually links to core production requirements of all our major clients. They use AI and computing as the important production and productivity tools and important embedded functions in their workflow. In that sense, as you understand, it is always about value.

I think our value proposition has been lifted from traditional IT services, which may have a limited ceiling of profitability, to a much higher potential in the current market because the GPU AI services become so important for our client internal workflow and for their final products. They are willing to pay much higher fees and pricing because of the necessity of the technology value. Putting all things together, as I mentioned, I think we do have a very different pricing model for the AI-related services, which are higher than the CPU and the traditional AI. We already shifted our strategy from a traditional IT cloud services company to an AI cloud and solution services company going forward. I think that's actually providing very fundamental different changes and approach for the pricing.

As I mentioned, because the pricing is so sensitive and right now we are doing so good compared with industry average, we do not want to say too much regarding our strategy because it is a business confidential, as you may understand. Thank you.

Nicole Shan (IR Director)

Thank you, Wenting. This concludes our earnings call today. Thank you all once again for joining us. If you have any further questions, please feel free to contact us. Look forward to speaking with you again next quarter. Have a nice day.

Operator (participant)

Please complete today's conference call. Thank you for participating. You may now disconnect.