Good day, ladies and gentlemen. Thank you for standing by. Welcome to Alibaba Group's March quarter and full fiscal year 2026 results conference call. At this time, all participants are in listen-only mode. After management's prepared remarks, there will be a Q&A session. I would now like to turn the call over to Lydia Liu, head of investor relations of Alibaba. Please go ahead.

Good day, everyone. Thank you for joining Alibaba Group's March quarter and full fiscal year 2026 earnings call. On the call with me are Joe Cai, Chairman, Eddie Wu, Chief Executive Officer, Toby Xu, Chief Financial Officer, Jiang Fan, Chief Executive Officer of Alibaba E-Commerce Business Group. As a reminder, this call is being webcast live. A replay of the call will be available on our website later today. On this call, we may make forward-looking statements and discuss certain non-GAAP financial The forward-looking statements reflect management's current expectations that are subject to risks and uncertainties. Our GAAP results and reconcilations of GAAP to non-GAAP measures is included in today's earnings press release and investor presentation. Our comments will be on year-over-year comparisons, unless we state otherwise. And with that, let me turn the call over to Eddie.

Welcome to Alibaba Group's fiscal year 2026 fourth quarter earnings call. Over the past quarter, Alibaba's high-intensity investment in our two strategic priorities of AI plus cloud and consumption is rapidly translating into tangible business results. with group revenue growing 11% year-over-year. This quarter, Cloud Intelligence Group's external revenue growth accelerated to 40% and AI-related product revenue achieved triple-digit growth for the 11th consecutive quarter. China e-commerce CMR grew 8% year-over-year on a like-for-like basis and the quick commerce market achieved significant unit economics improvement while maintaining market share. We are at a pivotal inflection point in the evolution from conversational chatbots to autonomous AI agents, which is directly driving explosive growth across three core workload categories, training, inference, and agent orchestration. Against this backdrop, Alibaba's AI has moved beyond the initial investment phase and progressed commercialization at scale. Next, let me walk you through four areas in detail. AI commercialization, cloud infrastructure, the AI application ecosystem, and our consumption business. First, the AI and cloud commercialization inflection point has arrived. This quarter, Cloud Intelligence Group's annualized AI-related product revenue has surpassed 35.8 billion RMB, continuing to maintain triple-digit growth. AI-related product revenue now accounts for 30% of Cloud Intelligence Group's external revenue. We expect that in about one year, AI-related product revenue will cross the 50% threshold, becoming the primary engine driving the cloud business's revenue growth. As a result, Cloud Intelligence Group's external revenue growth is expected to continue accelerating beyond its current 40% rate over the coming quarters. Given the certainty of long-term AI demand and our full-stack technology advantages, we expect this trajectory to sustain strong growth over the medium to long-term. This reflects AI's role in driving a comprehensive upgrade of Alibaba Cloud's entire business as its growth engine fully pivots from traditional compute and storage to models, AI compute, and agent services. We're also seeing exponential growth in AI model and application services revenue. A new revenue engine driven jointly by foundation model services and AI native software. Over the past three months, token consumption volumes on our model services platform grew substantially quarter over quarter as enterprise customers accelerated their shift from simple tasks to production scale and complex workloads, driving continued growth and demand for model and application services on the Model Studio platform. We expect Model and Application Services' annualized recurring revenue, ARR, inclusive of the Model Studio platform to surpass 10 billion RMB in the June quarter and 30 billion RMB by year-end. The hard margin profile of this revenue stream is becoming increasingly apparent, making it a source of healthy, high-quality growth. Second, our AI infrastructure underpins our full technology stack and constitutes a durable moat. T-Head's proprietary GPU chips have achieved scaled mass production with over 60% of compute capacity already serving external customers across internet, financial services, and autonomous driving verticals. As the only AI cloud provider in China capable of delivering self-developed AI chips at scale, we've secured autonomy over our compute supply chain while providing customers with highly competitive AI inference and training services. In an environment of compute scarcity, this structural advantage is favorable to our revenue growth and gross margin improvement. At the same time, our cloud products are accelerating their AI-oriented upgrade. The surge in agent workloads has significantly elevated demand for traditional cloud products built around CPU storage and containers, and we're upgrading these into infrastructure solutions optimized for the agent era. Third, at the application layer, we've built a complete closed loop spanning AI-native software to a full-agent ecosystem. Alibaba Token Hub ATH continues to launch new products, connecting consumer and enterprise environments with breakthrough progress in AI-native software and coding agents. The Q1 model continues to iterate across reasoning, coding, and agentic capabilities. On the enterprise side, we've launched a range of products spanning intelligent workplace tools, AI coding, and business operations management, helping enterprises unlock greater productivity. On the consumer side, the QN app fully integrated Taobao and Tmall's commerce service capabilities on May the 7th. With this, Q&A is now deeply embedded across the ecosystem, spanning Taobao, Alipay, AMAP, and Fliggy, making it China's first all-in-one personal assistant to seamlessly bridge everyday life, productivity, and learning. Fourth, across our consumption business and at the group level, we're prioritizing long-term value. Beyond AI, our consumption strategy continues to progress steadily, with CMR growth rebounding significantly. This quarter, CMR grew 8% year-over-year on a like-for-like basis as we continue to improve user experience and merchant operating efficiency. The quick commerce business achieved significant unit economics improvement while maintaining stable market share, market scale. In summary, the return on our investments in AI plus cloud and consumption are increasingly AI plus cloud revenue growth is accelerating with improving margins. Model and application services, ARR, continues to grow at pace. And operating efficiency across our consumption business continues to improve. Facing the historical opportunity that AI represents, Alibaba is at a pivotal juncture where our technology investments are beginning to pay off commercially. We'll maintain our strategic resolve and leverage our full-stack AI capabilities to support long-term growth. That concludes my prepared remarks. Next, I'll hand over to Toby to walk you through our financial results.

Thank you, Eddie. Our strategic priorities remain laser-focused on AI plus cloud and consumption businesses. Multiple growth catalysts, including technological advancement and business innovation, are aligning to create strong tailwinds. On AI plus cloud, our full stack capabilities span models, cloud infrastructure, and applications with established leadership in every layer, the strong growth of our AI plus cloud businesses and the clear path to monetization of our mass platform give us confidence to make significant investments to extend our leadership. On consumption, we achieved a strong CMR growth on a like-for-like basis during the quarter, And our quick commerce business continued to improve UE and AOV quarter over quarter. Now let's look at the financial results for this quarter. On a consolidated basis, total revenue was R&B $243.4 billion. Excluding revenue from Sunnah and InTime, revenue on like-for-like basis would have grown by 11%. Total adjusted EBITDA decreased 84%, primarily due to our strategic investments in technology businesses, quick commerce, and user experience, partly offset by the improved operating results supported by continued growth in consumer management service in the cloud business, and enhanced operating efficiencies across various businesses. Our gap net income was R&B $23.5 billion, an increase of 96%, primarily attributable to the year-over-year increase in net gain from mark-to-mark changes of our equity investments and disposal losses of Sunna and InTime in the same quarter last year, partly offset by the decrease in adjusted EBITDA. operating cash flow was an inflow of R&B 9.4 billion. Free cash flow was an outflow of R&B 17.3 billion. We are reinvesting our operating cash flow to enhance our competitive advantage in AI. As of March 31, 2026, we held approximately U.S. dollar 38 billion in net cash, excluding that with maturities beyond five years, our net cash position stands at approximately US dollar 59 billion. This balance sheet strength gives us confidence to invest for growth. Now let's look at our consumption businesses. Revenue from China e-commerce group was R&B $122 billion, an increase of 6%. Customer manual revenue increased by 1%. To help merchants grow their businesses and increase willingness to spend on our platform, we upgraded our business development program for select merchants during the quarter, under which the level of platform subsidies for these merchants is directly tied to their marketing spend on our platform. For accounting purpose, such subsidies previously recorded as sales and marketing expenses are now recorded as a contract revenue item to CMR. Accordingly, CMR grows 1% year over year during the quarter. Excluding the contract revenue impact from the program on a like-for-like basis, CMR would have grown 8% year over year. Revenue from our quick commerce business increased 57% to RMB 20 billion. The quick commerce business further improved UE and increased AOV quarter over quarter, primarily driven by order mix optimization. Alibaba China e-commerce group adjusted EBITDA was RMB 24 billion, a decrease of 40%, primarily due to the investment in quick commerce, user experience, and technology, while there's positive contribution from customer management service. Excluding loss from our quick commerce business, our Alibaba China e-commerce group EBITDA would have been stable year over year and will fluctuate quarter over quarter due to significant investment in merchant retention and the user experience. Revenue from AIDC grew 6% this quarter. AIDC's adjusted EBITDA loss narrowed significantly year over year, approaching break-even, driven by a combination of logistics, optimization, and operating efficiency. The unique economics of AliExpress's choice business continue to improve substantially on a sequential basis. Next, let's look at the business updates and results of Cloud Intelligence Group. Our cloud business delivered another quarter of accelerating growth. Revenue from external customers accelerated to growth 40%. AI-related products continue to lead this momentum. We delivered our 11th consecutive quarter of triple-digit growth in AI revenue. Its share of external cloud revenue continued to increase, now account for 30%. This quarter's AI revenue is R&B $9 billion, and the annual revenue run rate is R&B $36 billion, or U.S. dollar $5.3 billion. This is a clear reflection of the scale and acceleration in our AI business. The adjusted EBITDA margin remained relatively stable at 9.1%. All other segment revenue decreased by 21% to R&B $65.5 billion, mainly due to the disposal of Sunna and In-Time businesses, as well as the decrease in revenue from China, partially offset by the increase in revenue from Freshapal and AMAP. All others adjust the EBITDA was a loss of R&B 21.2 billion, primarily due to the increased investment in technology businesses, including foundation models and the consumer-facing QIN APP. As we close this fiscal year, we remain committed to delivering consistent shareholder returns. Our board of directors has approved an annual dividend of US dollar 1.05 per ADS. We will continue to invest decisively in AI and consumption businesses, where we see significant long-term growth potential in our competitive advantages of compounding. We believe these investments to deliver growth and returns over time, ultimately creating greater value for our shareholders. Thank you. We will now open for Q&A.

Hi, everyone. You're welcome to ask questions in Chinese or English. A third-party translator will provide consecutive interpretation. In the case of any discrepancy, our management statement in the original language will prevail. 大家好,欢迎提问。我们有第三方工作人员提供实时中英文交替传译。 如有任何意义,请以管理层原始语言所做的陈述为准。 Operator, please start Q&A session. Thank you.

Thank you. If you wish to ask a question, please press star 1 on your telephone and wait for your name to be announced. If you wish to cancel your request, please press star 2. If you're on a speakerphone, please pick up the handset to ask your question. To give people the opportunity to ask questions, please keep yourself to no more than one question at a time. Your first question comes from Ronald Kung with Golden Sacks.

Thank you, Joe, Eddie, Toby, Fan, and Lydia. And thanks for sharing the very sizable AI mass and applications ARR scale and the target for the first time. So I just want to ask how much of that ALR is driven by our in-house models like QAN versus third-party models? And given the recent token price hikes, what would be the implications to MASS and also our cloud margins as a result? Thank you.

I thank Joe, Eddie, Toby, John Van and Lydia. I also thank you for sharing the AI MASS and the ability to use the ability to use.

Thank you. 现在我们因为这个季度我们刚刚公布了一个最新的一个数字 也就是我们的模型及应用模型及应用服务收入 这个收入的主要的业务现在就现在而言主要的业务构成 其实分成两个部分就是我们的百链的MaaS的API服务 以及我们的AI原生软件的订阅值的这些这些收入 那么在现在这两部分的收入里面 绝大部分都是还是来自于 百链的MaaS的API服务 但同时阿里尼的百链 其实又是一个相对开放的一个平台 在我们这个平台上 我们hosting字眼的模型 也会hosting三方的开源模型 包括三方的B缘模型 那就现在我们的收入规模上来说 绝大部分还是我们的字眼模型 包括我们的签问 筋模 以及包括我们的语音模型

包括我们的视频模型 Thank you for that question This quarter marks the first time that we announced the latest figure for model and application service revenue That really comprises mainly two things On the one hand, it includes revenue from API calls on MAS on our BaiLian platform and it also includes revenues from our AI software subscriptions. At present, most of the revenue is coming from the first of those two pieces, but this is an open platform, so we are providing access both to our proprietary models as well as third-party models, including open-source models and closed models. But for the time being, most of that revenue is coming from our own proprietary models, including Q1 as well as Timor as well as our voice and video generating models.

非常复杂的推理任务 那在这个导致于客户的 对于模型的推理的需求 持续直线的上升 那在这个过程当中 由于可以帮助客户 完成更复杂的工作任务 所以整个API token的价格 也客户的接受度 其实虽然价格进行了上调 但是客户的接受度 其实还是非常高 而且需求也是非常持续的大 而且在这个需求 就现在而看 我们实际上 我们的供应 其实还没有办法 完全地满足这些客户的需求 排队的客户还有很多 所以就现在这个情况下 我们觉得 首先MAS业务 它也本身的一个 对比S业务来说 相对毛利就会 天然就会高一些 然后还有一些 还有几个比较重要的 一些点吧 第一个点就是说 整个推理的技术的 这个发展还在持续的发展在这个所以每个季度我们都会看到在推理技术上的一些优化带来的成果可以是我们在单服务器单卡的token产能上面产生一些持续的持续的增量效果同时由于模型的能力也在持续的加强而且模型的价格在未来的一年一两年内 我们看到持续的应该还是会是一个价格提升的一个过程

会对我们的毛利率产生一些非常积极的影响

Your second question was also a really important one because in the past quarter over the past few months we've seen a very large shift in the market where AI is shifting from functioning as a conversational chatbot to providing agentic capabilities. These agents are increasingly capable of solving for very complex problems, meaning that they need to do a lot more inferencing than in the past. And precisely because these agents can help to solve very complex tasks, customers' acceptance for higher prices, and we have increased per token prices, is good and the demand continues to be high and growing. In fact, our ability to supply this demand is not able to keep up with all the growth in demand. We actually have a lot of customers still waiting to access the service. Inherently, MAS will have higher gross margin than IAS. That's important to know. And I can also add a few important points on top of that. First is that the development of reasoning or inferencing technology still continues to advance. So every quarter we're seeing new results in terms of optimization in reasoning, in inferencing, with continuous incremental effects in terms of the token capacity of a single server and a single card. At the same time, as the capabilities of models continue to strengthen and price of the models continues to increase in the next year or two, we see that this should be a process of continued price improvement. So I think from this point of view, the rapid growth in this business over the next few quarters will result in a very positive impact, a gross profit margin.

所以从这个角度来看 我觉得我们未来记住 因为MAS业务的这个飞速增长

Your next question comes from Kenneth Fong with UBS.

Hi, good evening management. Thanks for taking my questions and congrats on the very strong progress on the AI. I have a question regarding the return on invested capital on the AI investment. So while our AI investment have driven impressive 40% cloud growth, they have also created significant drag on the group free cash flow as well as our EBITDA. So how should investors assess the return on this investment? And what's the management framework for balancing the aggressive AI spending versus earnings ability?

Thank you. 得以实现40%的云业务的增长 这个非常好 但同时呢 对于集团的这个自由现金流 以及EBTA带来了一定的这个拖累效果 那么我想知道集团如何评价这个投资的回报率

然后你们采用什么样的管理框架来平衡 就是一方面是AI投入 另一方面是公司的这个营收能力怎么平衡 OK那谢谢你的问题 我想我先来这个回答第一部分 因为我觉得大家可能现在特别是这个季度看到我们的这个自由现金流是一个负数 所以大家可能会比较关心我们是怎么来控我们的这个自由现金流的 我想首先大家可以看得到这样一个自由现金流的负数 最关键的还是我们在过去一年的在AI方面的投入 那正是因为我们看到了这样一个历史的一个机遇

我们是十分坚定的在这个方面进行这样的一个投入 所以这是最主要造成的一个整体的一个自由现金领维护的这样一个情况 Thanks Kenneth for that question This is Toby and I'm going to start by answering that first question Because I think it's important and of interest to everybody The question is the reason for the negative free cash flow and how we are managing that. So starting there, the answer is that the negative free cash flow is primarily due to the very significant investments we've been making in AI over the past year.

And we've been extremely resolute in making those investments precisely because we've seen the historic opportunity of AI. 往未来再看两年的话,我想我们的这样一个投入还是会继续的十分的坚定,因为这样一个窗口期可能对我们来讲就是几年的时间,所以在投入的角度我们会继续很坚定,同时回来再看从我们经营现金流的角度,其实并没有这个有大的一个变化,这里我想谈两个方面, 第一个方面从我们消费业务的角度 淘宝天猫就我们最重要的一个 经营现金流的产生的这样一个业务 这个整个经营现金流的产生还是十分稳定的 而未来两年随着整体的我们闪购的 这个相对的这个亏损的大幅的收窄 整体的消费业务会加上我们还有一个 AIDC的整体的从一个亏损

所以这是第一个我想说的

We've been very resolute

往盈利的方向走 整体的消费业务的 在未来两年来看 整体的现金流 经营现金流 会是十分的一个正向的一个发展

in making those investments over the past year and looking forward to the next two years we intend to be equally resolute in continuing these investments again because we see this as a critical window of opportunity that will be open for that period for the next couple of years. Additionally, there's really been no big change in the way that we look at cash flow. First of all, the major contributor of operating cash flow for the group is Taobao and Tmall, and that cash flow is very stable. And looking ahead over the next two years, in terms of quick commerce, the losses will narrow very substantially at the same time AIDC will develop from making a loss to being profitable so we see these developments over the next two years as being highly positive for our net cash flow 第二点 另外一个很重要的一点是 我们在云的这样一个基础设施的投资 刚才那个Eddy也提到了 会让我们的整个的云的收入 AI云的收入继续加速

同时整个的毛利率的改善 所有这些会进一步提升 我们在云这个板块的经营现金流 回来同样可以支撑我们 对云的基础设施的投入 第三点 我想说 我刚才提到了 我们自己的整个的balance sheet 还是十分strong的 我们目前的净现金 差不多380亿美金 如果我们拿掉 这个五年以上到期的债务的话 我们现在进现金大概有差不多590亿美金 所以这样一个十分坚实的这样一个资产负载表 也可以很好的支持我们的对这样一个云的基础设施的投入 最后我还想说是我们还有很强的资本市场的融资能力 可以通过不同的形式的市场融资去应对战略发展的需要 所以我这里我想先回应一下您刚才提到的关于现金流 包括我们怎么样准备好更充足的现阶楼这个问题 然后接下来看Eddie是不是有些补充 Another important point to be added is that our ongoing investments in cloud infrastructure

will increase the revenues that we can achieve from our AI and cloud offerings At the same time, we will increase gross margins in those very same offerings So in those ways, we expect that we can achieve higher net cash flow from our cloud and AI business, and that cash flow can in turn be used to support the development of the relevant infrastructure. An additional point is that we have a very strong balance sheet. As of March 31st, 2026, we held approximately US$38 billion in net cash, And if you exclude debt with maturities beyond five years, our net cash position stands at approximately 58 billion U.S. dollars. So that balance sheet strength also gives us confidence to reinvest for growth. And beyond that, I would also add that we have a very strong capacity for pursuing financing in capital markets. and we have the capability to raise capital from the markets as we need to support our development. So I wanted to open with that in response to your question on cash flows and I'll pause there to see if Eddie has anything to add.

我觉得AI更像一个制造业 也就是说 当我们要获得更多的收入的时候 我们其实要必须要去建立两个核心的工厂 这两个核心工厂的规模 会影响着我们未来有多大的一个收入规模 这两个核心工厂 一个我们可以成为叫做AI的训练工厂 一个叫做AI的推理工厂 那这两个工厂背后 其实是一个AI的数据中心的建设 而数据中心的这个建设 我相信一定会消耗比较大的集团的自由现金流 但是这些刚性的数据中心的这些基础设施的建设 我们在回报的路径上是非常清晰的 就是在我们这些数据中心 我们在2B的商业化这些路径上是非常清晰的 无论是云的AS的这种商业化的服务 还是通过我们的MaaS平台 以及通过我们的AI原生软件 去创造更多的模型之上的

也就是现在 我觉得在未来三到五年内的 这个需求的情况下 我们大量的投入的 AI数据中心的建设 这个投资回报是非常确定的

Thank you, you asked about our investments in AI and the ROI on those investments going forward So on top of what Toby's already said But I'd like to add a few notes regarding where we're heading. I think the best analogy is manufacturing. In other words, in order to be able to manufacture more and sell more in the future and achieve more revenue, what we're doing today is investing capital to build two factories, if you like. The first we can call the AI training factory. The second we can call the inferencing factory. And both of those factories need to be powered by our AI data centers, and that requires the investment of cash flow today. However, looking to the future, the pathway to achieving solid return on investment in those factories, in those areas, is very clear. on the 2B side by monetizing our 2B offerings, including our cloud-based IaaS, as well as MaaS, of course, and our AI native apps. And I can tell you that today there isn't a single card on our servers that is idle. So we see the ROI on this investment in the next three to five year period as being extremely clear.

这样的营收 我们几乎现在 在我们的服务器内 几乎没有一张卡是空

Next question, please.

Your next question comes from Thomas Chong with Jeffreys.

Hi, good evening. Thanks, management, for taking my question. My question is on quick commerce. I've talked about the improvement in Rui Yi in the prepared remarks. I just want to get some more color about the drivers behind in terms of AOV, subsidies ratio, fulfillment ratio, etc. And on top of that, I remember last time we talked about the outlook for quick commerce over the next couple of years. Is there any update or changes in terms of how we think about the landscape or the UE in the next few years? Thank you.

Thank you,管理层. My question is about the supply chain. 这一方面那么刚才在准备好的发言当中 你们已经介绍了UE得到了显著的改善 我想了解一下背后的驱动因素 包括这个是比单价还是补贴比率 还是履约等等各方面的一个贡献 另外的话上一季度的业绩发布会上 你们介绍了未来两年的这样一个 即是零售的展望 那么想知道从那个时候到现在 你们这种展望有什么样的新的变化 你们看待市场格局的方式有什么样的变化

包括这个EV方面和补贴等方面 我来回答一下这个问题 对 首先经过一年的投入 我们看到即是零售业务飞速发展 我们的市场份额取得了根本性的变化 那么对比这个三月季度 去年三月季度同期 就大规模的投入之前 Thank you very much.

Thank you. Well, first, as a result of our strong investments in QuickCommerce, we've achieved very rapid growth in QuickCommerce over the past year, marking a very fundamental shift in our market position. Compared to the same quarter last year, which was, of course, prior to all this large-scale investment, both our order volume and our market share have increased significantly. Overall, order volume was 2.7 times that of the same quarter last year, with non-food orders at three times. From April onwards, while maintaining order volume, we've continued to drive substantial improvement in UE through enhanced fulfillment, logistics efficiency, as well as order mix optimization. So we are confident that UE will turn positive by the end of fiscal year 27. 促进用户活跃度 满足用户多元化消费场景 拉动成交跟商业化以及物流基建等方面

那么推动淘天整体板块发展 我们继续看到闪购相关品类的带动明显 特别是食品生鲜等品类 那么也继续推动像河马 猫巢等技术零售相关业务的发展 那么我们的食物电商这个季度无论在成交跟CMR方面 这个就都取得了比较好的一个增长的势头 While optimizing UE, we will continue to innovate to improve the experience for both consumers and merchants, thereby sustaining our long-term competitiveness in QuickCommerce.

We're confident that our QuickCommerce business will achieve overall profitability in the future at new scale and market share. This quarter, QuickCommerce continued to generate synergies with our conventional e-commerce business, as demonstrated in driving customer acquisition, enhancing user engagement, fulfilling diverse consumer demands, increasing transactions, improving monetization, and supporting logistics infrastructure. In terms of categories, quick commerce continued to drive sales in various categories, especially food and fresh produce and health care, and contributed to fresh appos and Tmall supermarkets accelerated growth. So in our conventional e-commerce business, we saw GMV and CMR demonstrate strong growth momentum in the March quarter, and quick commerce played a vital role in driving that performance.

Next question, please. Your next question comes from Jia Longshi with Nomura.

Good evening, management for taking my question. I have a follow-up based on your opening remarks concerning mass. I'm wondering, first of all, what are the major advantages that Alibaba has when compared to the other major AI platforms in China, as well as Chinese AI startups? In the United States, we see that AI agents, and especially coding, are the fastest growth track in AI. I'm wondering when you think we'll see that kind of growth in China in terms of AI coding. We also know that Chinese customers are less willing than U.S. customers to pay. So do you think that that means that the future commercialization of Chinese AI coding products might have less potential than we see with the U.S. counterparts?

Thank you. 广度来说其实是我们会远远超过这些模型创业公司当然这些模型创业公司在专注于某一个模型或者某一个领域来说他们也有非常强的技术以及很快的很快的商业商业敏锐度所以他们的进展也是很快所以从这个角度上来说我们如果单纯在看Mars领域的话 我觉得这些创业公司 也是阿里云的合作伙伴 所以 但是对阿里巴巴而言 我们会更强调 我们在各种不同领域的模型 要做更广度的研发 也就是我们会做 像基于 coding能力为优先的 签问基座模型 像我们的视频模型 无论是万象 还是我们的happy house 包括更面向未来的世界模型 包括我们的语音模型 而我们相信未来在很多的业务场景上面 用户会需要多种不同的模型能力 综合来满足它的一些业务需求 所以这个是我觉得我们 与这些头部的AI创业公司相比的一个区别 同时和他们其实某种程度上 也是阿里云百链的合作伙伴 Thank you.

Well, the way we define our mass platform, Bailian Model Studio, is as an open AI inferencing platform. Certainly, at present, the majority of Bailian's revenue is driven by our own proprietary models. But in contrast to the AI startups in China, I think that we are investing at a much higher scale and across a much broader range of different model types. In contrast, those startups may tend to focus on a very narrow particular vertical segment and they can move rapidly ahead with that kind of focus. You know, and strictly in terms of our mass business, I think those AI startups really are partners rather than competitors. But at Alibaba, we particularly place emphasis on our model capabilities and developing them across all different spaces and all verticals to serve a very broad and diverse set of needs, including our coding model capabilities, including image-based models. so both Wan Xiang and Happy Horse as well as these new world models and of course voice models as well so we aim to provide all different kinds of models and this is different I think from those startups

and at the same time those startups are 您说的关于AI coding这个问题也很好 其实您说何时能迎来类似的帧数 其实我觉得我们的判断是在中国就现在已经迎来了类似的增速 也就是从我们自己在百链平台上看到的趋势 以及行业内其他的像这些和我们合作关系比较友好的这些AI创业公司那边看到 其实从去年11 12月份开始到现在到今年的5月份 这一段时间的大量的公司的API需求的一个增长 几乎大部分都是由AI coding的能力的提升来带来的 而AI coding它不是简单的只是说我们替代了软件工程师的工作 由于AI模型的能力的提升以及AI coding和整个的aging的运行环境的这些结合 也就是所谓的Honey's工程啊这些数据领域的这些结合 我们看到由AI coding驱动的复杂任务aging 已经可以几乎可以在各个数字化的工作任务当中去完成任务 那所以这无论是美国还是中国 这一波AI需求的带来的带来的增速 主要是靠的是AI coding能力的提升 因为AI coding能力的提升 结合电脑或者数字化的这些工具场景 理论上是可以解决几乎所有数字化工作的

The other part of your question was about when we can see the similar kind of growth in China as is being witnessed in the U.S. around AI coding. And I would say in terms of what we're seeing, based on the trends that we ourselves see on Bailian, as well as the experience of some of these AI startups in China who work closely with us, I would say China is already there. Most of the growth that we're seeing in utilization from, say, November or December of last year through to May of this year has been driven by capability upgrades in terms of coding. And these models are not just able to replace software engineers. basically they're able to solve a wide array of very complex tasks beyond just coding per se in any kind of digitalized productivity scenario. So, you know, we've seen AI coding capabilities improve significantly in both the U.S. and in China. and these capabilities are capable of supporting much more than just coding per se. It can address a whole wide range of very complex tasks insofar as those tasks can be digitalized.

So looking ahead to the next two to three years, we see this is a very, very important. 现在的AI的模型的能力越来越强,当它可以帮助这些用户真正的完成工作当中的复杂的任务的时候,我们看到无论是美国的企业还是中国的企业,对于为它帮助它完成工作任务,帮助为这个智能能力买单,这个需求在美国和中国都是一样的,而且都会非常强。 理论上来说只要帮助他完成的这个工作任务在他的企业内这部分工作任务创造的价值大于token的成本 某种程度上这个对于API的token的需求就会是无限的 所以从这个角度上来说我们也看到底层的对于AI的这个需求的增长还是会非常长期确定性 那从我们自己看到的也可以分享一下我们自己看到的一些数据 那我们在我们的摆链平台上看到的整体的增速非常非常之快 对比去年11 12月份的数据 那我们今年5 6月份的数据我们应该会增长增长10倍以上 而且就是在最近我们其实的我们其实的AR应该已经突破了80亿

那也就是说在这个季度我们突破100亿的AR是一个非常确定性的事情 所以我们看到无论是中国企业还是美国企业 愿意为智能能力买单 愿意为通过这个智能能力 调用这个智能能力帮助他完成真实的工作任务 这个事情买单是一个非常普世性的一个选择 Your other comment, you know, we have also taken note of the lower willingness in China to pay for SaaS However, I think that is poised to change as the models become increasingly powerful and are able to truly solve for very complex tasks, very complex problems as they are providing truly valuable intelligence. You know, I think we can expect to see the same demand for that kind of service in China as in the U.S. In a certain sense, when the value provided by tokens exceeds the cost of those tokens, the demand for tokens will become infinite in a sense. So we see growth in AI demand as a long-term certainty. And I can also share some numbers with you in terms of the growth that we're seeing on our own Bylien platform from November, December last year through to May of this year. It's higher than 10 times growth. In terms of our ARR, it's already over 8 billion. And I think this quarter, it's highly certain that we can achieve ARR of over 10.

Next question, please.

So our next question comes from Ellie Zhang with Macquarie. Good evening, management.

Thank you very much for taking my question. I just wanted to stay on the topics of that global comparison. So if you look at globally, the overseas peers seem to have captured the most immediate ROIs in enterprise agentic workflows, whereas for the consumer and the monetization would remain a bit lagged. So going forward, considering that Alibaba is investing kind of in multi-fronts for infrastructure models, cloud, and Q&A, how do we evaluate the strategic priority and resources allocation between 2B and 2C initiatives? If going forward, enterprise side continues to gain more traction, would we consider gradually shift more resources away from Q&A to cloud and mass?

Thank you. 感谢管理层接受我的问题 我还是想接着刚才我们讲的这样一个国际间的对比来提问 因为展望全球呢 我们似乎海外的一些同行 他们是在企业的这些智能体工作流方面 是更好的把握到了就是眼前的这个机会 但是对他们来说消费端的这个变现还是相对之后的 那么讲到未来现在我们阿里巴巴在投资多种这个地图设施 多种模型云签问APP等等 那么我们就是阿里巴巴对于2C2B这两边的战略优先级 是怎么考虑的以及资源分配的这个优先级如何

企业端发展得更加好的话 我们会不会把更多的资源 从2C端转移到 比如说MAS这一方面

对 非常感谢您的这个问题 就是你这个问题 其实问得也很好 但是从我们自己的 从AI本质上的原理来说 其实它更是一个 计算范式的革命 那在计算这个范式革命 其实最后还是要能够帮助用户去完成任务 或者帮助用户去更好的去解决问题 那么从这个角度上来说 我觉得2B和2C本质上是一样的 但是就现在来说 我们看到无论是全球还是在中国 最能够接受客户付费意愿的 却是在这个2B的领域 因为这和它的ROI更容易计算 也就是这些企业的付费意愿更强 所以我们的大部分的推理资源 其实也是投入在这个2B的这个商业化领域 但是另外一个层面 其实AI无非最终就是要作为人的一个助理 那作为人的助理而言有工作的助理 也有个人的助理 也有学习的助理 从某种程度上来说 其实它要本质要解决的问题是一样的 就是用AI怎么样帮助人去解决任务 而这个任务可能有2B2C或者学习工作生活 那么我们相信2C的业务可能他的就现在的客户的接受程度和付费意愿来说 他需要一定的投资周期 但是我们相信2C的AI助理一样会随着技术的进展 以及客户的接受程度 或者在帮助他完成更多的任务之后 慢慢形成一个他的商业模式

那么如果未来的话

这个商业模式我们其实在海外也已经看到 是吧 包括我们在国内其实也会在 我觉得在未来一两年之内 我们也会看到在2C的AI助理方面 有很多的商业化的这个进展会产生 And it's a good question But I think you know fundamentally from the perspective of AI development It's really all about a paradigm shift in computing And it's about leveraging this new technology to help users, whoever they are, to complete tasks and solve problems. And that applies equally on the 2C side as well as on the 2B side. Now, certainly at present, we see higher willingness to pay on the 2B side because it's easier to show a business case with compelling ROI for a business. And at present, most of our infrastructure resources are therefore channeled to the 2B side. But at the end of the day, AI ultimately is an invention that's there to be a helper and assistant to humans to help humans with a whole range of things spanning their own daily life, their studies, and their work. And what AI does really is the same across all of those scenarios. It's about solving problems to be introduced, you see. So certainly China technology improves as it's better able to help them solve real business model internationally And I'm from Jaleigh Xu with Bank of America

未来几个季度,随着我们的加速,我们会不会看到和国外类似的这个margin扩张的这么一个趋势。 Thanks, Management.

I have a follow-up question also on the future growth of the cloud business. And I just like to understand what your view is on EBITDA margins in the cloud business over the next few quarters. Do you think as this business accelerates, we can expect to see similar margins as we see in your international peers? 和这个和这个和这个市场份额以绝对领先的我们希望还是说能够超过行业平均增速的速度平均增长速度来进一步快速的获取市场份额来夯使我们的绝对的市场领先地位利润率是我们的第二目标

但是由于现在有几个行业的一个非常明确的一个行业特点 也就是说第一我们觉得判断未来三到五年内 其实AI的行业的这个需求还是很难 因为有很多的物理瓶颈的限制 无论是AI数据中心的建设的周期 以及行业内芯片或者memory 各方面的生产周期物理扩厂 整个的产能的增长 都觉得我们还是三到五年内 很难能够支撑AI的需求的增长 所以这个角度上来说 阿里云历史上因为有很强大的 一个客户规模效应和资产 我们历史上的IT CAPEX的规模效应 在这个规模效应下 由于现在的市场供需的紧张 现在实际上我们新的一台服务器 某种程度上相同的服务器 在今年去部署上线的成本是在我们两年前部署上线的一倍以上 也就是这个成本增长了百分之百分之一百以上 那这样的一个新的服务器的重置成本 对于我们的云的客户老客户的需求以及新客户的需求 都有一个定价的牵引作用 这个定价牵引作用我们相信对于未来一个比较长周期内云的各项服务的一个一个资产定价会是有一个比较正向的提升的一个提升的一个效果吧 那还有第二个就是我们还看到一个就是我们阿里云的Mars业务的快速增长 UMS 业务本身就是一个收入 相当于毛利 对比我们传统的US 就是整个的IT资产类的业务 相对来说是一个高毛利的业务 同时由于我们刚才已经说到了 由于整个推理技术的优化 单卡的产能还会持续地提升 所以我们现在看到的一个情况 如果我们一个同样的服务器 交给我们的百链平台 所创造的营收和毛利水平 相当于毛利对比我们传统的US 就是整个IT资产类的业务 相对来说是一个高毛利的业务 同时由于我们刚才已经说到了 由于整个推理技术的优化 单卡的产能还会持续的提升 所以我们现在看到的一个情况 如果我们一个同样的服务器 交给我们的百链平台所创造的营收 和毛利水平是高于我们传统的云计算 简单的算力服务的 所以这个业务的收入的占比提升 我们也会相信能够提升我们未来的毛利水平 那还有一个就是我们的权战技术优势 评头哥的我们的AI芯片的一个字眼 以及评头哥AI芯片的未来的规模放量 有之于我们在整体来说提供一个 可能在中国最好的一个性价比水平的一个推理平台 I think when it comes to the deep penetration of AI technology,

across all different industries. We're still really in the early days of that long process. But our objective is clear. Our objective is to achieve growth, to drive growth, to drive growth in token consumption, and to acquire larger market share. We aim to maintain growth that is faster than the market average in order to gain larger market share and firmly cement our position. So those are the primary objectives, and margin is still secondary. The other thing to be pointed out is that for the next three or even five years to come, there are physical constraints on production, production capacity for chips, for memory, for the physical things that are needed to support all this growth in demand. And an advantage that we have at Alibaba Cloud is the scale of our customer base as well as the scale effect from all of the capex that we've put in over these years. But in this environment of market scarcity, new server this year is double. Given that higher replacement cost effect, customers and also old customers. So I think in the long term, the asset pricing effect. Secondly, we see very rapid growth in mass, as we reported to you, and inherently, as we said, mass represents a much higher level of gross margin than IaaS or traditional types of IT operations. So as demand for inference continues to grow exponentially, due to the optimization of our reasoning technology, the output capacity, the productivity of a single card will continue. An additional factor is as we continue to scale up the deployment of T-Head, the T-Head chips represent the highest value for money compute power on the cloud platform. But for several objective reasons that I've outlined, I think overall in the next two to three years, we can expect to see a significantly higher gross margin for Alibaba Cloud, and we can expect to start to see that in the next one day.

Peter, let's take the last question. Thank you.

Your last question comes from Gary Yu with Morgan Stanley.

Hi, for the opportunity. I have a question regarding capex. so what kind of level of capex investment is required in order to satisfy the demand from both mass and also the long-term cloud revenue and also management mentioned about T-HAT opportunity what is the current penetration of T-HAT being deployed on Naughty Cloud and as this penetration increases the margin uplift we should expect from our in-house chip Thank you 啊我们的资本开支方面就是为了呃实现我们这个mass以及云业务方面的收入目标我们需要维持什么样的一个资本开支的水平啊另外一个问题刚才讲到啊平头哥想知道现在呃我们呃在我们的系统中平头哥芯片的渗透率已经达到百分之多少然后随着它的进一步渗透具体可以给我们的呃毛利带来呃

多大的提升

好 谢谢您的问题 就是您的第一个问题 也比较重要 就是其实 我们在上个季度的 我们的讲稿中 也已经提到了 就是我们其实 对未来五年的 一个营收的目标 有一个非常高的 营收目标 那在那个目标情况下 其实我们对比 2022年 2022年和2023年的 也就是AI大模型 还没爆发的这个年代的就是阿里云的外部营收 那个收入目标大概是是一个增长十倍的收入目标 那初浅的计算可以可以理解为至少是当时阿里巴巴阿里云所持有的数据数据机房资产的十倍以上才能够支撑我们的长期的业务目标 所以某种程度上我们未来要建的数据中心的规模基本上是一个对比2022年是一个十倍以上的增长 这是我们所有投入的CAPEX也好 包括我们现在还有不少方式 不少用OPEX方式去获取的这些算力 来说这是一个我们的总的目标 所以我们的对比前面三年 前面所说的三年3800亿来说 我们面向未来的这个五年目标来说 其实我们相信我们为了获取这些算力中心 所获取的这个投入的投入资金 会远远超过我们原来的原来的3800亿 当然现在这个情况 因为已经比较复杂 不见得会所有的算力中心 都会由我们自建的CAPEX去获取 比如说我们会用 OPEX租赁的方式去获取 包括我们随着我们 平头哥芯片的产能的扩大 我们也可能会通过去 销售平头哥AI服务器的方式 去销售给各个算力中心 或者说数据中心的服务商 同时和这些数据中心的服务商 共同去建设

The first question is quite an important one. And actually, in our prepared remarks delivered at the last quarter's earnings call, We set out a forecast for the coming five years for revenues, and it was a very high target. But essentially, I think if you compare where things were in the year 2022 before this explosive growth in AI models and what we expect to need in 2033, I think we're talking about 10x increase. So we need 10 times the amount of data center infrastructure compared to what we had in 2022. But there are different ways to get that compute capacity. Some of it can be CAPEX. Part of it can also be OPEX. And we're actually now acquiring quite a bit of computing capacity using OPEX. You know, the situation is complex today for reasons we've discussed. But I think it's likely, given that kind of investment, that we will overshoot the original CAPEX figure that we had stated of 380. But at the same time, we can acquire some compute. As we have our own proprietary T-head chips, we can actually also sell AI servers, leveraging those chips to other computing centers, or we can co-build computing centers with others. So there are different ways that we can get to where we need to get. But the bottom line is that the demand for our structure is going to be 10x.

所以未来我们有机会用我们一个全站 从我们的GPU CPU存储到网络芯片上 去进行一个全站的一个字眼 那现在这个比例还比较低 因为有一些各方面客观的原因 比如说在中国国内的这个整个半导体的产能 产能的量还是比较少 当然这几年中国国内的半导体的产能 也在持续的扩张当中 所以我们觉得随着这个平头哥 比如说我们自研全套芯片的渗透率的提升,对我们毛利率的提升影响会是非常大,但这个大现在就是说有一些情况在互相的影响,比如说由于中国的国产的国产半导体的整个的制程相对来说,比国外的相对来说还是相对差一点,那么总体同样的芯片,无论是能耗还是说还是效率上来说都会有一些对比国外先进的芯片来说, 还会有一些差距 但是由于我们可以看到 国外主流的AI芯片的毛利率 非常之高 是吧 基本上接近在60%到80%的毛利率 那么无论是就算是国内芯片的 总体的性能或者说功耗 会有一定的提升 但是在这个60%到对比国外先进芯片 60%到80%的毛利率的中间 其实有个非常大的 Yeah, so in terms of our T-Head proprietary chips,

They can be deployed across a very large part of our AI infrastructure, and not just compute chips, but we have a full stack, including memory. But at present, the ratio is still relatively low, and that's because of constraints around production capacity in China, which has been limited. Of course, it's been growing. But as we deploy more and more of our own proprietary T-head chips, the new chips will certainly contribute very, very significantly to gross margin expansion. It's true to say that domestically produced semiconductors in China lag behind the leading overseas ones in terms of energy efficiency and production efficiency. However, if you look at globally leading AI chip vendors today, their gross margins are as high as 60% or even 80%. And so as we ramp up domestic chip production and the capabilities improve, I think there's a lot of room for our chips to be providing very high value for money as compared to that 60% to 80% gross margin that other vendors are taking.

Thank you. This brings us to the end of today's earnings call. We appreciate your time and participation, and we look forward to speaking with you soon.