Alibaba Group Holding Limited

Good day, ladies and gentlemen. Thank you for standing by and welcome to Alibaba Group's December quarter 2025 results conference call. At this time, all participants are on 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 Group. Please go ahead.

Thank you. Good day, everyone, and welcome to Alibaba Group's December quarter 2025 earnings conference call. Joining us today are Jing Cai, Chairman, Eddie Wu, Chief Executive Officer, Toby Xu, Chief Financial Officer, Jiang Fan, Chief Executive Officer of Alibaba eCommerce Business Group. I would like to remind you that this call is also being webcast on our copy website. A replay of the call will be available on our website later today. Now, I will quickly cover the safe harbor. Today's discussions may contain forward-looking statements based on current expectations and assumptions that are subject to risks and uncertainties. Actual results may differ materially. Please refer to the safe harbor statements that appear in our press release and investor presentation provided today. Please note that certain financial measures are expressed on a non-GAAP basis. Our GAAP results and reconciliations 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 now I will turn the call over to Eddie.

Thank you, and welcome to this quarter's earnings call. Over the past quarter, we maintained strong investment momentum in our two strategic priorities, AI plus cloud and consumption. Cloud intelligence group revenue growth accelerated to 36%, while our quick commerce business continued to expand in scale with ongoing improvement in unit economics. With the dawn of the AI agent era, the addressable market for AI infrastructure providers like Alibaba is set to grow exponentially. AI models and their capabilities are rapidly being embedded into mainstream work environments across all industries, with token consumption surging across sectors. Cloud and software budgets for enterprise IT services have traditionally represented only around 5% of corporate revenue. As model-driven agents begin to handle mainstream work tasks across industries, our total addressable market will expand by several multiples. From AI infrastructure to the application layer, Alibaba has built a complete full-stack AI capability set to support the exponential growth in AI demand. Faced with an industry transformation and strategic opportunity of this magnitude, Alibaba Group is itself entering a new phase of entrepreneurial reinvention and critical investment oriented toward the future. Next, let me share Alibaba's AI strategic roadmap. We have complete full-stack AI capabilities. Ships and cloud computing form the AI infrastructure layer, while the AI application layer is anchored by Alibaba Token Hub and comprises foundation models, mass, and both enterprise and consumer applications. Together, these give us end-to-end coverage across the full stack from AI infrastructure to applications. Given the enormous and sustained growth momentum of the AI market, combined with Alibaba's full-stack positioning across the AI value chain, the business goal of Alibaba's AI strategy is very clear. Over the next five years, our goal is to surpass US$100 billion in combined cloud and AI external revenue, including mass. Regarding our infrastructure, driven by sustained strong AI demand, cloud intelligence groups' revenue from external customers accelerated to 35% this quarter, with AI-related product revenue delivering triple-digit year-over-year growth for the 10th consecutive quarter. Cloud Intelligence Group's market share has grown for three consecutive quarters, rising to 36%, with our lead continuing to widen. Alibaba Cloud's cumulative external revenue through February for fiscal year 2026 officially surpassed 100 billion RMB. Over the past three months, token consumption on the Model Studio platform has grown by six times.

We expect MaaS to become Cloud Intelligence Group's largest revenue product.

T-Head's proprietary GPU chips have achieved scaled mass production. As of February 2026, T-Head had cumulatively shipped 470,000 AI chips. In real-world business deployments through Alibaba Cloud, more than 60% of T-Head chips serve external customers, and we've completed scaled adoption for external customer AI workloads. T-Head now supports the AI workloads of over 400 enterprise customers across industries, including internet financial services and autonomous driving. We're confident that T-Head's compute supply capacity will continue to expand, contributing high quality compute to our cloud infrastructure and mass platform, strengthening the overall competitiveness of our cloud services. Regarding our application layer, Centered on the core mission of creating, delivering, and applying tokens, we established the new Alibaba Token Hub Business Group, ATH. It comprises Tongyi Laboratory, the Mass Business Line, the QN Business Unit, the Wukong Business Unit, and the AI Innovation Business Unit. It is the organizational foundation for executing Alibaba's AI strategy and the hub for efficient coordination across our AI businesses. During Chinese New Year, we launched our latest generation large model, QN 3.5+, which delivered outstanding performance across comprehensive benchmarks in reasoning, coding, and agentic capabilities, QN 3.5 Plus demonstrated significant improvements in inference efficiency through foundational architectural innovation. Building on QN 3.5, we will soon release the next generation of models optimized for coding and agentic use cases. On the consumer application side, Powered by the strength of our models, QN's consumer-facing monthly active users have surpassed 300 million. During Chinese New Year, we deepened integration across Alibaba's ecosystem, connecting QN app with Taobao Instant Commerce, Alipay, Fliggy, Damai, and AMAP, giving it unique capabilities relevant to everyday life and becoming China's first all-in-one personal AI assistant for life, work, and learning. We've also recently launched Wukong, our enterprise AI agent platform. Wukong is the world's first AI-native enterprise-grade agent platform, enabling AI-powered upgrades to enterprise workflows while remaining compatible with each organization's data permissions and management processes. It serves as the unified interface for Alibaba's AI capabilities in enterprise work environments, and the B2B capabilities of businesses across Alibaba's full ecosystem will be progressively integrated to support Wukong in becoming the best AI work system. On Alibaba's other strategic priority, the consumption segment, we continued to advance our strategic initiatives. This quarter, our quick commerce business further expanded in scale with continued share growth, high customer retention, and sequential improvement in both unit economics and average order value. At the same time, quick commerce and e-commerce demonstrated clear synergies, driving Taobao app monthly active consumers to double-digit year-over-year growth. That concludes my remarks. I'll now hand over to Toby to share the financial update. Thank you.

Thank you, Eddie. Our strategic priorities are clear. We remain focused on AI plus cloud and consumption businesses. We are seeing great momentum with gains in technology, customer adoption, market share, and user engagement. On AI plus cloud, we have the full stack AI capabilities with all three core elements, model, cloud infrastructure, and chips, and leadership in each with Queen, Alibaba Cloud, and T-Hat. We also operate the most comprehensive consumer ecosystem in China that can be monetized through AI. The launch of Queen APP was a major milestone, and it can bring our consumer applications together. On consumption, our quick commerce business continued to gain GMB market share in December quarter, while unit economics and AOV also continued to improve. Now let's look at the financial results. On a consolidated basis, total revenue was RMB 284.8 billion. Excluding revenue from Sunna and InTime, revenue on a like-for-like basis would have grown by 9%. Total adjusted EBITDA decreased by 57%, primarily due to our strategic investments in technology-related innovation initiatives and consumption front, including quick commerce business, partly offset by the improved operating results in cloud business and enhanced operating efficiencies across various businesses. Our gap net income was RMB 15.6 billion, a decrease of 66%. Operating cash flow was an inflow of RMB 36 billion. Free cash flow was RMB 11.3 billion, a decrease of RMB 27.7 billion from the same quarter last year. We are reinvesting our cash flow to be a leader in AI and quick commerce. As of December 31st, 2025, we held the U.S. dollar 42.5 billion in net cash, excluding debt with maturities beyond five years. Our net position stands beyond the U.S. dollar 60 billion. This balance sheet strength gives us confidence to reinvest for long-term growth. Now let's look at our consumption businesses. Revenue from China e-commerce group was RMB 159.3 billion, an increase of 6%. Customer management revenue increased by 1%. The slowdown in revenue growth was primarily due to weaker transaction activities and phase-out of the impact of software service fee implementation. The Taobao APP achieved a double-digit increase in MAC during the quarter, driven by the growing mindshare and increasing scale of our quick commerce business. Revenue from our quick commerce business increased 56% to RMB 20.8 billion. During the quarter, we executed our plan to further grow the scale of our quick commerce business improved user experience, improved UE, and increased AOV month over month during the quarter. Alibaba China e-commerce group adjusted EBITDA with RMB 34.6 billion, a decrease of 43%, primarily due to the investment in quick commerce, user experiences, and technology. Going forward, this adjusted beta will continue to fluctuate quarter over quarter due to intense competition and significant investment in user experience. Revenue from AIDC grew 4% this quarter. AIDC's adjusted beta loss narrowed significantly year over year, driven by a combination of logistic optimization and investment efficiency enhancement. The UE of the AliExpress's choice business also improved 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 grew 35% up from 29% last quarter. AI-related products continue to lead this momentum, we delivered our 10th consecutive quarter of triple-digit growth in AI revenue. Its share of external cloud revenue continued to increase. This is a clear reflection of the scale and acceleration in our AI business. The adjusted EBITDA margin remained relatively stable at 9%. We will continue to invest in customer growth and technology innovation to increase adoption of AI cloud infrastructure and strengthen our market leadership. All other segment revenue decreased by 25% to RMB 67.3 billion, mainly due to disposal of Sunnah and in-time businesses, as well as a decrease in revenue from China, partly offset by the increase in revenue from Fresh Apple and Alibaba Health. All others adjusted EBITDA with a loss of $9.8 billion, primarily due to the increased investment in technology businesses, including Queen Models and the consumer-facing Queen, partly offset by the improved results of China Hu Jing DME and other businesses. Queen model has become one of the most widely adopted open source model families globally, surpassing 1 billion cumulative downloads on Hugging Face by the end of this January. And the consumer-facing Queen has surpassed 300 million MAU cross-platforms, which reinforces user engagement and expands long-term monetization potential. We have been increasing investments on these technology fronts including the Spring Festival campaign. Building on the strong momentum and results achieved, as Eddie mentioned earlier, we will continue to invest substantially in Queen models and Queen APP. Our unallocated adjusted EBITDA was a loss of RMB 2.7 billion compared to a loss of RMB 0.2 billion in the same quarter last year, which reflected costs associated with tailoring retention incentive from the one-off replacement awards plan of Ulema. 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. If you are unable to hear the Chinese translation, bilingual transcripts of this call will be available on our website within one week after the end of the meeting. Operator please go ahead with the first question. Thank you.

Thank you. If you would like to ask a question, you can press star then 1 on your phone and wait for your name to be announced. If you are on a speakerphone, please pick up your handset to ask your question and to give more people the opportunity to ask questions, please keep yourself to no more than one question at a time. First question today comes from Robin Zhu at Bernstein. Please go ahead.

Thank you. Thanks, management, for taking my question. Could you give us some specific examples of how TokenHub will change how the different cloud and AI businesses work together going forward from an organizational standpoint? And strategically, what changes or goals are you hoping to achieve with this new structure that improves on the previous arrangement going forward? And then If management could share a hierarchy of priorities in cloud and AI, is it market share and revenue growth, such as the target you just announced, versus having the best first-party model capabilities versus consumer-side traction with customers using agentic AI or anything else?

Thank you. Thank you, Director.

That's my question. First of all, can I ask you to introduce what kind of new changes will occur from the point of view of organization design after TokenHub? In addition, from a strategic point of view, after having such a new structure, What kind of changes will it bring or what kind of new goals will it help to achieve? Is it better than the previous arrangement? In addition, I would like to ask the management to tell us about your priority in terms of cloud and AI. For example, the first priority is market share and income growth. As you just announced, Thank you for your question.

I think the purpose of our establishment of ATH is related to the current era or the background of technological change. Because from the second half of 2025 to the first two or three months of 2026, we have already seen that AI has entered the era of AI, which is the era of agent drive. The era of agent drive and the early stage of AI, the biggest difference is that . . . So from this perspective, we think that if we look at the five levels of AI, the first is the application layer, the second is the model layer. Great, thank you very much for your question.

The goal and purpose of the establishment of the ATH business group is very much connected to the era that we're now in as of the end of 2025 and going into the first few months of 2026 in terms of the development of AI. We're now in the agent-driven era of AI development. And this is different from the earlier period of AI development. In the agentic AI era, we need to achieve a very close integration of model with application. In the earlier AI era, a lot of model training data was static data, but in the agentic era, we need to enhance the integration between models and applications and achieve tight integration and a lot of the data is now coming from the customer side. So if you look at the different layers involved in AI deployment from application, model, the AI infrastructure through to chips, I think what's most different and most important about the agentic AI era is the need to achieve this tight integration between application, and model. That's the critical priority.

Okay, let me explain the connection between our ATH and our business. From the perspective of the future development trend in the industry, we can see very clearly that the future of AI Agents in the application layer will be very innovative. Foreign Foreign AI applications will have a lot of various industry applications, or vertical applications, or AI agent applications of all kinds of different scenarios, to support the development of various industries. So in addition to this application layer, we also need to have a powerful mass business. In fact, the mass business is a channel between the model and the application layer. In addition to supporting our own internal applications, we also need to use a powerful MaaS business to support more rich and colorful AI industry applications. We think this market value and market space will be very large. So from this perspective, we think AI applications are the biggest Next, let me address the

interconnection and synergies among the different businesses in relation to ATH. If you look at the trends of where this industry is going, and we think we see these trends very clearly, the AI agents will be tightly integrated together with the application layers. and there will be a multitude of highly diverse applications. In the to-consumer or 2C space, we're strongly developing the Q1 app as a personal assistant for individuals, and in the 2B space, we're positioning Wukong as a 2B assistant. In the AI application layer, there will be a multitude of different industry and vertically specialized applications to serve different industry use cases. And all of this needs to be supported by a very robust model as a service, MAS layer. So MAS supports, of course, our own internal applications, as well as a multitude of external and industry-specific use cases that leverage AI. So in this context, we see massive value that we can provide and a huge total addressable market or TAM. So going forward, we see the AI application layer as the main channel through which tokens will be distributed and the stronger the model capabilities that you can offer at the mass layer, the more attractive and compelling all of these different offerings will be to customers. That is the business logic that we have laid out within this new business unit.

So if we look at the model and application layer, our priority is undoubtedly to build the most intelligent model. But I want to emphasize that only the strongest model can drive the expansion of application scenarios in all industries. Only the strongest model can attract applications from all industries to use our MaaS business. But I want to emphasize that in fact, how to build the strongest model requires our own 2C, 2B applications, and applications that connect to all industries through MaaS, to use our model to gradually open up the data wheel and close down the business. Through this method, through more scenarios, more data, and more users to relay our model capabilities, we can gradually form a data wheel to continuously improve the ability of the model. This is one of the reasons why we have established ATH as a business group. So, I would like to To sum up, I think our priority is to improve our model capabilities. However, improving our model capabilities requires us to work together with ATH on both the model and application parts, and on the mass part, in order to improve our model capabilities for the long term.

From the perspective of both the model and the application layers, our top priority absolutely is to develop the most intelligent models. I really need to emphasize that only when you have the most powerful models can you truly drive the deployment of AI applications across all kinds of different industries. Only with the strongest models can you attract applications from across diverse industries to adopt our mass offering. However, in order to build the most robust models, you need to have very close collaboration with various industries and with our own 2C and 2B applications to connect with our mass to applications across all kinds of different industries and use cases. So we need to get more users to leverage and make use of our models in order to gradually be able to leverage the data flywheel effect. Only in that way can we continuously enhance the capabilities of our models. So that's one of the reasons why we have established the ATH business unit at this time. So to summarize, I would say our top priority is definitely to enhance model capabilities. However, to enhance model capabilities requires concerted efforts across the entire model pipeline as well as on the application and infrastructure side in order to achieve sustained improvements over the long term.

Next question, please.

Thank you. Your next question comes from Joyce Chu at Bank of America. Please go ahead.

Good evening, management. Congrats on the solid progress you've made in cloud and AI, and thanks for taking my question. My question is, we see CMR growth slowing notably in the December quarter, given the macro pressures. We have seen China's online retail sales only up 2% year-over-year in the fourth quarter, 25. But more recently, MBS data point to a reacceleration in January and February. Could you share your latest view on the CMR trends heading into the March quarter and and whether you have started to see any improvement in consumer sentiment.

Thanks.

Good evening, management. I also congratulate you on the solid progress you have made in terms of cloud and AI. My question is about CMR. We have seen a slowdown, especially in the December quarter, due to China's overall online retail uh uh Introduce your latest views on the CMR trend in March. And then the consumer side, did you start to see any improvements? Thank you.

Let me answer this question. As you mentioned, the December quarter caused us to grow in December due to the hot weather and the late spring festival this year. At the same time, due to the long initial period, we have increased our investment in consumer rights. So, we have slowed down our CMA and e-commerce profits in the December quarter. We have also seen that since the first quarter of this year, we have seen that consumption has resumed, which is very obvious. At the same time, under the joint pull of our real-time sales strategy, our food e-commerce transactions and CMA increases have significantly recovered, and our e-commerce profits have significantly improved.

Thank you for your question.

Indeed, in the December quarter, weak macro consumption, a warm winter, the later timing of the Chinese New Year challenged the growth for the December quarter. And due to the extended promotional season, our investments in consumer benefits increased compared to previous years. So as a result, the CMR and EBITDA trend softened. Going into the March quarter with the improving consumer sentiment that we've observed and momentum from our quick commerce strategy, our physical goods GMV and CMR trend have significantly recovered from the December quarter and EBITDA is expected to improve accordingly.

All right, let's move on to the next question.

Thank you. Your next question comes from Gary Yu at Morgan Stanley. Please go ahead.

Hi. Thank you, management, for the opportunity. My question is related to quick commerce. I understand that, you know, in the past couple of months, we have achieved certain milestones in terms of GDP market share and also UE improvement. How should we look at the priority going forward? Are we aiming for, you know, you know, share or, you know, hoping to take this opportunity to improve unit economics, reduce loss, and how should we look at the synergy between quick commerce and traditional e-commerce, and how should we see this synergy to translate into CML better growth going forward? Thank you.

感谢管理总接受我的提问。 我的问题是关于及时零售这一方面。 我知道过去几个月公司已经在市场份额还有UV改善这些方面取得了一些阶段性的成果。 那么面对未来我们的优先级会是什么呢? Is it market share or do you want to take advantage of this opportunity to further optimize UE and reduce losses? In addition, how should we view the collaboration between real-time sales and traditional e-commerce? And how will these collaborations turn into a faster increase in CMR in the future? Thank you.

Let me answer this question. Yes, we continue to see that while the market style is improving, we have improved the efficiency of logistics, the improvement of commercialization, and the optimization of our order structure. They have further driven the obvious improvement of the advantage. Our advantage is still in the continuous improvement. We think that in the next few seasons, our advantage will improve. Let me see.

Certainly. While growing our market share, we have continued to significantly improve UE, driven by improvement in fulfillment logistics efficiency, by improvement in monetization, as well as by order mix optimization. Driven by those factors, we expect to further optimize UE in the coming quarters.

We also saw that in the past year, flash purchases have significantly increased the overall pull of the platform. The number of e-commerce buyers increased by 150 million per year. The number of e-commerce buyers increased by 100 million per year. The number of e-commerce buyers increased by 100 million per year. The number of e-commerce buyers increased by 100 million per year. The number of e-commerce buyers increased by 100 million per year. The number of e-commerce buyers increased by 100 million per year. The number of e-commerce buyers increased by 100 million per year. The number of e-commerce buyers increased by 100 million per year. The number of e-commerce buyers increased by 100 million per year.

In terms of the positive impact that QuickCommerce is bringing to our conventional e-commerce business and to our entire ecosystem, we saw a very significant increase in AACs on the platform in the past year. Our AAC number increased 150 million in 2025, including 100 million conventional e-commerce physical goods AACs, which is more than the previous three years combined. Now, new consumers' RPU and purchase frequency are lower than that of existing users, so we aim to continually increase their RPU and purchase frequency which will serve as a new growth engine for our platform in the coming years. QuickCommerce is clearly driving sales in various categories, such as food and fresh produce and healthcare, and is contributing to Freshipo and Tmall supermarkets' accelerated growth.

Regarding the future development expectations, first of all, we maintain the goal of over 10,000,000,000,000,000,000

In terms of the outlook, we maintain our target of achieving over 1 trillion RMB in quick commerce GMV by FY28. We expect to generate positive cash flow when the GMV target is achieved and we expect the quick commerce business to be profitable in FY29.

Uh, uh, uh, uh,

QuickCommerce has become a cornerstone of our e-commerce business, playing a strategically vital role in the AI era by driving customer acquisition, enhancing user engagement, fulfilling diverse consumer demands, increasing transactions and improving monetization and supporting logistics infrastructure. We are committed to investing in QuickCommerce in the next two years towards achieving the 1 trillion RMB GMV target as a market leader.

Operator, let's move on to the next question.

Thank you. Your next question comes from Alicia Yap at Citigroup. Please go ahead.

Hi. Good evening, management. Thanks for taking my question. I have a question regarding your cheap business, THET, Pintoker. So there have been reports that Alibaba plans to spin off the THET unit as a separate listing. Can management provide any information of this? And if so, what is the expected timeframe for this to occur? And in the meantime, can you share more operating metrics? So in addition to the 470,000 chips that you mentioned you shipped to external customers, how we reconcile that number, the shipments, to the revenue size, and also what is the expected growth rate for your chip business in the coming year. And I think you mentioned currently it's 60% of this is from external customers. So maybe can you also share with us, are these chips for external customers mainly used for inferencing? And then for internal, is it used for model training and also inferencing? And then lastly, how do the Pinto Girls chips or TH chips compare to other domestic chips? If management can share some detail, it would be great. Thank you.

感谢管理层接受我的提问。 我想问一些关于你们芯片业务, 就是平头哥方面的问题。 那么最近有报告, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, Okay.

Thank you very much for your question. Alibaba is an important part of Alibaba's full-time AI layout. I would like to take this opportunity to introduce it in detail. In the current domestic AI chip ecosystem, we think that Alibaba's technical capabilities and product capabilities are at the forefront. Our products cover a complete AI workflow from training to reasoning. Our Pinduog's AI chips have been used on a large scale in the training scene, in the training scene, and in the training scene. At the same time, we also have more than 60% of Pinduog's chips in our public cloud and mixed cloud products that are being used by external commercialized customers. And external commercialized customers cover multiple industries such as the Internet, finance, smart driving, and smart manufacturing. . . . . In addition, I think that for Alibaba, Pintuco is important because we hope that in terms of domestic chips, compared to foreign chips, in terms of manufacturing and manufacturing, and in terms of chip performance, if foreign chips are lagging behind, we hope that with Alibaba's cloud-based facilities, as well as the model of communication and communication, to have a deeper co-design to improve the cost-effectiveness. So this is a point that is different from those of other chip companies. So what we hope to do is to create an AI-capable platform . . . . In addition to reducing costs, I think there is a special situation in the AI industry in China. I think there is a significant value for us, which is the supply and security of the entire AI algorithm. Because now I think in the next three to five years, the global AI algorithm will be in a very tight three to five years. Especially in the Chinese market, it will be even tighter. OK, thank you very much for this question.

And I'd like to take the opportunity to expand on this a bit because T-Head is a very important component of Alibaba's company-wide AI strategy. So, in the context of China's domestic AI chip ecosystem, we firmly believe that T-Head is ranked in the top tier of the domestic AI chip ecosystem in terms of the technology capabilities and product capabilities. Our products cover the entire AI workflow from model training and fine-tuning through to inference. And our T-Head AI chips are already in extensive large-scale use via Alibaba Cloud, both for training workloads and for bai lian inferencing use cases. At the same time, over 60% of TH chips are being used by external commercial customers across Alibaba Cloud's public and hybrid cloud offerings. The external commercial clients span multiple industries, including internet finance, autonomous driving, and intelligent manufacturing. And these external commercial customers are utilizing T-Head chips in both their training and inferencing workloads. Moreover, on the T-Head software stack we have excellent compatibility with the Linux ecosystem so customers can migrate their systems easily without spending a lot of time on the migration another another point I would make is that in my view he has significance to Alibaba and lies not only in our aspiration to close the gap between domestically produced chips and foreign counterparts, foreign produced chips in terms of manufacturing processes and overall performance across various dimensions. But given that our chips still lag behind foreign counterparts in performance in various respects, We aspire to engage in more profound co-design with Alibaba's cloud infrastructure and the QN model to provide improved cost effectiveness. So this is one key differentiator in how we approach chip design at T-Head that sets us apart from other chip companies. Our primary goal is to create AI capabilities that offer superior value for money This will make it a key product for the Bailian platform, allowing us to reduce inference costs going forward. Beyond generally improving our AI efficiency and reducing costs, there's another factor at play, namely the unique circumstances currently facing the AI industry in China. In that context, one significant benefit for us is the guaranteed supply. of AI computing power because I believe that over the next three to five years, global AI computing power will be in extremely short supply, especially in the Chinese market. As the only cloud computing company in the Chinese market with proprietary chip development capabilities, T-Head is of paramount importance, therefore, to the Alibaba group. Increasing the supply of AI computing power will help our cloud and AI businesses, including our mass business, to achieve stronger growth momentum.

In the past two years, we have successfully commercialized and launched more than 470,000 chips. Our annual revenue has also reached the level of one-hundredth of a million. In this year, two six, two seven years, this year, two six, and next year, two seven years, we expect that the scale of high-quality AI chips that Pinduog can produce will continue to expand and provide sufficient computing security for our entire group's AI business. can also provide a very strong growth power for the entire AI business. At the same time, it will also be very helpful for the future improvement of profit level. In general, the value of Pinduoguo is not only cost-efficient for Alibaba, but also supply and insurance. In an era of lack of computing power, I think it is very important for Alibaba's AI strategy. So Pinduoguo

At T-Head, over the past two years, we've successfully commercialized and launched chips with total volume exceeding 470,000 units with annual revenue reaching the 10 billion yuan level. Looking ahead to 2026, this year through 2027, next year, we expect T-Head's production capacity for high-quality AI chips to continue to expand. This will provide robust computing power support for our group's AI business and serve as a powerful growth driver for our overall AI initiatives. We also believe that future improvements in profitability will be achieved further enhancing profit levels, which will also be very beneficial. Overall, T-Head's value to Alibaba goes beyond cost optimization. It primarily serves to ensure supply chain resilience. And in an era of scarce computing power, I see this as crucial to Alibaba's AI strategy. So it is possible, and we don't rule out the idea of T-Head considering an IPO in the future, although we currently do not have any definitive timeline.

Next question, please.

Thank you. The next question comes from Yuan Liao at Citix. Please go ahead.

管理层网上好,感谢接受我的提问。 我的问题是关于你们提到的阿里巴巴集团AI战略的商业目标, 未来五年的收入要突破一千亿美元, 那管理层能否给出这一目标的更多的细节? 那比如说未来五年是到2031年嘛, Thank you for taking my question, management.

My question is about the business objectives for your AI strategy that you just mentioned. Revenue for the next five years is expected to exceed $100 billion. Could you provide more details on this target? For example, if the next five years is through to 2031, what kind of CAGR would that correspond to in this five-year period? And could you also break out what will be Thank you very much for your question.

Thank you for your question.

So, yes. We certainly believe that within five years, revenues from our AI and cloud-related business will exceed $100 billion. We think that that is very clear. If you look at the market growth that we're seeing today, the strength of our product portfolio and the roadmap to get there.

I think the biggest growth is due to the breakthrough of AI big model capabilities. In the past few months of 2026, we have seen some very obvious trends. We have seen that big models have started to have the ability to complete 2B complex workflow. The more and more companies are using large-scale models to produce agents to help them complete end-to-end work tasks. In fact, the entire IT budget market facing the original AI and cloud has actually undergone a fundamental change. That is, when a large number of companies use tokens to consume, they do not treat tokens as I think that the major driver

underlying all this really is continued breakthroughs in the capabilities of large AI models. And we've certainly seen a very clear trend over the past couple of months, the first two months of 2026, whereby large models have now gained the capability to execute complex B2B workflows. More and more enterprises are deploying agents powered by large models to handle end-to-end business tasks. And that marks a fundamental transformation in the way that the market looks at IT budgets, IT budgets traditionally allocated to AI and cloud services. The shift really is that Many enterprises now, when consuming tokens, don't treat token consumption as part of their IT budget any longer. Instead, they see tokens as part of their overall operational or R&D costs. Tokens are a key component of their production inputs, not just a part of their IT budget. So this is the most fundamental long-term factor that we see driving future AI growth.

There are three major growth factors. The first factor is the growth engine at the core of the large-model mass business. The application of the mass business includes our own applications, our own application clients, and various types of AI applications in the industry. um . . . . . There are some application scenarios that will choose to use the public MaaS API service. There are many application scenarios that will also rely on its private deployment in the enterprise. These application scenarios are also a very big positive scenario for our Aliyun AI infrastructure.

I believe that the largest drivers of growth will come from three areas. First is the mass-driven business, which really is the core growth engine, and the growth of our mass business will be supported by a variety of different use cases, including our own applications as well as a diverse array of of AI application scenarios from across our customer base and across various different industries, including AI application software. And we believe that the growth driven by mass initiatives will be a key driver of future revenue for both AI and cloud services. But secondly, for AI and cloud computing, there's another very important growth opportunity. Of course, we believe that public mass will be a substantial market in the future, but In a considerable number of larger, medium and large-sized enterprises, there'll also be a demand for enterprise-level internal inference and training, a new marketplace. And that market will continue to exist in the long term. It's not one that will disappear simply because each enterprise makes decisions based on its own business model and the security requirements of its specific use case or the particularity of an application scenario. So, for some application scenarios, enterprises will opt to use public mass API services, while many others will be based on privately deployed solutions within the enterprise. So, those kinds of application scenarios represent a large incremental growth opportunity for Alibaba Cloud's AI infrastructure.

There's another one that I think has been ignored by many of us, . . . . . . . . . Thank you. Third, there's another important driver, important opportunity that I think tends to get ignored a lot of the time.

And I'm talking about CPU centric cloud computing, the traditional cloud computing, which still has significant room for expansion in this AI-enabled era. So traditional cloud computing is designed for IT engineers, which in China may number a few million, say perhaps no more than 10 million potentially traditional IT engineers. and those have been the traditional cloud computing customers. However, in the future, there could be billions of agents that are created by large AI models, and their operating environments, the operating environments of these agents will also require substantial support from traditional CPU-centric cloud computing. They need these traditional CPUs as well as databases storage and large amounts of memory to support their long-term problem solving and sustained operations. So the challenge lies in transforming the traditional cloud computing market, shifting from a cloud platform designed for human users, those IT engineers, to one that's optimized for agent-based invocation. So I believe there's tremendous room for growth there and So a key challenge for us this year is transforming traditional cloud computing into a platform that is better suited for agentic use. And that's a key focus of Alibaba Cloud's upgrade.

I think as the company's revenue scale continues to grow, at the same time, the AI business is moving from selling resources . . . . the profit rate of the cloud should be visible. It should have a continuous improvement process. But this continuous improvement process is not a linear process. It is possible that with a sudden change in the scale effect or a huge improvement in the scale of our flat-head chips, I think it has something to do with the scale effect of the entire product. So it will not be a linear event. It's the same as what you said about the growth rate from 2026 to 2031. I think the average growth rate, if you use a computer to calculate it, is still very clear. But we say that in the development investment and the growth of the market, it will not be linear. Our investment today may take a year or two to produce a bigger growth. But the overall goal of five years, we think it's still very...

As the revenue from this business continues to grow, our AI business will undergo transformation and upgrading, shifting from selling resources to selling intelligence, selling intelligent capabilities. And I think that represents a massive upgrade to the business model. At the same time, by integrating our proprietary THET chips, we are achieving and will achieve cost reduction and efficiency gains. We believe that as our AI and cloud business continues to grow in revenue scale, Cloud profitability should become increasingly visible and we see it is on a steady path of improvement. However, the process of continued improvement is not a linear one. It's possible that there could be a scale effect breakthrough, the achievement of an economy of scale, or the scaling up of our TH chips, and there could be a massive leap forward. But I think that's a function of the product as a whole and those kinds of economies of scale. But it will not unfold in a linear fashion. So you asked about the CAGR, the compound annual growth rate from 2026 through to I think you can plug that into your calculator and figure out what it would be, assuming it were to be linear, but I don't think that it will be linear. Our R&D investment and growth in the market will not be linear, and some of the investments we're making today may not yield significant growth until one or even two years from now. However, regarding that overall five-year goal, we are highly confident in our ability to achieve it.

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

Thank you. The last question comes from Alex Yao at JPMorgan. Please go ahead.

Thank you for giving me the opportunity to ask this question. I want to adjust the topic to look at the development stage in the direction of e-commerce. 我之前也有曾经说过 电商会进入到一个三年的投资周期 那现在这个三年投资周期 是不是因为我们看到 闪购外卖的机会 包括现在Agented Commerce的机会 会一些调整 如果不调整的话 那是不是我们已经差不多 进入到这个三年周期的中期 接下来就是逐渐的财务改善 和到三年底进入到相对稳定的一个 Thank you.

I'd like to shift the topic a little bit and ask a question about e-commerce. You'd previously said that we were in a three-year investment cycle for e-commerce. I'm wondering, if that is now being adjusted or being driven by the new opportunities that have arisen in instant commerce and in agentic commerce, or if we're still thinking of it in terms of the original three-year plan, which would put us now in the middle really of that three-year period where I guess we would start to be reaping the returns on a stable basis from those investments. So if you could speak to us about the overall direction of e-commerce in the context of that three-year investment cycle that you told us before and also share with us how you're thinking about being positioned and your strategies on this e-commerce track. Thank you.

Let me answer this question. As I mentioned earlier, First of all, we have a big investment in the market this year. This is an opportunity for us to be sure. As I said, in the next two years, we will continue to invest in the market with the goal of breaking half a billion in the market. We also believe that in two years, the investment in the market will bring positive economic profits to the entire e-commerce segment.

Thank you.

So as I just mentioned, we are making a very significant investment in the instant retail business this year, the quick commerce business this year. And at this point in time, we're seeing a highly definitive opportunity in this space. So again, as I just mentioned, we will continue to invest heavily over the next two years in order to achieve our goal of surpassing 1 trillion RMB in a quick commerce sales. We also believe that in two years time, our investments in a quick commerce will generate positive economic returns for our e-commerce business as a whole. Thank you.

Then talk about this AI, because everyone actually talked a lot about some of the effects of AI. Uh, uh, But I'd like to add to that by bringing in the dimension of AI, because Eddie's talked a lot about AI.

I believe that AI will also have a very, very significant impact on e-commerce. However, three years is too long a time to talk about when it comes to AI because AI today is evolving at a pace that's measured in weeks or in months. But that's precisely why we're making significant investments on the AI front, and we are leveraging AI to roll out new experiences for consumers and for merchants, as well as upgrading merchants business models with AI.

We believe that AI will

allow us to make huge upgrades in e-commerce across different parts of the e-commerce business. It's beneficial for our B2B business, where we're seeing tremendous opportunities for its deployment, and we will actively seize on all of these new opportunities.

Okay, that wraps up the Q&A session of today's earnings call. Thank you very much for joining us today and we look forward to speaking with you soon.

Thank you. That concludes the call for today. Thank you for participating. You may now disconnect your lines.