Home / Transcripts / Alibaba Group Holding Limited (BABA) · August 20, 2026

Alibaba Group Holding Limited (BABA) Earnings Call Transcript

August 20, 2026

New York Stock Exchange US Consumer Discretionary Broadline Retail earnings 96 min

Earnings Call Speaker Segments

Lydia Lu executive
#1

Thanks so much. Hello, everyone, and thanks for joining us on the Alibaba Group June Quarter 2026 Earnings Conference Call. On the line with us today are Chairman Joe Tsai; Eddie Wu, our Chief Executive Officer; Toby Xu, our Chief Financial Officer; and Jiang Fan, Chief Executive Officer of Alibaba E-commerce Business Group. Before we begin, please note that remarks made during this call may include forward-looking statements that reflect management's present expectations and involve risks and uncertainties. We will also refer to non-GAAP financial measures. You can find reconciliations of GAAP to non-GAAP figures in today's earnings press release and investor presentation. Except where we say otherwise, our comments will address year-over-year comparisons. A replay of this call will be posted on our website later today. With that, I'll hand things over to Eddie.

Yongming Wu executive
#2

Hello and welcome to the earnings call for Alibaba Group, covering our first quarter of fiscal year 2027. Our strategic wagers on AI paid off well during the past quarter, pushing total group revenue to a 9% gain year over year. In addition, AI commercialization has accelerated across the board. Alibaba Cloud's external revenue rose 45%, and EBITDA surged 133% from the same period last year, reflecting how we continue to honor our commitment to faster growth. For a 12th consecutive quarter, revenue from AR rated products has now grown by triple digits, and our annual revenue run rate has surpassed RMB 49.5 billion, which is about USD 7.3 billion. That is the main force propelling Alibaba Cloud's quicker expansion. I'll walk you through four main topics now: how we're commercializing AI and cloud, our full-stack AI strengths, the ecosystem around AI applications, and the consumption side of the business. First off, monetization of AI and cloud picked up pace in each of our segments, and we expect that momentum to remain strong going forward. During this quarter, revenue from external customers at Alibaba Cloud climbed 45% — a pace we haven't seen in 22 quarters — and adjusted EBITDA margin reached 11.6%. Notably, growth of 45% came from a wide range of sources, including compute, storage, model as a service, mass, and AI applications; we made a conscious choice to step away from business with thin margins, and that continues to improve how solid our growth is. In this quarter, annual revenue run rate from AI-related products topped RMB 49.5 billion, and its portion of Alibaba Cloud's external revenue climbed to 35%. AI-related products deliver gross margins well above the average of the cloud portfolio. Our recurring AI-related product revenue covers several layers: AI compute, mass and AI applications. Because AI revenue sources and monetization models are layered in this way, rising customer demand at any layer turns straight into commercial opportunity for us. This structural edge will support continued rapid growth in recurring AI-related product revenue ahead. The boom in AI agents is directly driving up the need for tokens and GPU compute, and it's also giving a strong boost to demand for our more established cloud products — CPU compute, storage, databases, and networking. Alibaba Cloud is undergoing a full-scale transformation into an Agentic Cloud. Based on the most recent figures, ARR from our model and application services, mass included, has now passed RMB 16 billion. Given what we're seeing in market feedback and our contract pipelines, compute demand is going to continue exceeding supply. As we continue increasing supply, growth in our AI and cloud revenue will accelerate even more in the coming quarters, along with ongoing improvement in profitability. Additionally, our comprehensive in-house AI stack continues to gain momentum. We're seeing scaled commercialization of our own chips, faster model iteration, and a flourishing open source community. During this quarter, tighter integration between our proprietary T-Head chips and our foundation models drove further gains in AI commercialization efficiency. T-Head has assembled a complete in-house silicon lineup spanning GPU, CPU, and networking chips. By early August, Zhenwu chips had been used by over 650 customers on Alibaba Cloud. The super node instance powered by T-Head's next-generation Zhenwu M890 AI processor has just launched commercially at scale on Alibaba Cloud; we expect supply to continue ramping through the second half of the year to meet strong customer demand. Inference workloads for foundation models exceeding 2 trillion parameters can be handled efficiently by Alibaba Cloud's Zhenwu M890 super node, and both KBK3 and Q1 3.8 MAX already rely on it to deliver mass services to outside customers. Turning to the data center layer, Alibaba Cloud has reduced hyperscale AI data center delivery time to 100 days, a pace that leads the world and will markedly accelerate how quickly we build out compute infrastructure globally. On the model front, our release cadence has picked up over recent months, with major iterations spanning our large language, image, audio, video and music models, every one of them ranked among the world's top tier. During the past week, we released the modeled weights for Q1 3.8 MAX, a model containing 2.4 trillion parameters, as well as the Q13.827B model series. Downloads of the Q1 model series have surpassed 3 billion globally at this point, and more than 300,000 derivative models have been created from it. It's our belief that a robust open source model ecosystem drives greater demand for our cloud computing offerings, creating a virtuous cycle. Third, our AI native applications span both enterprise and consumer use cases, which is fueling fast growth in token consumption. On the enterprise front, we introduced Q1 work, a fresh AI productivity offering built for enterprise workforce scenarios, providing agent capabilities at scale; we expect productivity agents to emerge as an additional engine of ARR growth. When it comes to consumers, the Q1 app continued to add users at a steady pace and is expanding what it offers in terms of value-added services. Because Alibaba Token Hub and Alibaba Cloud worked closely together, we run a commercial flywheel that is highly efficient, covering compute, models, tokens, applications and monetization. Fourth, our e-commerce operations remained solid during this quarter; in quick commerce, we continued to reduce losses by a substantial margin while business scale rose 45%, and unit economics got better compared with the prior quarter. Now that we have passed the AI commercialization inflection point in this quarter, growth is picking up speed and margins are widening during this quarter. The ability of our AI businesses to fund themselves and keep going is getting stronger, which makes us more confident about continuing to invest. Turning to what lies ahead, artificial intelligence now stands as Alibaba's surest source of expansion. We'll remain disciplined in our strategy, pursuing durable growth by leveraging our complete AI technology stack. At this point, I'll pass things to Toby, who will take you through the financials. Thanks.

Toby Xu executive
#3

Thanks, Eddie. This quarter's strong performance came from our strategic focus on AI along with cloud and consumption businesses, all supported by disciplined investment. Revenue growth in our cloud segment sped up even more, reaching 45%, while its EBITDA margin climbed quarter over quarter to 12%. Revenue from AI-related products kept fueling this momentum, marking a 12th straight quarter of triple-digit growth and making up 35% of external cloud revenue. These results show customers increasingly adopting our full-stack AI offerings, which cover AI agents, models, cloud infrastructure, and preparatory chips, alongside our improved scale efficiencies and strong pricing power in a market where supply is constrained. On the consumption side, Taobao instant commerce kept enhancing its unit economics while holding onto market share. Broadly, e-commerce EBITDA stayed fairly flat compared with a year earlier. To capture synergies across our commerce platforms and bolster our full-stack AI capabilities, we carried out a strategic realignment of certain businesses within our financial reporting. Beginning this quarter, our segment reporting will show the following: one, Alibaba e-commerce group; two, AI cloud and computer services; three, AI labs and applications; and four, all others. Let's now turn to this quarter's financial results. Total revenue rose 9% from a year ago to RMB 269 billion, fueled by strong momentum in cloud business and quick commerce. Total adjusted EBITDA fell 30% to RMB 27.3 billion, largely because of investment in technology, partially offset by better operating results in our cloud business and improved operating efficiencies across various businesses. Our GAAP net income came to RMB 10.4 billion, down 75%, mainly driven by lower income from operations and reduced net gains from disposal of investments and mark-to-market changes of our equity investments. Operating cash flow for this quarter grew 11% to RMB 22.9 billion, up from RMB 20.7 billion in the same quarter last year. Free cash flow was an outflow of RMB 44.7 billion, versus an outflow of RMB 18.8 billion in the same quarter last year. That decline stemmed mainly from investment in cloud infrastructure. CapEx totaled RMB 67.7 billion this quarter, mirroring our ongoing investments in AI infrastructure to satisfy strong and growing customer demand. The sizable year-over-year rise comes from several factors, such as fluctuations in procurement cycles, growth in CPU compute capacity driven by anticipated growing customer adoption of AI agents, and higher pricing across a broad range of chip components. As of June 30, 2026, we held roughly USD 30.7 billion in net cash excluding debt with maturities beyond 5 years; our net cash position stands at approximately RMB 46.5 billion. This balance sheet strength gives us confidence to invest for robust growth. Our AI cloud investment has a clear path to attractive ROIC; our service equipment with chips typically reach breakeven within 3 years. With a 5-year useful life, we expect them to get positive free cash flow, at least in the 2 years following breakeven. During the period that closed on June 30, 2026, we bought back stock with total consideration amounting to USD 162 billion. Our commitment stays firm when it comes to boosting returns for shareholders over the long haul via disciplined deployment of capital, spanning investments aimed at expanding the AI plus cloud business, repurchases of shares, and dividend payments. As market conditions shift and strategic requirements change, we will reorder what we prioritize. Let's begin by examining our e-commerce operations. Our newly formed Alibaba e-commerce group demonstrates how we are strategically concentrating on realizing major synergies spanning both domestic and cross-border e-commerce operations. Beginning with the present quarter, Alibaba e-commerce Group's revenue will be reported in the following manner: number one, China e-commerce. Number two, China quick comments. Number three, international e-commerce. And number four, global wholesale. Alibaba e-commerce group posted revenue of RMB 205.9 billion, which represents a 4% rise. Customer management revenue fell 7%. If you set aside the contra revenue effect stemming from the new business development program, customer management revenue would have posted 1% growth compared with the prior year. The China quick commerce business brought in RMB 53.3 billion in revenue, up 45%, with Freshippo and Taobao instant commerce fueling that gain. Alibaba e-commerce Group's adjusted EBITDA stayed roughly flat against the prior year at RMB 39.7 billion, which highlights how we control costs even while putting more money into user experiences and technology. Taobao instant commerce kept narrowing its losses on a quarter-over-quarter basis while holding onto market share, thanks to larger average order values and better fulfillment logistics efficiency. Beyond that, Ad Express reached profitability ahead of schedule during this quarter. Our goal is to keep profits steady in our conventional e-commerce operation while continuing to push profitability gains in our quick commerce operation. Let's now go through the business updates and results for Ali Cloud — AI Cloud and compute services, which brings together the Cloud Intelligence Group and T-Head. Growth compared with a year earlier for both total revenue and revenue from external customers sped up to 45%. Alibaba Cloud's revenue likewise accelerated, rising 45% versus the prior year. We have confidence that the growth rate will speed up even more across upcoming quarters. AI-related product revenue for this quarter came to RMB 12.4 billion, which works out to an annual revenue run rate of RMB 49.5 billion. That marked a 12th straight quarter of triple-digit growth and made up 35% of external cloud revenue. The adjusted EBITDA margin widened to 12%, fueled by better economies of scale and stronger pricing power for AI-related products against a backdrop of tight market supply. We anticipate the EBITDA margin will keep widening at a steady pace in the quarters ahead through better resource utilization, optimization of our model portfolio, and innovation in new scenarios, as we speed up expansion of the AI plus cloud business and capture greater scale benefits. AI Lab applications consists of AI model labs Qwen Consumer Business Group and Qwen work. Its adjusted EBITDA came in as a loss of RMB 13.9 billion, mostly because we invested more in AI capabilities and faced higher influence costs tied to Qwen APP. That loss shrank considerably on a quarter-over-quarter basis thanks to lower marketing expenses for Qwen ABB. We anticipate the segment loss will shrink across the quarters ahead, propelled by better efficiency in both model training and marketing spend on Qwen ABB. Our frontier models for language coding, video, audio, image, and music are now out, and they deliver top-tier performance. Since Qwen APP launched, $250 million have enjoyed their first AI-driven shopping experience via Qwen APP's agentic features across a broadening array of e-commerce and other services. Revenue for all other segments held steady at RMB 28.8 billion. Adjusted EBITDA for all other segments came in as a loss of RMB 3.3 billion, mainly because we invested more in technology. At this point AI has moved from incubation into commercialization at scale, and as we grow our market share, reinforce AI leadership, and enhance operating efficiency, we are securing greater strategic and financial flexibility to make disciplined and sustained investments in both full stack AI capabilities and consumption opportunities that drive secular growth and greater value for our shareholders. Thank you. That concludes our prepared remarks. We can now open up for Q&A.

Lydia Lu executive
#4

Toby, thanks. At this point we're opening the Q&A portion. Feel free to pose your questions in either Chinese or English. A translator from an outside firm will handle consecutive interpretation. Should anything conflict, what our management said in the source language takes precedence. Operator, kindly open the Q&A now. Appreciate it.

Alicis Yap analyst
#5

I'd also like to congratulate you on the strong results in cloud. Could the management team talk about what's behind the big jump in CapEx we saw this quarter, and what's driving it? What trend should we expect for CapEx over the next few quarters? Is there anything new regarding that 3-year CapEx plan of $380 billion that was mentioned previously? We'd also find it helpful if management could break down how CapEx is allocated among the various services, including training costs and so on. And finally, what return has management seen on the capital invested in these efforts?

Executive executive
#6

<<foreign_lang>> < > Many thanks for raising this. This is a significant question, and I'd like to use this chance to lay out in general terms what our AI business model looks like and what we anticipate for CapEx moving forward; so back in February of last year, we laid out a 3-year capital investment plan totaling RMB 380 billion, and by the close of the June quarter of this year, RMB 190 billion had already been used, with progress largely matching what we had expected. Although spending this quarter came to RMB 67.1 billion, which is a bit on the high side, hardware deliveries occur on differing procurement cycles, so the cadence and pace of those deliveries can fluctuate. That means it isn't spread evenly from one quarter to the next. The rise, then, mainly comes from swings in the timing of equipment deliveries. At the same time, we purchased more CPUs this quarter, since we're seeing a major spike in demand fueled by the agent-centric era. Naturally, higher prices for semiconductor components have played a part in this as well. So my view is that we shouldn't take this quarter's spending and multiply by 4 to derive an annualized number for the year, nor should we assume a steady linear progression. The build-out has moved forward at a steady rate, but that's the situation overall. Now let me go into more detail about our full stock AI business model. This model is asset-heavy. When you consider every way AI can be monetized, whether through software subscriptions, API calls, models as a service, training, or inference, in each of these areas you need compute centers in order to operate and generate revenue; so monetization is only achievable once that compute capacity exists. What this means is that upfront investment is required so that this business model can grow and be monetized across all those areas. That's the reason we started a heavy investment cycle in hardware in 2025. As I said, this stems from that asset-heavy business model, and to seize that future growth, we must first make these CapEx investments to build out the compute capacity required. Now I'll walk through the reasons we regard returns on capital tied to AI as extremely dependable. Everyone in the sector agrees that today's scarcity of AI compute won't be worked out before 2030 at the earliest, so viewed across the whole industry, it stands to reason that our AI compute investments carry a high degree of certainty. Using today's average gross margins as a rough guide, we can recover our AI-related CapEx within 3 years. Naturally, average gross margin keeps climbing, and we anticipate being able to trim that payback window down, perhaps to 2.5 years. Once that 3-year payback window has passed, the AI assets we've put money into are able to generate strongly positive and sturdy cash flow; for some concrete illustrations, consider an A100 bought in 2020, or an A100 bought in 20 — pardon me, a V100 bought in 2018. Even those are still operating at full capacity. On top of that, we have 3 levers we can pull to push gross margin and return on invested capital even higher. The first is that we can keep building state-of-the-art models and lift gross margin on the AI products themselves, keep growing the higher-margin model-as-a-service mass businesses, and we can shift our product mix across IAS and across software so the portfolio overall lands at a higher gross margin. Thanks to that gross margin improvement, you've already witnessed a total gain of 4.4 percentage points in Alibaba Cloud's overall segment profitability, which brought it to 11.6% this quarter. That amounts to early proof of the thesis. A crucial element here is our capacity to deploy chips we've developed ourselves. As you're aware, our own in-house proprietary chip spend covers GPUs, CPUs and networking chips, which are the essential chipsets for AI. And within AI data centers, the priciest components are, needless to say, chips and storage. That gives us a very substantial edge, since we can deploy our own proprietary chips; as we scale up deployment of our own proprietary chips inside our data centers, as those chips make up a growing share of total ships and take the place of chips bought commercially, we should anticipate seeing considerably better gross margin along with profitability. Third, and equally important, we have ways to monetize and extract better utilization efficiency from our own cash flow. These take forms such as co-building data centers alongside partners, and pre-charging plus taking prepayments for compute-based services. So these matter as methods for us to further boost ROIC. Via these 3 distinct approaches, we can cut the payback window on AI CapEx down, for instance, to 2.5 years or possibly even 2 years. And a simple framework lets us grasp this. Given our present gross margin level for AI products and assuming a 3-year payback window on CapEx. In theory, holding our growth rate under 33% would already produce positive cash flow; even so, that isn't the strategic path we've chosen right now. Since artificial intelligence is still at quite a nascent point, we intend to put money into capital expenditures in a forceful way and to expand capacity ahead of demand so that our business can grow quickly. With product gross margin getting better and a rising share of our in-house chips replacing others, the time to recover what we spend will come down to 2.5 years or perhaps shorter; accordingly, in that situation, even as we aim for expansion above 40%, we will still keep cash flow in positive territory. That, then, is the strategic path we are taking for the long run.

Charlene Liu analyst
#7

I'm calling in from HSBC. First off, can you bring us up to speed on the latest happenings in quick commerce, and also explain how a number of business lines are being reclassified now that they've been placed together under the Alibaba e-commerce group? What strategic priorities will those business lines have going forward? I'll go ahead and translate my own question real quick. <<foreign_lang>>

Executive executive
#8

Okay. Thank you very much for asking that, and for doing the translation too. In this new fiscal year, we've actually restructured our e-commerce segments. From here on, we'll be sharing updates on how things are going in 4 main areas: China e-commerce, quick commerce, international e-commerce, and global B2B global wholesale. I'd like to walk through what we're prioritizing strategically and what matters most for each of those 4 segments going forward. Starting with China e-commerce. The domestic e-commerce space is facing near-term macroeconomic pressure, but our focus over the long haul is on strengthening core supply capabilities, and we also plan to leverage AI to improve the overall shopping experience and boost operational efficiency throughout the business. So first, regarding supply, since last year Taobao and Tmall have zeroed in on supporting original merchants, branded sellers among them, while also tapping the potential of high-quality white label suppliers from key industrial clusters. We're going to continue strengthening the ties we have with top brand merchants, supporting them in reaching business growth that's both stable and sustainable. For large brands as well as lots of original merchants, Tmall is still the most essential operational hub. Meanwhile, we're pushing further into industrial clusters so we can source quality goods straight from where they originate. We're bringing more manufacturing factories onto our platform, where they operate directly, and we're using our platform's AI capabilities so white-label merchants can move to a managed operation model that's simpler and more efficient. Transactions coming from that industrial cluster managed model keep climbing at a steady pace. Last quarter, through the recent 618 shopping festival, results matched what we expected even with some macroeconomic headwinds, and it's worth highlighting that core merchants posted solid growth. Meanwhile, there are major AI possibilities we're seeing on e-commerce's demand side as well as its supply side. When it comes to consumers, we'll keep rolling out fresh AI-powered experiences and scenarios, including multimodal search along with virtual try-ons. Two aims guide us here: first, applying AI tech to make existing shopping scenarios both more efficient and more enjoyable — and we've already seen AI produce notable efficiency improvements in our product recommendations; second, fueling novel forms of interaction driven by AI. Turning to merchants, we've noticed they're already putting AI to wide use in their operations. We're looking into how AI might be applied throughout different operational links to strengthen merchant capabilities, especially across data analytics, advertising and marketing, and customer service — areas where merchants stand to gain clear advantages. Looking ahead, we'll additionally team up with Qwen Office to roll out AI agents built specifically for e-commerce scenarios. Next, on quick commerce. Following over 12 months of spending and building, Taobao Instant Commerce has seen major shifts in its scale and share of the market, alongside notable gains in user mind share, supply diversity, logistics experience and order volume. During the latest quarter, even as we kept up growth in both users and orders unit economics, UE improved considerably and losses came down markedly. Given that, we will speed up bringing together operations like Freshippo and Tmall supermarket to grow the nonfood categories within the Quick Commerce business, and we'll put special emphasis on growing our front warehouses. Across the last 12 months, Freshippo has moved faster on building front warehouses, producing a year-over-year rise in GMV. Meanwhile, Quick Commerce is going to continue broadening the range of categories it serves and rolling out new ideas in important areas so that customers get a better experience. We anticipate that in the coming fiscal year, the order count for quick commerce in nonfood segments will surpass the order count in food segments, which will drive expansion across numerous physical goods categories spanning the entire e-commerce operation. For the Quick commerce unit, we expect it to become profitable on an overall basis in FY '29. Looking further ahead, we think it could account for 30% of the platform's total GMV, becoming the second growth engine for our e-commerce business. The third item is international e-commerce. In the near term, growth in our international e-commerce operation has actually been under pressure from tariff policies as well as the geopolitical environment. Despite the challenging conditions in the market, our cross-border operations still managed to significantly improve their profitability, all while transaction volume continued to grow. We see lasting potential for growth in the cross-border business, both in terms of transaction scale and profitability. In addition, our local e-commerce platforms in international markets, including Turkey and the Middle East, are growing rapidly, and operating efficiency continues to improve in markets like Southeast Asia. Over the past 2 decades, our B2B businesses — which include the 1688 and alibaba.com platforms — have grown steadily, and we believe AI technology will bring about profound changes to our B2B platforms and could even fundamentally reshape existing business models. In B2B deals, the genetic model is going to play an increasingly significant part. Accio Work is an AI agent we introduced for merchants operating across borders, and soon after it went live it had already attracted more than 50,000 paying merchants. The way B2B merchants conduct business—cross-border ones in particular—is being completely reshaped by AI; our view is that, drawing on the 2 decades of know-how we've accumulated in this field along with our 2 years of experience, we have a chance to build completely new business models and commercial opportunities within B2B and cross-border trade during the AI era. Taking everything into account, we have finished establishing a new strategic positioning for our e-commerce businesses across a number of key areas over these past few years. Looking ahead, our goal is to keep building on what we do well — from synergies within the supply chain through to AI technology — so that we can open up bigger growth opportunities for the e-commerce business in the age of AI, and at the same time create a revenue and profit structure that is more diversified, which will help the segment as a whole develop in a steadier way.

Yang Bai analyst
#9

I want to ask about the cloud and AI business. Alibaba Cloud's revenue growth has been speeding up each quarter, and this quarter it hit 45%. We know that in the past the company set a long-term target of more than USD 100 billion in external cloud revenue across the next 5 years. And now you've said as well that growth will stay on a faster path in the quarters to come. So I have 2 questions. First, for the quarters ahead, how fast do you expect the cloud business to grow, and what are the main drivers behind the ongoing acceleration in cloud computing growth? Second, as you noted, the industry is currently in a period where capacity is fairly tight when it comes to compute supply. And you just said that this supply and demand situation could change around 2030. So from a much longer-term angle, I'd like to ask what the fundamental drivers of cloud business growth are, and whether they are different from the near-term ones?

Executive executive
#10

<<foreign_lang>> < > Thanks for asking. I believe I can go into this across 3 areas. I'll begin with our present business and the pertinent figures. Second, I can talk about what's driving growth; and third, I can offer our longer-term view. based on that analysis. Let me start with part one, which covers our current business and the key metrics. As you've observed, external revenue for the AI and Cloud segment has been accelerating for 9 quarters in a row now; and during this most recent quarter, growth has already risen to 5%. Customer demand is very strong, and our offerings hold a clear competitive edge relative to what other cloud providers offer. Because of that, we anticipate revenue growth will keep accelerating through the quarters ahead. We've seen that AI-related products brought in RMB 12.4 billion in revenue this quarter. Converting that to an annualized U.S. dollar amount, it comes to USD 7.3 billion in yearly revenue. For the next quarter, our own projection is that the same annualized revenue for AI quarters in the following quarter will get close to USD 10 billion. So our growth rate is still extraordinarily strong. At the same time, we also anticipate our EBITDA margin will rise sequentially, quarter over quarter, across the next few quarters. Also, something quite important regarding the cloud business is demand growth for mass we've witnessed a very substantial increase in demand for mass this quarter, along with continued gains in inference efficiencies; so the ARR of our mass business has now gone past RMB 16 billion. And let me be clear; that's the most recent data as of August. It has already exceeded RMB 16 billion. Now, let me go further into the growth drivers inside our business model. You have to grasp one key point first: when it comes to AI, the way Alibaba puts money to work is nothing like what you see at companies focused solely on AI. Our approach is intensive and spans the entire stack — semiconductors, cloud infrastructure, and models alike. Across each of those 3 essential domains, we hold a leading place in the sector. On top of that, we think AI and technology development throughout the industry remains at an early point. Going forward, as technology advances through its various phases, the primary commercial value inside the AI sector could migrate between layers — semiconductors, cloud computing models, and applications. Because we invest across the whole stack, we can provide superior service capabilities and the strongest value-for-money proposition, which puts us in a good position within the industry down the road and guarantees that at every phase of technological progress, we can stay competitive and keep up sustained growth momentum. Now let me turn to what we consider the biggest growth driver ahead over the coming 1 to 2 years, near term. At the close of 2025, we observed exponential demand for commercial insurance services. This explosive rise in inference demand signals a basic shift in the paradigm: compute has become the central asset generating AI revenue; and currently, every AI-related revenue model revolves around AI compute; and concurrently, there's an industry-wide consensus, as I noted, that supply of compute will stay tight for a while. Meanwhile, the fatter gross margins on mass inference services have made a big difference too — if compute used to be a cost center, conventionally, it has now turned into a core productive asset whose value creation moves in positive correlation with revenue. So costly computing power is still scarce throughout the industry, exactly at a moment when GPUs are being adopted widely across many different use cases; consequently, pricing models are gravitating toward the highest-margin, most margin-generative monetization methods. That is what's pushing pricing models for virtually all GPU-related products. What's more, Alibaba, with both comprehensive multimodal model capabilities — our models rank at state-of-the-art level in the industry, handing us a clear edge in capturing the value of that compute power and giving a solid anchor for our pricing strategy — whenever we price for new customers or re-sign — resign contracts with existing customers upon renewal, we can move to healthier pricing models. And so we anticipate this will serve as a strongly positive near-term driver for margin improvement in the year ahead and beyond. Now let me address scale effect and networking effects, which matter greatly as long-term growth drivers in AI cloud; over the last couple of years, many people have posed the question: what is the super app for AI. The response to that is that the actual super app is really compute — cloud-based AI compute — since every one of these workloads has to operate on a complete stack of AI cloud compute. That covers training, inferencing, AI software, and agents that require GPUs, CPUs, storage, databases, virtualization, plus harness tools, among other things. So AI cloud resembles a super city, where the workload plays the role of residents, and continuously iterating full-stack AI cloud services act as the urban infrastructure, which then draws in additional new residents and boosts the stickiness of those already there. That's where you find an extraordinarily potent network effect along with a scale effect. Since we run more data centers than any other cloud provider in Asia, we enjoy the greatest economies of scale. Meanwhile, deploying our proprietary T-Head AI chips at large scale lets us sidestep the steep price premiums tied to buying costly commercial GPUs, thereby preventing erosion of our gross margins. And thanks to our state-of-the-art performance in our proprietary models, we hold strong pricing power over our compute resources. So going forward, from the standpoint of industry development trends and our own product strength, the long-term revenue growth trend and margin expansion trend are extraordinarily strong. Consequently, we have great confidence in our capacity to hit our target of RMB 100 billion in external cloud revenue by 2030; and we have solid visibility into reaching a gross margin of 20%.

Yuan Liao analyst
#11

Congrats on the impressive results this quarter, and particularly on how far the AI business has come. I want to ask a follow-up about the mass business. As Eddie said before, ARR for August had gone past RMB 16 billion in the prior quarter. I think you said the goal for the end of the year is to get above RMB 30 billion in mass ARR. So, with the progress so far, I'm curious whether you expect to make any changes to that year-end target? Also, in the MaaS business, how do the shares break down between our own proprietary models and those from third parties; and with competition around models heating up and more open source models coming out, how might all of that influence gross margin and profitability in the MaaS business.

Executive executive
#12

Thanks for asking. Yes, growth in Bailian's MaaS business truly is very rapid; and in — as of August, we hit RMB 16 billion or went beyond RMB 16 billion in ARR. Given the current momentum and the upcoming wave of new models in our pipeline, our confidence remains intact that we will reach the RMB 30 billion ARR target we set for the close of this year. When it comes to our MaaS platform, the bulk of revenue is still generated by our in-house proprietary model. That being said, contributions from third-party models are far from insignificant. With that in mind, perhaps I should share our perspective on how various model capabilities differ. A lot of customers often need, or prefer, to incorporate several distinct models into their own AI applications, since each model brings its own unique traits and strengths to draw from. So for us and for Bailian, it's beneficial when additional open source models are available on platforms such as ours to handle inferencing. Now, regarding gross margin: whether we're hosting our own proprietary models or third-party models on something like Bailian, the gross margin we can get is quite similar—highly comparable, in fact. On one hand, we're building those proprietary models specifically to keep pushing model intelligence higher, and that's also tied to our broader pursuit of AI. But if you just look at it from the mouse business angle, the gross margin is really very comparable between those two types of models. On the whole, though, when you've got a thriving, vibrant open ecosystem featuring a lot of these open source models, that works out really well for a cloud provider such as Alibaba Cloud.

Alex Yao analyst
#13

I want to return to what Eddie mentioned before, since he discussed in detail how Alibaba is building a complete stock AI ecosystem. What I really want to ask is this: within that full SAC ecosystem, at which layer do you expect value to build up and monetization to become concentrated? Only 3 months after release, we saw that you made the weights of your flagship model, Q1 3.8 MAX, open source. Meanwhile, your own chips are also showing success, and they now serve more than 600 external customers. I'd like to know if what you're describing points to most of tomorrow's worth accruing in the compute tier—or perhaps the orchestration tier—instead of the model tier as such. Or is your view that worth accumulates across varying tiers depending on where the industry stands in its evolution? And looking far out, assuming value and monetization end up mostly packed into the hardware and compute tiers, what's the right lens for thinking about rivalry down the road, given that a government-driven process will handle a big share of allocating that hardware and compute capacity?

Executive executive
#14

Thank you. That's a highly professional question, and truly it boils down to judgment over the long haul. So my sense is that it's inherently bound up with a great deal of uncertainty. What I'm able to tell you, though, is that our investments span every layer of the stack. The implication being that regardless of which layer ends up holding the most worth, and regardless of how that balance might move between layers over various time frames, every one of those layers sits within our ecosystem. I suppose I could offer you my personal near-term perspective, which is that over the short run, I believe chips together with AI cloud infrastructure will capture the bulk of the value. We observe this pattern not only inside China but around the world at many different companies whenever a technology is still young, and particularly when supply is scarce. A significant portion of the worth usually sits within the infrastructure layer and the fundamental hardware—specifically, chips alongside storage. Thus, for Alibaba, we have merged our computing capacity, cloud foundation, and AI inference capabilities into a single primary business division. Now let me address where the greatest commercial payoff from these models will ultimately materialize. That is a matter generating considerable discussion across the sector, and in fact, opinions differ even among people within our own organization. Okay, so what I'm about to say is just my own take, my personal perspective. From where I sit, the way large language models are monetized right now via APIs strikes me as something temporary — a stopgap phase, really — and it's definitely not where the business ultimately lands. We've poured an enormous amount of compute into our whole platform as a company, yet the goal isn't merely capturing that near-term API revenue. My sense is that once we get to AEI, or as we approach AGI, the endgame business model shifts toward shipping actual products and actual outcomes that customers are asking for. It'll involve doing the actual R&D that produces those products and that runs the operations. So the reason all these various AI model companies are pouring so much money in and racing against each other right now isn't just so they can compete in offering that API-based service. It's because their eyes are on that final end game, where I believe the monetization level will be substantially higher — far higher than what you see today from selling the service through API calls. On the hardware side, I'd like to share a few thoughts about our in-house proprietary chips. I realize we haven't shared a great deal on this topic with investors up to now, but the latest T-Head chip generation has already been manufactured and delivered in excess of 500,000 units. In August, the newest generation went live as super nodes within Alibaba's AI cloud. In my view, we're one of just a handful of firms able to roll out this type of proprietary, domestically produced chip at scale. One thing that truly sets our T-Head chips — domestically made chips — apart is that GPU architecture serves as the core technical basis, and they perform strongly across both training and inference workloads. A few hundred businesses have already adopted these chips via Alibaba Cloud, using them both to run inference and to train models, and they span areas like embodied AI, autonomous driving, and large model companies. As for our generation 2 chips, development kicks off in the back half of this year. We anticipate they'll pack remarkably high compute power together with very strong interconnection bandwidth, so they can step in directly for the chips we have now. That said, I'd argue our standing in the chip space is genuinely one of a kind, especially when it comes to training models at large scale. So my view is that no allocation program run by a government for compute supply could produce chips that are this truly competitive. I'd say T-Head's path forward is very clear — it's a central, essential piece of Alibaba Cloud, and our confidence in the competitive edge we hold there remains very high. Having spoken with numerous engineers in various parts of China, I can share that these chips enjoy wide appeal among engineers working across many different engineering fields. In short, when it comes to domestic Chinese chips that can handle both training and inference for many kinds of industries, ours at T-Head are unquestionably the strongest. That makes us the industry's clear number one. When it comes to capacity for making and rolling out products down the road, I believe we can say with confidence that we rank no worse than second. That said, regarding our capability to genuinely get AI chips into the hands of customers, Alibaba Cloud holds the biggest market share in China's cloud and AI sector. So my view is that our advantage in distribution channels is quite strong. Looking at it this way, I have a great deal of confidence in the commercial worth of T-Head chips over the long term.

Lydia Lu executive
#15

Thanks so much. We’re grateful for your backing, and we’re eager to share our next-quarter progress update with you. Thanks.

Operator operator
#16

Appreciate it. Today's call has now reached its end. We're grateful that all of you joined us. You're free to hang up at this point.

Read the full transcript via the API

You're viewing the first half of this call. Get the complete Alibaba Group Holding Limited transcript - plus 132,000+ transcripts from 7,000+ companies, speaker segments and full-text search - through the EarningsAPI REST API or hosted MCP server.

Get an API key View API docs →

For developers and AI pipelines

Programmatic access to Alibaba Group Holding Limited earnings transcripts and 132,000+ others is available through the EarningsAPI REST API and the hosted MCP server. Quarterly plans from $162 - full transcripts, speaker segments, full-text search, and the /api/v1/transcripts/recent polling endpoint for ETL pipelines.