On 19 November 2026, Alibaba Group released its earnings for the quarter ending 30 September 2026. As Alibaba’s FY runs from April to March, the July-September quarter was Q2 of its 2027 Fiscal Year. 

Revenue rose 10% year on year to RMB271.5 billion (US$40.5 billion), and adjusted EBITA rose 260% to RMB32.6 billion. Revenue from external customers in AI Cloud and Compute Services grew 64%, while the segment’s adjusted EBITA margin reached 14.5%. Customer management revenue – a core measure of the group’s ecommerce operations – fell 5%.

During the earrings call, only one of eight analysts who had questions asked about ecommerce and quick commerce. The RMB100 billion war seems to have been pushed aside as AI became the top focus. Questions on AI included tighter US chip export controls, Alibaba Cloud’s growth and investment plans, demand for Alibaba’s T-Head chips, and development of the Qwen model. 

Alibaba Group CEO Wu Yongming, CFO Toby Xu and Alibaba E-commerce Group CEO Jiang Fan participated in the call. We’ve listened through the call in Chinese – here is our translated transcript of the Q&A session in English. Do note that it might differ from the official translation provided during the call. 

Interestingly, PDD Group conducted their own earnings call at almost the same time. AI was not mentioned once there. You can read our transcript on PDD’s Q3 2026 earnings call here. 


Q&A Session

Q: Gary – Morgan Stanley

Good evening, management, and thank you for taking my question. My question is about the outlook for cloud revenue. First, thank you for providing the target of US$100 billion in cloud revenue by 2030. In light of the latest announcement from the US Department of Commerce’s Bureau of Industry and Security (BIS), how should we think about the trajectory of cloud revenue growth over the remaining two quarters of this fiscal year?

Are you confident that revenue can grow by more than 50% next fiscal year, FY2028? If so, what gives you that confidence under the current circumstances? Thank you.

A: Wu Yongming, Alibaba CEO

Thank you for the question. Our AI infrastructure plans for meeting our long-term 2030 target have always been based on the supply chain for T-Head’s domestically developed chips. As production ramps up, domestically sourced computing capacity will account for a growing share of our total capacity and become the main driver of cloud revenue growth.

Global demand for AI computing capacity still exceeds supply. We expect supply to be particularly tight in China over the next two years. As we continue to develop and scale our full-stack AI products and the supply chain built around T-Head, we still see a clear path to our 2030 targets of US$100 billion in external cloud revenue and a profit margin above 20%.

As for next year, we believe the new global export controls on computing capacity will affect next year’s products, and there is considerable uncertainty across the global supply chain. At the same time, our progress on domestically sourced computing capacity has exceeded our initial expectations. We expect our operating target for next fiscal year to fall in a growth range of 30% to 45%, and we are still assessing it. As we continue to scale up products based on domestically sourced computing capacity in calendar 2028, we expect growth to return to above 50% that calendar year.

Q: Kenneth Fong – UBS

Good evening, management. My question follows up on the tightening of US restrictions on advanced GPUs and overseas computing assets. How should we think about the capital expenditure roadmap for reaching your target of 20 gigawatts of data centre capacity?

We understand that the situation is still evolving and that the pace of capital spending could change. Is it possible that free cash flow could turn positive sooner than originally expected? Thank you.

A: Wu Yongming

Thank you for the question. Policies governing computing capacity are changing significantly around the world, and that will affect Alibaba Cloud’s long-term plans to build data centres in China and overseas. Based on our preliminary assessment, we may lower the long-term 20-gigawatt target and reduce planned capital expenditure accordingly. We will move more quickly on our plans for domestically sourced computing capacity and the associated supply chain.

As the T-Head supply chain continues to scale in calendar 2027 and 2028, we are targeting positive cash flow in the second half of 2028. Our heaviest sustained investment will be in 2027 and the first half of 2028. Once the domestic supply chain can sustain large-scale expansion beyond 2028, we will have passed the most capital-intensive phase of our AI infrastructure buildout. We remain very confident in our long-term cash flow plans.

Q: Joyce – Bank of America

Thank you, and congratulations on a very solid quarter. My question is about the core ecommerce business. Given the weak macroeconomic environment and your cautious fourth-quarter guidance, what is your outlook for Taobao and Tmall Group in the December quarter, particularly for customer management revenue (CMR) growth and adjusted EBITA?

On quick commerce, could you share how the business has performed so far in the December quarter and what timeline you expect for further loss reductions next year? Thank you.

A: Jiang Fan, CEO of Alibaba’s Ecommerce Business Group

I’ll take this question. Overall, our ecommerce segment performed in line with expectations in the third quarter. Domestic ecommerce remained stable. Quick commerce continued to grow strongly year on year while substantially reducing its losses, and profitability improved noticeably across our other businesses.

As you mentioned, China’s consumer spending environment has remained under pressure from the third quarter into the fourth. Consumers have become more cautious about spending, and competition remains intense in some categories. Since the start of the fourth quarter, our platform’s key operating metrics have moved in line with China’s online retail sales of physical goods. In the current environment, essential goods are performing better than discretionary categories, and we will strengthen the platform’s competitiveness in everyday sales of essentials.

Quick commerce remains a key growth driver for the platform over the next few years. In the third quarter, the value of online transactions in non-F&B categories grew 40% year on year, leading the industry. Our local fulfilment warehouse business grew particularly quickly. In categories such as fresh food and everyday goods, quick commerce continues to change how consumers shop and is driving growth in those categories.

On the F&B side, unit economics improved in line with expectations, market share remained stable, and average order value continued to rise. We expect quick commerce losses to keep narrowing. We maintain our target of halving losses next year and reaching profitability in FY2029.

Outside B2C ecommerce and quick commerce, other businesses in our ecommerce segment, such as B2B and Xianyu (Alibaba’s 2nd-hand marketplace), have continued to make progress over the past two years. This quarter, we integrated our international and domestic B2B businesses to connect domestic supply with overseas demand. As we make greater use of AI, we believe the integration will unlock more potential for B2B. Xianyu’s orders grew 29% year on year last quarter, and the business has been profitable for multiple quarters. We expect these businesses to continue contributing to revenue growth and profit across the ecommerce segment.

Q: Robin – Bernstein

Thank you for taking my question. I have two parts. First, could you discuss the impact of the new BIS restrictions on model training? Will they affect how quickly you can iterate on your flagship Qwen model?

Second, we understand some funding for your data centre expansion will come from partners (i.e. your customers). Could you share an update on that? Have you reached any such agreements with customers that would validate the model?

A: Wu Yongming

Thank you for the question. The new export controls on computing capacity are being implemented, and there are many detailed provisions. We use computing resources around the world and have always complied with applicable laws and regulations. We may need more specific information before we can fully assess the impact on our products and financial results.

If the rules affect the timelines for some upcoming products, I expect the impact to be fairly short-lived. In the first quarter of 2027, our domestically developed T-Head chips will be fully capable of training very large models at scale. We have already started training the next generation of our models on this year’s T-Head chips.

As production of T-Head’s domestically developed chips continues to ramp up, those chips will support our model development over the long term. In the future, the vast majority of models from us, our partners and our customers will be trained entirely on domestic chips, from pre-training through post-training. As production continues to expand in 2027 and 2028, we will have more control over the pace of model development.

Q: Ellie Zhang – Macquarie

Thank you for taking my question. My question is mainly about T-Head. It’s encouraging to hear that chip production is scaling steadily and that the product pipeline looks promising. Over the next few years, what share of your additional computing capacity for training and inference can T-Head supply?

One quick follow-up. Could you share an update on the progress and timing of T-Head’s spin-off listing? Once more details are available, how should we think about how T-Head allocates resources between internal and external customers, and what that means for the profit margins of the cloud business? Thank you.

A: Wu Yongming

Thank you for the question. Under the new circumstances, we expect most of our additional AI chip supply in 2027 and 2028 to come from our own domestically developed T-Head chips. We will also buy some other domestic chips to support some of our internal inference needs.

We will prioritise large-model training when allocating our flagship chips. That includes training for key partners we work closely with. We will take a balanced approach, securing the capacity we need to keep developing our own models while also taking into account the needs of important partners and customers. For inference, we will try to buy a wider range of chips, including domestic chips from a number of other brands. That is broadly what we expect.

There are a few factors to consider when it comes to cloud business profit margins. First, computing capacity in China could be at its tightest in 2027 and 2028. That scarcity should support compute prices and gross margins. We also have advantages in procurement and supply chain management because we use our own chips and production continues to scale. We expect that to improve margins compared with buying chips externally for our cloud and AI businesses.

A: Xu Hong, Alibaba CFO

I’ll take the question about T-Head’s spin-off listing. As we announced recently, T-Head has confidentially submitted a listing application to the relevant regulatory authorities. Listing rules limit what we can disclose at this stage about T-Head’s finances and other non-public information. We can confirm that domestically sourced computing capacity will account for a growing share of Alibaba Cloud’s total capacity. That is a structural benefit for our cloud business as a whole.

Q: Ronald – Goldman Sachs

Thank you, Eddy, Toby, Fan and Lydia. It’s good to see that losses in the AI Labs and Applications segment narrowed quarter on quarter. We are seeing other major Chinese internet companies increase investment at the model and application layers. Where does management think most of the value in AI will ultimately accrue? Will it be primarily in applications and agents?

If so, how should we think about the allocation of investment in Qwen? How should we understand the differences between Qwen and the other models in the Qwen family, and how will that affect the size and allocation of investment and the trajectory of losses over the next few quarters or the next year or two?

A: Wu Yongming

Thank you for the question. People and companies in the industry have different views. We’re still in the early stages of AI, and I think we have to consider models and applications together when we ask where the value will be. Models and applications need to develop together. Models improve applications, and applications in turn can support model growth.

If an application has no stronger model behind it, or no distinctive, more capable model integrated into it, its ability to retain users over the long term is uncertain. If everyone uses the same underlying model, the application is easy to replace. We believe models and applications need to develop together to create products with distinctive capabilities that stay in use inside companies or on customers’ desktops over the long term.

For now, our strategy is to develop applications alongside models and build stronger product capabilities. We are taking a measured approach to aggressive user acquisition and subsidies. Over the long term, we see more promise in pairing a distinctive model with a distinctive application to give customers something valuable that is hard to replace. That combination is what can deliver lasting value.

Q: Thomas Chung – Jefferies

Good evening, and thank you for taking my questions. The first is about annualised revenue (ARR). You previously highlighted that ARR crossed a new milestone of RMB 20 billion in August. Do you think ARR can reach RMB 30 billion by the end of the fiscal year? I’m sure you are closely tracking ARR and analysing the competitive landscape. Could you share what you’re seeing on the competition?

My second question is about your international strategy. How do you expect the mix of domestic and overseas cloud services revenue share to evolve? Thank you.

A: Wu Yongming

Thank you for the questions. Revenue from our model-as-a-service (MaaS) business depends quite heavily on new model releases. Over the past two months, releases from both us and our partners have generally been slower than we expected. We just launched Qwen4-Max, and we have high expectations for our next-generation video model, which we will start rolling out soon. As we and our partners release new models, I expect ARR growth to accelerate soon. As for December, our ARR target of RMB 30 billion for this fiscal year remains unchanged.

We are assessing how the new export controls on computing capacity will affect our overseas cloud business. We need more time to understand the specific impact. Alibaba Cloud currently operates 109 availability zones across 32 regions worldwide, and our data centres in the Netherlands, France and Brazil are gradually coming online.

Our overseas business will remain a priority. For AI overseas, we could pursue different types of partnerships and may need more international partners. Over the longer term, exporting our own chips could be one option, provided we first meet domestic demand. We remain confident in the long-term prospects of Alibaba Cloud’s full-stack infrastructure as we serve Chinese companies expanding overseas and international customers.

Q: Alicia Yap – Citigroup

Good evening, management, and thank you for taking my question. I’d like to follow up on cloud margins. Could you explain the main drivers of the significant improvement in cloud margins this quarter? How should we think about the trend over the next few quarters?

Given the constraints on computing capacity, does Alibaba have room to raise prices for its cloud products in response to stronger demand? And could AI Labs’ losses narrow more quickly as a result? Thank you.

A: Wu Yongming

Thank you for the question. A few factors are driving the continued improvement in AI cloud margins.

First, and most importantly, higher-margin AI products account for a growing share of recurring revenue, lifting overall cloud business profit margins.

Second, computing capacity is scarce globally, and the shortage is especially pronounced in China, as I mentioned earlier. That gives us more leverage when negotiating prices with customers across our product range. Compared with last year, we have more room on gross margins for essentially every product.

Third, as our data centre footprint expands, we are seeing greater economies of scale in our supply chain. On the customer side, switching costs rise as customers use more of our products, and customers’ growing acceptance of our offerings is also helping improve margins.

Over the longer term, advances in AI have structurally changed the cloud business. AI cloud services, particularly those built on in-house chips, differ from our traditional general-purpose cloud computing business. With our own chips and models and a full stack of AI services, Alibaba Cloud has become a differentiated offering. Customers have different requirements across the full stack, which makes straightforward price comparisons difficult.

I have met quite a few customers recently. My impression is that in the AI era, their relationship with Alibaba Cloud has expanded from buying cloud services to developing products together. That is a structural change for the business. We can support customers across the entire product stack, and they are more willing to work with us over the long term. Product differences also make it harder to compare bids and prices directly. I believe those factors will structurally improve margins in our AI cloud business over the long term.

 

[THE END]



Source note

Momentum Works translated this account from Chinese-language notes of Alibaba’s 19 November 2026 earnings call and edited it for readability. The translation may contain errors or omissions. For the official figures and webcast, refer to Alibaba’s results announcement and earnings webcast page.

Follow our thinking

Follow our thinking

Get our latest research, analysis, and invitations in your inbox.