--- title: "BABA (Trans): E-com Steady, AI Push" type: "Topics" locale: "en" url: "https://longbridge.com/en/topics/40743509.md" description: "Below is Dolphin Research's summary of BABA's FY26 Q4 earnings call transcript. For our in-depth read, please see 'Amid the AI boom, is Alibaba bleeding?'.Shareholder returns: the BOD approved an annual dividend of $1.05 per ADS. Guidance: no specific next-quarter or full-year revenue/profit outlook. Management reiterated firm investment in AI and consumer biz, and expects on-demand retail UE (unit economics) to turn positive in FY29." datetime: "2026-05-13T16:50:36.000Z" locales: - [en](https://longbridge.com/en/topics/40743509.md) - [zh-CN](https://longbridge.com/zh-CN/topics/40743509.md) - [zh-HK](https://longbridge.com/zh-HK/topics/40743509.md) author: "[Dolphin Research](https://longbridge.com/en/news/dolphin.md)" --- # BABA (Trans): E-com Steady, AI Push **Below is Dolphin Research's Trans of**$Alibaba(BABA.US) **FY26 Q4 earnings call. For our take, see** [**AI boom, is Alibaba 'bleeding'?**](https://longbridge.com/zh-CN/topics/40743081)**.** **I. Key takeaways** 1\. **Shareholder returns**: BOD approved an annual dividend of $1.05 per ADS. The payout underscores the company's commitment to returning cash. 2\. **Outlook**: No specific revenue or profit guidance for next quarter or the full year, while reiterating firm investment in AI and consumer businesses. Instant retail unit economics (UE) are expected to turn positive in FY29, with potential monthly breakeven in FY27, and actual CapEx may exceed the previously announced RMB 380 bn target. 3\. **Key metrics**: Taobao/Tmall revenue RMB 122.0 bn (+6% YoY), CMR +1% (+8% ex. scope change). Instant retail revenue RMB 20.0 bn (+57% YoY), and Taobao/Tmall adj. EBITDA RMB 24.0 bn (-40% YoY), roughly flat YoY ex. instant retail losses. AIDC revenue +6% YoY, with adj. EBITDA loss narrowing sharply toward breakeven. Cloud external-customer revenue +40% YoY, quarterly AI revenue RMB 9.0 bn (ARR RMB 36.0 bn/$5.3 bn), 30% of external cloud revenue and the 11th straight quarter of triple-digit growth; cloud adj. EBITDA margin was 9.1%. Other businesses revenue RMB 65.5 bn (-21% YoY), with adj. EBITDA loss of RMB 21.2 bn. 4\. **Cash**: As of Mar 31, 2026, net cash was approx. $38 bn. Excluding debt maturing beyond five years, net cash was approx. $58 bn. **II. Call details** **2.2 Q&A** **Q: For Bailian platform AI revenue, what is the split between in-house models and third-party models? How do recent token price hikes impact Bailian and cloud margins?** **A:** This is our first disclosure of revenue data by models vs. application services, which primarily includes two streams: API call revenue on the platform and AI software subscriptions. Most revenue currently comes from API calls. Bailian is an open platform offering both proprietary and third-party models (open- and closed-source), but the bulk of revenue today is from our own models. Those include Tongyi Qianwen, Tongyi Wanxiang, and our speech and video generation models. On token pricing, the market has shifted materially in recent months as AI moves from chatbots to Agents providing capabilities. Agents solve far more complex tasks and therefore require more reasoning compute, which makes customers more willing to accept higher-priced tokens and keeps demand strong and growing. In fact, our supply is not keeping up, and many customers are waiting to onboard. Bailian (MaaS) naturally carries higher GP than IaaS, which is important for margin mix. Inference efficiency continues to improve, with tokens per card/server rising each quarter. As model capabilities advance, token prices are likely to trend up over the next 1–2 years, and rapid growth in this business over the coming quarters should have a very positive impact on overall GPM. **Q: How do you assess ROI on AI investments? How do you balance aggressive AI spend with earnings stability?** **A:** Negative FCF is primarily driven by large-scale AI investments over the past year. We are investing with conviction because we see a historic opportunity. We expect to maintain this pace over the next two years, as we view this as a critical window. On cash flows, Taobao/Tmall is the major contributor and remains very stable. Looking ahead two years, instant retail losses should narrow substantially, while AIDC is expected to move from loss to profit, both favorable for net cash flow. Continued investment in cloud infrastructure will scale AI and cloud revenue and lift GPM, enabling higher net cash flows to fund infra, and our balance sheet is strong with approx. $38 bn net cash as of Mar 31, 2026 and approx. $58 bn ex. \>5-year maturities. We also have strong financing capability in capital markets. (Eddie Wu) The best analogy is manufacturing: we are building two 'factories' now — one for AI training and one for inference, both underpinned by AI data centers. The monetization path is clear on the 2B side via IaaS, Bailian, and AI-native apps. We have no idle cards in our servers today, so the ROI over the next three to five years is very clear. This underutilization risk is minimal given current demand. **Q: What drives UE improvement in instant retail? Any updates to the multi-year outlook?** **A:** After heavy investment, instant retail grew very rapidly over the past year, triggering a fundamental shift in market position. Vs. the same period last year (pre-scale investment), total orders are 2.7x and non-food orders are 3.0x, with both volume and share up meaningfully. Since Apr, we have kept order volumes while materially improving UE through fulfillment/logistics efficiency gains and mix optimization. We are confident UE can turn positive in some months of FY27. While optimizing UE, we will continue to innovate to improve consumer and merchant experience to maintain long-term competitiveness. We are confident the business can achieve overall profitability on the back of scale and share gains. This quarter, instant retail continued to drive synergies with our core e-com business in acquisition, user stickiness, demand coverage, conversion, monetization, and logistics infrastructure. By category, instant retail boosted food & fresh and healthcare, aiding faster growth at Hema and Tmall Supermarket. In the Mar quarter, core e-com GMV and CMR showed strong momentum, and instant retail was a major contributor. The integration is proving accretive to both growth and monetization. **Q: What advantages does Alibaba have in AI vs. other China AI platforms and startups? When will China see a US-like boom in AI coding? Given lower SaaS willingness to pay in China, does that imply weaker monetization for AI coding products than in the US?** **A:** We position Bailian as an open AI inference platform. While most revenue today comes from our proprietary models, our investment spans larger scale and a broader set of model types than startups. Startups may focus on a vertical and move quickly, but from Bailian's perspective they are more partners than competitors. We emphasize model capabilities across domains and verticals — including code models, image models (Tongyi Wanxiang and Happy Oyster), world models, and speech — to provide a full lineup meeting diverse needs. On the coding breakout, China is already there. From Nov–Dec last year to May this year, we saw usage growth on Bailian primarily driven by coding capability upgrades. These models not only augment software engineers, they also solve a wide range of complex tasks across digital productivity scenarios. Coding capabilities have improved substantially in both China and the US, and we see this as a key growth driver over the next 2–3 years. On SaaS willingness to pay, we think this is about to change. As models become stronger, able to solve complex problems and deliver valuable intelligent services, China demand will converge toward the US. In a sense, when the value per token exceeds its cost, token demand becomes almost unlimited. From Nov–Dec last year to May this year, our platform ARR rose more than 10x to over RMB 8.0 bn, and crossing RMB 10.0 bn in the Jun quarter is highly likely. **Q: How do you prioritize and allocate resources between 2B (cloud, Bailian) and 2C (Tongyi app)? If 2B keeps gaining traction, would you shift resources from the Tongyi app toward cloud and Bailian?** **A:** Fundamentally, AI is a shift in the computing paradigm, aimed at helping users complete tasks and solve problems across both 2C and 2B. Currently, willingness to pay is higher on 2B because ROI is easier to demonstrate, so most infrastructure resources are allocated to 2B. However, AI ultimately serves as an assistant for everyday life, learning, and work, solving the same kinds of problems across scenarios. 2C willingness to pay is low today, but the US already shows a per-use model emerging, and we believe China will follow, especially as technology improves and solves real problems better. We expect the commercial model to converge globally, with China likely reaching that stage in the next 1–2 years. Resource allocation will remain dynamic to reflect demand and ROI. **Q: How do you see cloud EBITDA margins over the next few quarters? Can they reach levels comparable to global peers?** **A:** We are still early in AI penetration across industries. Our priority is clear — drive token consumption, gain share, and grow faster than the market to cement absolute leadership, with growth taking precedence over margins. Over the next 3–5 years, there are physical capacity constraints in chips and memory that underpin demand growth. Alibaba Cloud benefits from customer scale and years of CapEx that deliver economies of scale. With scarce supply, the cost to deploy a new server this year is more than twice last year, implying \>100% cost inflation. Given higher replacement costs, we have pricing power with both new and existing customers, and asset repricing should support revenue over time. Secondly, Bailian is growing very fast and carries structurally higher GP than IaaS or traditional IT ops. As inference demand grows exponentially, this will lift GPM. Inference optimization keeps boosting tokens per card. In addition, as our in-house Pingtouge chips scale — offering the best price-to-performance compute on the platform — they will contribute to margin expansion, and we expect a meaningful step-up in cloud GPM over the next 2–3 years, with signs emerging in the next 1–2 quarters. **Q: What CapEx scale is needed to support Bailian and long-term cloud revenue growth? What is the current penetration of Pingtouge chips, and how much could margin improve as penetration rises?** **A:** On last quarter's call we shared a five-year revenue outlook with an ambitious target. Comparing pre-2022 (pre-AI model boom) with 2033 expected demand levels, data center infrastructure needs to grow 10x. Compute can be sourced in multiple ways — partly via CapEx and partly via OpEx — and we are already procuring significant compute through OpEx. Overall, actual CapEx may exceed the previously announced RMB 380 bn target, and with our in-house Pingtouge chips, we can sell AI servers to other compute centers or co-build compute centers. Bottom line, compute infrastructure demand will be 10x the 2026 level. On in-house chips, Pingtouge can be deployed across much of the AI stack, not only compute but full-stack including storage. Penetration is still low due to domestic capacity constraints, though capacity is ramping. As deployment increases, new chips should drive a significant GPM uplift, and while domestic chips trail leading overseas products in energy and manufacturing efficiency, global leading AI chip vendors enjoy 60%–80% GPMs, leaving considerable room for our chips to improve on price-performance as capacity and capabilities scale. **Risk disclosure and statements:**[**Dolphin Research Disclaimer and General Disclosure**](https://support.longbridge.global/topics/misc/dolphin-disclaimer) ### Related Stocks - [09988.HK](https://longbridge.com/en/quote/09988.HK.md) - [89988.HK](https://longbridge.com/en/quote/89988.HK.md) - [BABA.US](https://longbridge.com/en/quote/BABA.US.md) - [KBAB.US](https://longbridge.com/en/quote/KBAB.US.md) - [BABO.US](https://longbridge.com/en/quote/BABO.US.md) - [BABX.US](https://longbridge.com/en/quote/BABX.US.md) - [HBBD.SG](https://longbridge.com/en/quote/HBBD.SG.md)