阿里巴巴 (BABA) 2027 Q1 法說會逐字稿

Good day, ladies and gentlemen. Thank you for standing by. Welcome to Alibaba Group's June quarter 2026 results conference call. (Operator Instructions)

各位女士、先生，大家好。感謝各位撥冗等候。歡迎參加阿里巴巴集團 2026 年 6 月季度業績電話會議。（接線員指示）

I would now like to turn the call over to Lydia Lu, Head of Investor Relations of Alibaba Group. Please go ahead.

現在我想把電話交給阿里巴巴集團投資者關係主管 Lydia Lu。請開始。

Thank you. Good day, everyone, and welcome to Alibaba Group's June quarter 2026 earnings conference call. Joining the call today are Joe Tsai, Chairman; Eddie Wu, Chief Executive Officer; Toby Xu, Chief Financial Officer; Jiang Fan, Chief Executive Officer of Alibaba E-commerce Business Group.

謝謝。各位好，歡迎參加阿里巴巴集團 2026 年 6 月季度業績電話會議。今天與會者包括主席蔡崇信、首席執行官吳泳銘、首席財務官徐宏，以及阿里巴巴電商事業群首席執行官蔣凡。

Before we get started, I would like to remind you that today's discussion may contain forward-looking statements based on management's current expectations that are subject to risks and uncertainties. We also make reference to non-GAAP financial measures. Reconciliations between GAAP and non-GAAP measures are included in today's earnings press release and investor presentation. Our comments will be on year-over-year comparisons unless we state otherwise. A replay of the call will be available on our website later today.

在開始之前，我想提醒各位，今天的討論可能包含前瞻性陳述，係基於管理層目前的預期，並可能受到風險與不確定性影響。我們也會提及非 GAAP 財務衡量指標。GAAP 與非 GAAP 指標之間的調節表已包含於今日的財報新聞稿與投資者簡報中。除非另有說明，我們的評論將以同比比較為基礎。本次電話會議的重播將於今日稍晚在我們的網站提供。

With that, I would like to turn the call over to Eddie.

接下來，我想把電話交給 Eddie。

(interpreted) Good evening, good morning, and welcome to Alibaba Group's Earnings Call for the First Quarter of Fiscal Year 2027. Over the past quarter, Alibaba's strategic AI investments have translated into robust results with a total group revenue growing 9% year over year. AI commercialization has also accelerated across the board. Alibaba Cloud's external revenue grew 45% and EBITDA increased 133% year over year, continuing to deliver on our commitment to accelerate growth.

（口譯）各位晚上好、早上好，歡迎參加阿里巴巴集團 2027 財年第一季度業績電話會議。在過去一個季度，阿里巴巴在 AI 方面的策略性投資轉化為強勁成果，集團總收入同比增長 9%。AI 商業化也在各條業務線全面加速。阿里雲外部收入同比增長 45%，EBITDA 同比增長 133%，持續兌現我們加速增長的承諾。

Revenue from AR rated products has maintained a triple-digit growth for the 12th consecutive quarter with annual revenue run rate surpassing RMB49.5 billion around USD7.3 billion. It is the core engine of Alibaba's Cloud's growth acceleration.

來自 AI 相關產品的收入已連續第 12 個季度保持三位數增長，年化收入運行率超過人民幣 495 億元，約合 73 億美元。這是阿里雲增長加速的核心引擎。

I'll now walk you through four key areas: AI and cloud commercialization, full-stack AI capabilities, AI application ecosystem and consumption business.

接下來我將從四個重點領域進行介紹：AI 與雲商業化、全棧 AI 能力、AI 應用生態，以及消費業務。

First, AI and cloud commercialization accelerated across the board and is expected to sustain high growth going forward. This quarter, Alibaba Cloud's external revenue growth accelerated to 45%, a 22 quarter high, while adjusted EBITDA margin reached 11.6%. Notably, this 45% growth was broad-based, driven by compute storage Model as a Service, MaaS, and AI applications.

第一，AI 與雲商業化全面加速，並預期未來將維持高增長。本季度，阿里雲外部收入增速加快至 45%，創 22 個季度新高，同時調整後 EBITDA 利潤率達到 11.6%。值得注意的是，這 45% 的增長具有廣泛性，主要由算力與儲存、模型即服務（MaaS）以及 AI 應用所帶動。

We proactively scaled back low-margin business continuing to improve the quality of our growth. This quarter, annual revenue run rate from AI-related products exceeded RMB49.5 billion and its share of Alibaba Cloud's external revenue rose to 35%.

我們主動縮減低毛利業務，持續提升增長品質。本季度，AI 相關產品的年化收入運行率超過人民幣 495 億元，其在阿里雲外部收入中的占比提升至 35%。

AI-related products generate significantly higher gross margins than the average cloud portfolio. Our recurring AI-related product revenue spans multiple layers, AI compute MaaS and AI applications. This multilayered mix of AI revenue sources and monetization models means growing customer demand at any layer converts directly into commercial opportunity for us. This structural advantage will underpin sustained rapid growth in recurring AI-related product revenue going forward.

AI 相關產品的毛利率顯著高於雲產品組合的平均水準。我們的經常性 AI 相關產品收入覆蓋多個層次，包括 AI 算力、MaaS 與 AI 應用。這種多層次的 AI 收入來源與變現模式組合，意味著任何一個層面的客戶需求增長，都能直接轉化為我們的商業機會。這一結構性優勢將支撐未來經常性 AI 相關產品收入持續快速增長。

The surge in AI agents directly drives demand for tokens and GPU compute while also significantly boosting demand for our traditional cloud products across CPU compute, storage databases, and networking. Alibaba Cloud is undergoing a comprehensive upgrade to an agentic cloud.

AI 智能體的激增直接帶動對 token 與 GPU 算力的需求，同時也顯著提升對我們傳統雲產品的需求，涵蓋 CPU 算力、儲存、資料庫與網路等。阿里雲正在進行全面升級，邁向智能體雲（agentic cloud）。

Based on the latest data, the ARR of our model and application services, including MaaS, has surpassed RMB16 billion. Based on current market feedback and our contract pipelines, compute demand will continue to outstrip supply. As we continue to ramp up our supply, our AI and cloud revenue growth will accelerate further in the coming quarters, alongside continued improvement in profitability.

根據最新數據，我們的模型與應用服務（包括 MaaS）的 ARR 已超過人民幣 160 億元。基於目前的市場回饋與我們的合約管線，算力需求將持續超過供給。隨著我們持續提升供給，未來幾個季度 AI 與雲收入增長將進一步加速，同時盈利能力也將持續改善。

Second, our full stock AI capabilities continue to strengthen, marked by the scaled commercialization of proprietary chips, faster model iteration, and a thriving open-source ecosystem. This quarter, deepening synergy between proprietary T-Head chips and proprietary foundation models further improved our AI commercialization efficiency.

第二，我們的全棧 AI 能力持續增強，體現在自研晶片的規模化商業化、更快的模型迭代，以及蓬勃發展的開源生態。本季度，自研平頭哥晶片與自研基礎模型之間的協同進一步加深，提升了我們 AI 商業化的效率。

T-Head has established a full stock proprietary silicon portfolio, spanning GPU, CPU, and networking chips. As of early August, the Zhenwu chips have served more than 650 customers on Alibaba Cloud. The supernode instance powered by T-Head's next-generation Zhenwu M890 AI processor recently launched on Alibaba Cloud at commercial scale. We expect supply to continue ramping up in the second half of the year to meet strong customer demand.

平頭哥已建立完整的自研矽產品組合，涵蓋 GPU、CPU 與網路晶片。截至 8 月初，鎮武晶片已在阿里雲上服務超過 650 家客戶。由平頭哥下一代鎮武 M890 AI 處理器驅動的超節點實例，近期已在阿里雲以商業化規模上線。我們預期今年下半年供給將持續爬坡，以滿足強勁的客戶需求。

Alibaba Cloud's Zhenwu M890 supernode can efficiently run inference workloads for foundation models with more than 2 trillion parameters, both Kimi K3 and Qwen 3.8 Max are already using it to provide MaaS services to external customers.

阿里雲的鎮武 M890 超節點可高效運行參數超過 2 兆的基礎模型推理工作負載；Kimi K3 與 Qwen 3.8 Max 均已使用該超節點為外部客戶提供 MaaS 服務。

At the data center layer, Alibaba Cloud has cut the delivery time for hyperscale AI data centers to 100 days, a world-leading pace that will significantly speed up our global compute infrastructure buildout.

在資料中心層面，阿里雲已將超大規模 AI 資料中心的交付時間縮短至 100 天，達到全球領先水準，將顯著加快我們全球算力基礎設施的建設進度。

At the model here, our model release cadence has intensified over the past months with major iterations across our large language, image, audio, video, and music models, all ranking among the world's top tier.

在模型層面，過去數月我們的模型發布節奏明顯加快，針對大語言、圖像、音訊、影片與音樂模型均進行了重要迭代，整體表現位居全球第一梯隊。

Last week, we opened the modeled weights of Qwen 3.8 Max with 2.4 trillion parameters and the Qwen 3.8 27B model series. To date, the Qwen model series has been downloaded more than 3 billion times globally with more than 300,000 derivative models built on it. We believe a thriving open-source model ecosystem drives greater demand for our cloud computing services creating a virtuous cycle.

上週，我們開放了參數達 2.4 兆的 Qwen 3.8 Max 以及 Qwen 3.8 27B 模型系列的模型權重。截至目前，Qwen 模型系列在全球下載量已超過 30 億次，並基於其構建了超過 30 萬個衍生模型。我們相信，繁榮的開源模型生態將帶動對我們雲計算服務的更大需求，形成良性循環。

Third, our AI native applications span both enterprise and consumer use cases driving rapid growth in token consumption. On the enterprise side, we launched QwenWork, a new AI productivity product built for enterprise workforce scenarios, delivering agent capabilities at scale. We expect productivity agents to become another engine of ARR growth. On the consumer side, the Qwen app continued to steadily grow its user base and is expanding the range of its value-added offerings.

第三，我們的 AI 原生應用覆蓋企業與消費者兩類場景，推動 token 消耗快速增長。在企業端，我們推出了 QwenWork，這是一款面向企業員工工作場景打造的全新 AI 生產力產品，可規模化提供智能體能力。我們預期生產力智能體將成為 ARR 增長的另一個引擎。在消費者端，Qwen App 持續穩步擴大用戶基礎，並拓展其增值服務範圍。

Through close coordination between Alibaba Token Hub and Alibaba Cloud, we're running a highly efficient commercial flywheel across compute models, tokens applications, and monetization.

透過阿里巴巴 Token Hub 與阿里雲的緊密協同，我們在算力、模型、token、應用與變現之間運行著高效率的商業飛輪。

Fourth, our e-commerce businesses remained solid this quarter. In quick commerce, we continued to narrow losses substantially while growing business scale by 45% with unit economics improving quarter over quarter. Having crossed the AI commercialization inflection point this quarter, we're now seeing growth accelerate and margins expand this quarter. Our AI businesses own capacity to self-fund and sustain itself is strengthening giving us greater confidence to keep investing.

第四，我們的電商業務本季度保持穩健。在即時零售方面，我們在業務規模同比增長 45% 的同時，持續大幅收窄虧損，且單位經濟效益環比改善。本季度跨越 AI 商業化拐點後，我們已看到本季度增長加速、利潤率擴張。我們的 AI 業務自我造血並可持續發展的能力正在增強，讓我們更有信心持續投入。

Looking ahead, AI has become Alibaba's most certain growth engine, we will stay strategically disciplined and drive long-term growth through our full stack AI capabilities.

展望未來，AI 已成為阿里巴巴最確定的增長引擎；我們將保持策略紀律，依託全棧 AI 能力推動長期增長。

I'll now hand over to Toby to walk you through our financial results. Thank you.

接下來我把電話交給 Toby，請他帶大家回顧我們的財務業績。謝謝。

Thank you, Eddie. Our strategic priorities in AI plus cloud and consumption businesses backed by disciplined investments delivered strong results this quarter. Cloud segment revenue growth further accelerated to 45% with its EBITDA margin sequentially rising to 12%. AI-related product revenue continued to drive this momentum marking the 12th consecutive quarter of triple-digit growth and accounting for 35% of external cloud revenue.

謝謝你，Eddie。我們在 AI+雲與消費業務上的策略重點，配合有紀律的投資，推動本季度取得強勁成果。雲業務收入增長進一步加速至 45%，其 EBITDA 利潤率環比提升至 12%。AI 相關產品收入持續帶動這一動能，實現連續第 12 個季度三位數增長，並占外部雲收入的 35%。

The strong performance demonstrates growing customer adoption of our full stack AI capabilities, spanning AI agents, models, cloud infrastructure, and preparatory chips as well as our enhanced scale efficiencies and the robust pricing power in a supply-constrained market.

強勁的表現顯示客戶對我們全棧 AI 能力的採用持續提升，涵蓋 AI 代理、模型、雲端基礎設施與前置晶片，同時也反映我們規模效率的提升，以及在供給受限市場中的強勁定價能力。

On consumption, Taobao instant commerce continued to improve its unit economics while maintaining market share. Overall e-commerce EBITDA remained relatively stable year over year.

在消費端，淘寶即時商業在維持市佔率的同時，持續改善其單位經濟效益。整體電商 EBITDA 與去年同期相比保持相對穩定。

To realize synergies across our commerce platforms and strengthen our full stack AI capabilities, we have implemented strategic realignment of certain businesses in our financial reporting. Starting from this quarter, our segment reporting will present the following. First, Alibaba E-commerce Group. Second, AI Cloud and Compute Services. Third, AI Labs and Applications. And number four, All others.

為了在我們的商業平台之間實現協同效應並強化全棧 AI 能力，我們已在財務報告中對部分業務實施策略性重整。自本季度起，我們的分部報告將呈現如下。第一，阿里巴巴電商集團。第二，AI 雲與算力服務。第三，AI 實驗室與應用。第四，其他所有。

Now let's look at the financial results for this quarter. Total revenue increased 9% year over year to RMB269 billion, driven by the strong momentum in cloud business and quick commerce. Total adjusted EBITDA decreased 30% to RMB27.3 billion, primarily attributable to the investment in technology, partly offset by the improved operating results in our cloud business as well as enhanced operating efficiencies across various businesses.

現在我們來看本季度的財務業績。總收入同比增長 9% 至人民幣 2,690 億元，主要由雲業務與即時零售的強勁動能帶動。調整後 EBITDA 總額下降 30% 至人民幣 273 億元，主要由技術投入所致，部分被雲業務經營業績改善以及各項業務營運效率提升所抵銷。

Our GAAP net income was RMB10.4 billion, a decrease of 75%, primarily due to the decrease in income from operations and decreasing net gains from disposal of investments and mark-to-market changes of our equity investments.

按 GAAP 計算的淨利為人民幣 104 億元，同比下降 75%，主要由營業利潤下降，以及投資處置收益與權益投資按市價計量變動的淨收益減少所致。

Operating cash flow this quarter increased by 11% to RMB22.9 billion compared to RMB20.7 billion in the same quarter last year. Free cash flow was an outflow of RMB44.7 billion compared to an outflow of RMB18.8 billion in the same quarter last year. The decrease was mainly attributed to the investment in cloud infrastructure. CapEx was RMB67.7 billion this quarter. reflecting our continued investments in AI infrastructure to meet strong and growing customer demand.

本季度經營活動現金流同比增長 11% 至人民幣 229 億元，去年同期為人民幣 207 億元。自由現金流為流出人民幣 447 億元，去年同期為流出人民幣 188 億元。下降主要歸因於雲基礎設施投資。本季度資本開支為人民幣 677 億元，反映我們持續投資 AI 基礎設施以滿足強勁且不斷增長的客戶需求。

The significant year-over-year increase is due to several reasons, including fluctuations in procurement cycles, increasing in CPU compute capacity driven by anticipated growing customer adoption of AI agents and higher pricing of a broad range of chip components.

同比大幅增加的原因包括多項因素，例如採購週期的波動、因預期客戶對 AI 代理的採用增加而帶動 CPU 算力容量提升，以及多類晶片元件價格上升。

As of June 30, 2026, we held approximately USD30.7 billion in net cash excluding debt with maturities beyond five years, our net cash position stands at approximately RMB46.5 billion. This balance sheet strength gives us confidence to invest for robust growth.

截至 2026 年 6 月 30 日，在剔除到期日超過五年的債務後，我們持有約 307 億美元的淨現金；我們的淨現金部位約為人民幣 465 億元。這一資產負債表實力使我們有信心為強勁增長進行投資。

Our AI plus cloud investment has a clear path to attractive ROIC, our service equipment with chips typically reach breakeven within three years. With a five-year useful life, we expect them to get positive free cash flow, at least in the two years following breakeven. For the quarter ended June 30, 2026, we repurchased a series of an aggregate consideration of USD162 billion.

我們的「AI+雲」投資具備清晰的通往具吸引力 ROIC 的路徑；我們搭載晶片的服務設備通常可在三年內達到損益兩平。在五年的使用年限下，我們預期其在達到損益兩平後至少接下來兩年可產生正向自由現金流。截至 2026 年 6 月 30 日止季度，我們回購了一系列股份，合計對價為 1,620 億美元。

We remain committed to maximizing long-term shareholder returns through disciplined capital allocation across investments for AI plus cloud business growth, share buybacks, and dividends. We will adjust our priorities as market conditions and the strategic needs evolve.

我們仍致力於透過嚴謹的資本配置，在「AI+雲」業務增長投資、股份回購與股利之間取得平衡，以最大化股東長期回報。我們將隨市場環境與策略需求的演進調整優先順序。

Now let's first look at our e-commerce businesses. The new Alibaba E-commerce Group reflects our strategic focus on unlocking significant synergies across our domestic and cross-border e-commerce businesses. Starting from this quarter, we will present Alibaba E-commerce Group's revenue as the following. First, China eCommerce. Second, China Quick Commerce. Third, International eCommerce, and fourth, Global Wholesale.

接下來先看我們的電商業務。新的阿里巴巴電商集團反映我們的策略重點：釋放國內與跨境電商業務之間的重大協同效應。自本季度起，我們將按如下方式呈列阿里巴巴電商集團的收入。第一，中國電商。第二，中國即時零售。第三，國際電商；第四，全球批發。

Revenue for Alibaba E-commerce Group was RMB205.9 billion, an increase of 4%. Customer management revenue decreased by 7%. Excluding the contra revenue impact from the new business development program, customer management revenue would have grown by 1% year over year.

阿里巴巴電商集團收入為人民幣 2,059 億元，同比增長 4%。客戶管理收入下降 7%。若剔除新業務發展計畫帶來的抵減收入影響，客戶管理收入同比將增長 1%。

Revenue from China Quick Commerce business was RMB53.3 billion, an increase of 45%, driven by Freshippo and Taobao Instant Commerce. Alibaba E-commerce Group's adjusted EBITDA remained relatively stable year over year at RMB39.7 billion, underscoring our cost discipline against the backdrop of increased investments in user experiences and technology. Taobao Instant Commerce continued to improve its uneconomic quarter over quarter while maintaining market share. driven by higher average order value and enhanced fulfillment logistics efficiency.

中國即時零售業務收入為人民幣 533 億元，同比增長 45%，主要由盒馬與淘寶即時商業帶動。阿里巴巴電商集團調整後 EBITDA 同比保持相對穩定，為人民幣 397 億元，體現我們在加大用戶體驗與技術投入背景下仍保持成本紀律。淘寶即時商業在維持市佔率的同時，季度環比持續改善其單位經濟效益，主要由更高的平均客單價與履約物流效率提升所驅動。

In addition, AliExpress achieved operating profit this quarter. We aim to maintain steady profit in our conventional e-commerce business of continuing to drive profitability improvement in our Quick Commerce business.

此外，AliExpress 本季度實現經營利潤。我們的目標是在傳統電商業務保持穩定盈利的同時，持續推動即時零售業務的盈利能力改善。

Now let's review the business updates and results of Ali Cloud -- AI Cloud and compute services, which comprises the Cloud Intelligence Group and T-Head. The year-over-year growth of total revenue and revenue from external customers both accelerated to 45%. Revenue from Alibaba Cloud also accelerated growing 45% year over year. We are confident the growth rate will further accelerate in the coming quarters.

接下來回顧阿里雲——AI 雲與算力服務（包括雲智能集團與平頭哥）的業務進展與業績。總收入以及外部客戶收入的同比增速均加快至 45%。阿里雲收入同比增速亦加快至 45%。我們有信心在未來幾個季度增速將進一步加快。

This quarter's AI-related product revenue was RMB12.4 billion, implying an annual revenue run rate of RMB49.5 billion. It delivered a 12th consecutive quarter of triple-digit growth and accounted for 35% of external cloud revenue.

本季度 AI 相關產品收入為人民幣 124 億元，意味著年化收入運行率為人民幣 495 億元。其已連續第 12 個季度實現三位數增長，並佔外部雲收入的 35%。

The adjusted EBITDA margin expanded to 12%, driven by improved economies of scale and a stronger pricing power of AI-related products amid tight market supply. We expect EBITDA margin to further expand steadily in the coming quarters by improving resource utilization, optimizing model portfolio, and innovating new scenarios, we are accelerating the growth of AI plus cloud business and driving greater benefits of scale.

調整後 EBITDA 利潤率擴大至 12%，主要由規模經濟改善以及在市場供給偏緊背景下 AI 相關產品更強的定價能力所帶動。我們預期未來幾個季度 EBITDA 利潤率將持續穩步擴張；透過提升資源利用率、優化模型組合並創新新場景，我們正加速「AI+雲」業務增長並釋放更大的規模效益。

AI Labs and Applications comprises AI Model Labs, Qwen Consumer Business Group, and QwenWork. Its adjusted EBITDA was a loss of RMB13.9 billion, primarily due to our increased investment in AI capabilities and higher influence costs related to Qwen app. The loss significantly narrowed quarter over quarter due to the reduction in marketing expenses for Qwen app. We expect the segment loss to narrow over the coming quarters driven by improving efficiency in both model training and marketing spend on Qwen app.

AI 實驗室與應用包括 AI 模型實驗室、通義千問消費者業務集團以及 QwenWork。其調整後 EBITDA 虧損為人民幣 139 億元，主要由我們加大 AI 能力投入以及與通義 App 相關的更高影響力成本所致。由於通義 App 行銷費用下降，虧損較上季度顯著收窄。我們預期未來幾個季度該分部虧損將進一步收窄，主要由模型訓練與通義 App 行銷投放效率提升所驅動。

We have launched our frontier language coding, video, audio, image, and music models, all delivering top-tier performance. 250 million have had their first AI-driven shopping experience through Qwen app's agentic features across an expanding range of e-commerce and other services since the launch of Qwen app.

我們已推出前沿的語言、程式碼、影片、音訊、影像與音樂模型，均展現頂尖水準的表現。自通義 App 上線以來，已有 2.5 億人透過通義 App 的代理式功能，在不斷擴展的電商及其他服務範圍內獲得其首次 AI 驅動的購物體驗。

All other segment revenue remained stable at RMB28.8 billion. All other adjusted EBITDA was a loss of RMB3.3 billion primarily due to our increased investment in technology.

其他所有分部收入保持穩定，為人民幣 288 億元。其他所有調整後 EBITDA 虧損為人民幣 33 億元，主要由我們加大技術投入所致。

AI has progressed from incubation to commercialization at scale as we expand our market share, strengthen AI leadership and improving operating efficiency, we are gaining greater strategic and financial flexibility to make disciplined and sustained investments in both full stack AI capabilities and consumption opportunities driving secular growth and greater value for our shareholders.

隨著我們擴大市佔率、鞏固 AI 領導地位並提升營運效率，AI 已從孵化階段走向規模化商業化；我們正獲得更大的策略與財務彈性，以有紀律且可持續地在全棧 AI 能力與消費機會上進行投資，推動長期結構性增長並為股東創造更大價值。

Thank you. That's the end of our prepared remarks. We can open up for Q&A.

謝謝。以上是我們準備好的發言。我們可以開始問答環節。

Thank you, Toby. We will now begin the Q&A session. 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.

謝謝你，Toby。我們現在開始問答環節。歡迎用中文或英文提問。第三方翻譯將提供逐句口譯。如有任何差異，以管理層以原語言所作陳述為準。

Operator, please start the Q&A session. Thank you.

接線員，請開始問答環節。謝謝。

(Operator Instructions) Alicia Yap, Citigroup.

(接線員指示) Alicia Yap，花旗集團。

Also congrats on your solid cloud performance. Could management please comment on the reasons and the drivers for the significant increase in the CapEx this quarter? And also, what is the expected CapEx trend for the coming quarters? Are there any updates to the existing three-year CapEx budget that you have of this RMB380 billion that you mentioned before?

也恭喜你們雲端業務表現穩健。請管理層評論一下本季資本支出（CapEx）大幅增加的原因與驅動因素？另外，未來幾季的CapEx趨勢預期如何？對於先前提到的人民幣3,800億元三年CapEx預算，是否有任何更新？

And also, we would appreciate if management can also provide a breakdown of the CapEx allocation across the different services, like the training part and all that. What is management expected return on the invested capital for this investment?

另外，也希望管理層能提供CapEx在不同服務之間的配置拆分，例如訓練部分等等。管理層對這項投資的投入資本報酬率（ROIC）預期是多少？

(interpreted) Thank you very much for the question. It's an important question, and I'd like to take the opportunity perhaps to explain generally what our business model is for AI and our expectations around CapEx going forward. So indeed, last February, we announced a three-year capital investment plan with total investment of RMB380 billion as of the end of the June quarter this year, we had already spent RMB190 billion with progress broadly in line with our expectations. While this quarter spending of RMB67.1 billion is somewhat higher hardware deliveries follow different procurement cycles, there can be fluctuations in the cadence and pace of hardware deliveries.

(口譯)非常感謝這個問題。這是個重要的問題，我也想藉此機會，或許從整體上說明我們在AI方面的商業模式，以及對未來CapEx的預期。確實，我們在去年2月宣布了一項三年資本投資計畫，總投資額為人民幣3,800億元；截至今年6月季度末，我們已經投入人民幣1,900億元，進度大致符合我們的預期。本季支出人民幣671億元略高，主要是因為硬體交付遵循不同的採購週期，硬體交付的節奏與速度可能會出現波動。

So it's not evenly distributed across different quarters. So the increase primarily reflects volatility in those equipment delivery schedules. At the same time, we increased procurement of CPUs this quarter as we are witnessing a substantial surge in demand driven by the agent-centric era. Of course, rising prices for semiconductor components have also contributed to this trend.

因此，它不會在各個季度之間平均分配。所以，本季的增加主要反映了設備交付排程的波動性。同時，我們本季也增加了CPU的採購，因為我們看到由「以代理（agent）為中心」時代所帶動的需求大幅上升。當然，半導體元件價格上漲也促成了這一趨勢。

So I don't think we should take the spending for this quarter and multiply it by 4 to come up with an annualized figure for the year or to expect that there'll be a steady linear progression. The build-out has been progressing at a steady pace, but that is the overall situation.

因此，我認為不應該把本季的支出乘以4來推算全年年化數字，或預期會呈現穩定的線性成長。整體建置仍以穩健的步伐推進，這就是目前的整體情況。

Next, let me expand on our full stock AI business model. This is an asset-heavy business model. If you think about all of the different ways that AI is monetized and can be monetized, be it through software subscriptions, be it through API calls through models as a service through training, inference. In all of these different respects, you need compute centers to run and to monetize.

接下來，我想進一步說明我們的全棧AI商業模式。這是一種資產密集型的商業模式。如果你思考AI各種變現與可變現的方式，無論是軟體訂閱、透過API呼叫、以模型即服務（Model as a Service）的形式、或是訓練與推理。在所有這些面向上，你都需要運算中心來運行並實現變現。

So it's only possible to monetize when you have that compute capacity in place. So what that means is that we need to be investing upfront in order to be able to grow this business model and monetize across all of those different areas. So that's why beginning in 2025. we began a heavy investment cycle in hardware. And this is really a function of that asset-heavy business model, as I explained, in order to be able to capture that future growth. We first need to make these CapEx investments to build out the necessary compute capacity.

因此，只有在具備相應的運算能力之後，才有可能實現變現。這意味著我們需要先行投入，才能推動這個商業模式成長，並在上述各個領域實現變現。因此，自2025年起，我們開始進入硬體的重投資週期。正如我所解釋，這確實是資產密集型商業模式所決定的，為了把握未來成長。我們首先需要進行這些CapEx投資，以建置必要的運算能力。

Next, let me explain why we see return on invested capital in AI-related as highly certain. There's consensus across the industry that the current shortage in AI compute will not be resolved until at least 2030. So industry-wide then, it makes sense that there should be high certainty in our investments in AI compute.

接下來，我想說明為什麼我們認為AI相關投資的投入資本報酬率（ROIC）具有高度確定性。業界普遍共識是，目前AI算力短缺至少要到2030年才可能緩解。因此，從全產業角度來看，我們在AI算力上的投資具有高度確定性是合理的。

Based on average gross margins today, roughly, we can break even on AI-related CapEx in three years. And of course, average gross margin continues to rise, and we expect to be able to shorten that payback period, say, to 2.5 years.

以目前的平均毛利率來看，大致上我們可以在三年內讓AI相關CapEx達到損益兩平。當然，平均毛利率仍在上升，我們預期能把回收期縮短，例如縮短到2.5年。

Following that three-year payback period then these AI assets that we've invested in can achieve very positive and robust cash flow. So to give you some direct examples and A100 purchased in 2020 or A100 purchased in 20 -- sorry, V100 purchased in 2018. Even are still running at full capacity.

在三年的回收期之後，我們所投資的這些AI資產就能帶來非常正向且強勁的現金流。舉一些直接例子：2020年購買的A100，或是——抱歉——2018年購買的V100，至今仍在滿載運行。

Additionally, we have three means that we can leverage to further enhance gross margin and return on invested capital. First is we can continue to develop state-of-the-art models and enhanced gross margin on AI products themselves and continue to expand a higher-margin Model-as-a-Service MaaS businesses.

此外，我們還有三種方式可以進一步提升毛利率與投入資本報酬率。第一，我們可以持續開發最先進的模型，提升AI產品本身的毛利率，並持續擴大高毛利的模型即服務（MaaS）業務。

And we can adopt our product mix across IaaS and across software to achieve higher gross margin on the portfolio as a whole. And as a result of improving gross margin, you've already seen an overall increase of 4.4 percentage points in Alibaba Cloud's overall segment profitability, bringing it this quarter to 11.6%. So that represents initial validation of that thesis.

第二，我們可以在IaaS與軟體之間調整產品組合，以提升整體組合的毛利率。隨著毛利率改善，你們已經看到阿里雲整體分部獲利能力提升了4.4個百分點，使本季達到11.6%。這代表我們的論點已獲得初步驗證。

A very important piece of this is our ability to deploy our own proprietary chips. As you know, our own T-Head proprietary chips span GPUs, CPUs, and networking chips, which are the critical chipsets for AI. In AI data centers, the most expensive components are of course chips and storage. We have a very significant advantage in being able to deploy our own proprietary chips.

其中非常重要的一環，是我們部署自研晶片的能力。如你所知，我們自研的平頭哥（T-Head）晶片涵蓋GPU、CPU與網路晶片，這些都是AI的關鍵晶片組。在AI資料中心中，最昂貴的元件當然是晶片與儲存。我們能部署自研晶片，具備非常顯著的優勢。

As we ramp up deployment of our own proprietary chips in our data centers, as they account for an increasing proportion of total chips and replace commercially procured chips, we can expect to see substantially higher gross margin as well as profitability.

隨著我們在資料中心加速部署自研晶片，當其在總晶片中的占比提高並取代商用採購晶片時，我們預期毛利率與獲利能力都將顯著提升。

Third, also very importantly, we have means to monetize and get better efficiency of utilization of our own cash flow. These include, for example, co-building data centers with partners, as well as pre-charging and receiving prepayments for compute-based services. These are important ways in which we can further enhance ROIC.

第三，同樣非常重要的是，我們也有方式可以變現，並提升自有現金流的使用效率。例如與合作夥伴共同建置資料中心，以及對以算力為基礎的服務進行預先收費並取得預付款。這些都是我們進一步提升ROIC的重要方式。

So through these three different methods, we can shorten the payback period for AI CapEx, for example, to 2.5 years or even 2 years. And we can apply a simple framework to understand this. At our current level of gross margin for AI products and under the assumption of a three-year payback period on CapEx. Theoretically, keeping our growth rate below 33% would already enable positive cash flow.

因此，透過這三種不同方法，我們可以縮短AI CapEx的回收期，例如縮短到2.5年甚至2年。我們也可以用一個簡單框架來理解：在目前AI產品毛利率水準、且假設CapEx三年回收期的情況下，理論上，只要把成長率控制在33%以下，就已能實現正向現金流。

However, that is not our strategic choice at this time. Given that AI remains in a very early stage, we're committed to aggressively investing in CapEx and proactively scaling up to drive our rapid business expansion.

然而，這並不是我們目前的策略選擇。鑑於AI仍處於非常早期的階段，我們承諾將積極投入CapEx，並主動擴大規模，以推動業務快速擴張。

As our product gross margin improves and our proprietary chip substitution rate increases, our payback period will shorten to 2.5 years or even less. And so under those circumstances, while pursuing growth of over 40%, we'll also be able to maintain positive cash flow. So that is our long-term strategic direction.

隨著產品毛利率提升，以及自研晶片替代率提高，我們的回收期將縮短至2.5年甚至更短。在這種情況下，即使追求超過40%的成長，我們也能維持正向現金流。這就是我們的長期策略方向。

Charlene Liu, HBSC.

Charlene Liu，匯豐。

I'm from HSBC. First, when we get an update on the latest developments in quick comers and under the reclassification of multiple business lines, which are regrouped under the Alibaba E-commerce Group. Can you talk about the future strategic focuses of these lines of businesses. Let me quickly translate the question myself. (spoken in foreign language)

我來自匯豐。首先，關於即時零售（quick commerce）的最新進展，以及在多條業務線重新分類、並重新歸入阿里巴巴電商集團之後，何時能有更新？能否談談這些業務線未來的策略重點？我先快速把問題自己翻譯一下。(以外語發言)

(interpreted) Okay. Thank you very much for the question as well as for the translation. In the new fiscal year, indeed, we've realigned our e-commerce business segments. And moving forward, we'll be updating progress on four core areas: China e-commerce, quick commerce, international e-commerce, and global B2B global wholesale. Let me then briefly share the strategic priorities and key considerations for each of these four segments in the period ahead.

(口譯)好的。也非常感謝你的提問以及翻譯。在新的財政年度，我們確實重新調整了電商業務分部。展望未來，我們將就四個核心領域更新進展：中國電商、即時零售（quick commerce）、國際電商，以及全球B2B全球批發。接下來我將簡要分享未來一段時間內，這四個分部各自的策略重點與關鍵考量。

So starting with China e-commerce. While the domestic e-commerce landscape faces short-term macroeconomic challenges, our long-term strategy centers on strengthening core supply capabilities and at the same time, we aim to leverage AI to enhance the overall shopping experience and improve operational efficiency across the board.

那麼先從中國電商開始。儘管國內電商環境短期面臨宏觀經濟挑戰，我們的長期策略聚焦於強化核心供給能力，同時也希望運用 AI 來提升整體購物體驗，並全面提高營運效率。

So first, regarding supply, since last year, Taobao and Tmall have focused on supporting original merchants, including branded sellers, while simultaneously unlocking the potential of high-quality white label suppliers from key industrial clusters.

首先，在供給方面，自去年以來，淘寶與天貓聚焦支持原創商家（包括品牌商家），同時釋放來自重點產業集群的高品質白牌供應商的潛力。

We will continue to strengthen our partnerships with leading brand merchants, helping them achieve stable and sustainable business growth Tmall remains the most critical operational hub for both major brands and many original merchants. At the same time, we are diving deeper into industrial clusters to source high-quality products directly from their origins. We are supporting more manufacturing factories and operating directly on our platform and leveraging our platform AI capabilities to enable white-label merchants to adopt a simpler and more efficient managed operation model.

我們將持續強化與頭部品牌商家的合作，協助其實現穩定且可持續的業務成長。天貓仍是大品牌與眾多原創商家最關鍵的經營陣地。同時，我們也將更深入產業集群，直接從源頭採購高品質商品。我們正在支持更多製造工廠在平台上直接經營，並運用平台的 AI 能力，讓白牌商家採用更簡單、更高效率的託管式經營模式。

And the share of transactions being generated through that industrial cluster managed model continues to rise steadily. In the past quarter, during the recent 618 shopping festival despite certain macroeconomic challenges, the outcomes were aligned with our expectations and notably, core merchants achieved solid growth.

透過該產業集群託管模式所產生的交易占比仍在穩步提升。在過去一個季度，於近期 618 購物節期間，儘管面臨一定的宏觀經濟挑戰，整體結果符合我們的預期，且核心商家實現了穩健成長。

At the same time, we see significant opportunities for AI across both the supply and demand sides of e-commerce. On the consumer side, we will continue to launch new experiences and scenarios powered by AI, such as multimodal search and virtual try-ons, our goal is twofold: first, to use AI technology to enhance the experience and efficiency of existing shopping scenarios, and we've already observed that AI has driven significant efficiency gains in our product recommendations and secondly, to drive new kinds of AI-driven interaction.

同時，我們看到 AI 在電商供給端與需求端都存在顯著機會。在消費者端，我們將持續推出由 AI 驅動的新體驗與新場景，例如多模態搜尋與虛擬試穿；我們的目標有兩點：第一，運用 AI 技術提升既有購物場景的體驗與效率，我們已觀察到 AI 在商品推薦方面帶來顯著的效率提升；第二，推動新型態的 AI 驅動互動。

On the merchant side, we observed that merchants are already widely adopting AI in their operations. We're exploring ways to leverage AI across various operational links to boost merchant capabilities, particularly in data analytics, advertising and marketing, and customer service, where merchants can derive clear benefits. And going forward, we'll also collaborate with Qwen Office to launch AI agents that are specifically tailored for e-commerce scenarios.

在商家端，我們觀察到商家已在營運中廣泛採用 AI。我們正在探索在各個營運環節運用 AI 以提升商家能力，特別是在數據分析、廣告與行銷，以及客服等領域，商家可獲得明確收益。展望未來，我們也將與通義辦公（Qwen Office）合作，推出專為電商場景量身打造的 AI 代理（AI agents）。

Next, on quick commerce. After more than a year of investment and development, Taobao Instant Commerce has undergone substantial changes in scale and in market share with significant improvements across user mind share, supply diversity, logistics experience and order volume. Last quarter, while maintaining growth in both users and orders unit economics, UE substantially improved and losses significantly reduced.

接下來是即時零售（Quick Commerce）。經過一年多的投入與發展，淘寶即時零售在規模與市占率方面都發生了顯著變化，在用戶心智、供給多樣性、物流體驗與訂單量等方面均有明顯提升。上個季度，在用戶與訂單保持成長的同時，單位經濟效益（UE）大幅改善，虧損顯著收窄。

On that basis, we will accelerate the integration of businesses such as Freshippo and Tmall Supermarket to develop the nonfood categories growth within the Quick Commerce business, and we'll place a particular focus on expanding our front warehouses. Over the past year, Freshippo has accelerated the development of front warehouses leading to a year-over-year increase in GMV.

在此基礎上，我們將加速整合盒馬與天貓超市等業務，推動即時零售業務中非食品品類的成長，並將特別聚焦於擴張前置倉。過去一年，盒馬加速前置倉建設，帶動 GMV 同比提升。

Meanwhile, Quick Commerce will continue to expand its category coverage and innovate in key areas to enhance the consumer experience. We expect the transaction volume of quick commerce for nonfood categories to surpass that of food categories within the next fiscal year, driving growth in many different physical goods categories across the overall e-commerce business.

同時，即時零售將持續擴大品類覆蓋，並在關鍵領域創新以提升消費者體驗。我們預期在下一財年，即時零售的非食品品類交易量將超過食品品類，從而帶動整體電商業務中多個實物商品品類的成長。

The Quick Commerce business is expected to achieve overall profitability in FY29. In the long term, we believe it has the potential to contribute 30% of the platform's total GMV, becoming the second growth curve for our e-commerce business.

即時零售業務預計在 FY29 實現整體盈利。長期來看，我們認為其有潛力貢獻平台總 GMV 的 30%，成為我們電商業務的第二成長曲線。

Third is international e-commerce. In the short term, our international e-commerce business has indeed been affected by tariff policies and the geopolitical environment pressuring growth. That said, despite the complex market environment, our cross-border business has delivered significant improvement in profitability while maintaining growth in transaction volume.

第三是國際電商。短期內，我們的國際電商業務確實受到關稅政策與地緣政治環境影響，對成長造成壓力。不過，儘管市場環境複雜，我們的跨境業務在保持交易量成長的同時，盈利能力已顯著改善。

In terms of both transaction scale and profitability, we believe the cross-border business holds long-term growth potential. In addition, our local e-commerce platforms in international markets such as Turkiye and the Middle East are growing rapidly and operating efficiency in markets such as Southeast Asia continues to improve.

無論在交易規模或盈利能力方面，我們都認為跨境業務具備長期成長潛力。此外，我們在土耳其與中東等海外市場的本地電商平台增長迅速，而在東南亞等市場的營運效率也持續提升。

Fourth is global B2B. Our B2B businesses, including the 1688 and alibaba.com platforms have grown consistently over the past two decades and we see that AI technology will bring profound changes to our B2B platforms and may even fundamentally reshape existing business models.

第四是全球 B2B。我們的 B2B 業務（包括 1688 與 Alibaba.com 平台）在過去二十年持續成長；我們也看到 AI 技術將為 B2B 平台帶來深刻變革，甚至可能從根本上重塑既有商業模式。

In particular, the agentic model will play an increasingly important role in B2B transactions. We've launched Accio Work, which is an AI agent for cross-border merchants, and it had already attracted over 50,000 paying merchants shortly after its launch.

尤其是代理式（agentic）模型將在 B2B 交易中扮演愈來愈重要的角色。我們已推出面向跨境商家的 AI 代理 Accio Work，上線後不久便吸引了超過 50,000 名付費商家。

AI is comprehensively transforming the way that B2B merchants do business especially cross-border merchants. We believe that building on our two years of know-how in the two decades of know-how in this field, we have the opportunity to create entirely new business models and commercial opportunities in B2B and in cross-border trade in the AI era.

AI 正在全面改變 B2B 商家的經營方式，尤其是跨境商家。我們相信，基於我們在該領域二十年的積累，以及近兩年的相關實戰經驗，在 AI 時代我們有機會在 B2B 與跨境貿易中創造全新的商業模式與商業機會。

Overall, over the past few years, we have completed a new strategic positioning for our e-commerce businesses across several key areas. And going forward, we aim to continue leveraging our strengths from supply chain synergies to AI technology to unlock greater growth potential for the e-commerce segment in the AI era, while building a more diversified revenue and profit structure to drive steadier development of the overall segment.

總體而言，過去幾年我們已在多個關鍵領域完成了電商業務的全新戰略定位。展望未來，我們希望持續運用從供應鏈協同到 AI 技術等優勢，在 AI 時代釋放電商板塊更大的成長潛力，同時打造更為多元的收入與利潤結構，推動整體板塊更穩健的發展。

Yang Bai, CICC.

楊柏，中金公司。

(interpreted) My question is about the cloud and AI business. We've seen that Alibaba Cloud's revenue growth has been accelerating quarter by quarter reaching 45% this quarter. We know the company has previously set a long-term goal of exceeding USD100 billion in external cloud revenue over the next five years. And you've also now indicated that growth will remain on an accelerated trajectory in the quarters ahead.

（口譯）我的問題關於雲與 AI 業務。我們看到阿里雲的收入增速逐季加快，本季度達到 45%。我們知道公司此前設定了長期目標：未來五年外部雲收入超過 1,000 億美元。而你們也表示，未來幾個季度增長仍將保持加速態勢。

So I'd like to ask two questions. First, looking ahead to the coming quarters, what do you anticipate being the pace of growth in the cloud business. what are the core drivers underpinning the continued acceleration of cloud computing growth?

因此我想問兩個問題。第一，展望未來幾個季度，你們預期雲業務的增長節奏如何？支撐雲計算增長持續加速的核心驅動因素是什麼？

And then secondly, as you mentioned, the industry is now in a phase of relatively tight capacity in terms of supply of compute. And you just mentioned that, that supply demand dynamic may shift around 2030. So I'd like to ask from an even longer-term perspective, what are the fundamental growth drivers for the cloud business? And do they differ from those in the short term?

第二，如你所提到，行業目前在算力供給方面處於相對偏緊的產能階段。你剛才也提到，供需格局可能在 2030 年前後發生變化。因此我想從更長期的視角請教：雲業務的根本增長驅動因素是什麼？它們與短期驅動因素是否有所不同？

(interpreted) Thank you for the question. And I think I can expand on this in three different areas. I can start by looking at our current business and the relevant data. Secondly, I can discuss the drivers for growth. And then thirdly, I can share with you our long-term perspective based on that analysis.

（口譯）謝謝你的提問。我想可以從三個方面展開。我可以先從我們當前的業務與相關數據談起。第二，我可以討論增長的驅動因素。第三，我可以基於上述分析與你分享我們的長期視角。

So let me begin with the first part, covering our current business and the key metrics. So as you've seen, external revenue for the AI and Cloud segment has been accelerating now for nine consecutive quarters. And in this last quarter, growth has already accelerated to 5%. We're seeing very strong customer demand and our offerings boast a distinct competitive advantage compared to those of other cloud providers.

那我先從第一部分開始，介紹我們當前的業務與關鍵指標。如你所見，AI 與雲板塊的外部收入已連續九個季度加速增長。而在上一季度，增速已加快至 5%。我們看到客戶需求非常強勁，且相較其他雲服務商，我們的產品與服務具備明顯的競爭優勢。

As a result, we expect revenue growth to continue accelerating over the coming quarters. We've observed that AI-related products generated RMB12.4 billion in revenue this quarter. And so if we convert that into an annualized US dollar figure, that works out to USD7.3 billion in annual revenue.

因此，我們預期未來幾個季度收入增長將持續加速。我們觀察到，本季度 AI 相關產品帶來了人民幣 124 億元的收入。若換算為年化的美元口徑，約相當於 73 億美元的年化收入。

Looking ahead to the next quarter, our own forecast is that, that same annualized revenue for AI quarters next quarter will approach USD10 billion. So our growth rate remains exceptionally strong. At the same time, we also expect our EBITDA margin to improve quarter by quarter sequentially over the next few quarters.

展望下一季，我們自己的預測是，AI 相關季度的同口徑年化營收在下一季將接近 100 億美元。因此，我們的成長率仍然異常強勁。同時，我們也預期在接下來幾季，我們的 EBITDA 利潤率將逐季環比改善。

Additionally, something very important in respect of the cloud business is growth in demand for MaaS. We have seen very significant growth in demand for MaaS this quarter, coupled with ongoing improvement in inference efficiency. The ARR of our MaaS business has now surpassed RMB16 billion. Actually, let me clarify, that is the latest data as of August. It has already surpassed RMB16 billion.

此外，就雲端業務而言，有一件非常重要的事情是對 MaaS 的需求成長。本季我們看到 MaaS 需求非常顯著的成長，同時推理效率也持續提升。我們 MaaS 業務的 ARR 現已超過人民幣 160 億元。其實我澄清一下，這是截至 8 月的最新數據。已經超過人民幣 160 億元。

Next, let me expand on the growth drivers within our business model. So it's important to understand that Alibaba's investment model for AI is fundamentally different from that pure-play AI companies. We are pursuing an intensive strategy across the full stock including chips, including cloud infrastructure, and including models. And we maintain a leading position in the industry across all three of those most critical domains. Moreover, we believe that the development of AI and technology across the industry is still in its early stages.

接下來，我想進一步說明我們商業模式中的成長驅動因素。因此，理解阿里巴巴在 AI 上的投資模式，從根本上不同於那些純 AI 公司，這點很重要。我們在全栈上採取密集型策略，包括晶片、雲端基礎設施，以及模型。而且在這三個最關鍵的領域，我們都維持產業領先地位。此外，我們認為整個產業的 AI 與科技發展仍處於早期階段。

Looking forward, different stages of technological development. The core commercial value within the AI industry may shift across different layers, including chips, cloud computing models, and applications. Our full stack investments ensure that we can deliver optimal service capabilities and the best value for money positioning us favorably in the industry going forward and ensuring that within each stage of technological development, it's possible for us to maintain competitiveness and sustained growth momentum.

展望未來，在技術發展的不同階段，AI 產業的核心商業價值可能會在不同層次之間轉移，包括晶片、雲端運算、模型與應用。我們的全栈投資確保我們能提供最佳的服務能力與最佳性價比，使我們在未來產業競爭中處於有利位置，並確保在每一個技術發展階段，我們都有可能維持競爭力與持續的成長動能。

Next, let me look ahead to what we think is going to be the most important growth driver over the next one to two years in the short term. So we've seen exponential demand for commercial insurance services as of the end of 2025. This exponential growth in demand for inference has marked a fundamental shift in the model whereby compute has now become the core asset driving AI revenue.

接下來，我想展望一下我們認為在未來一到兩年短期內最重要的成長驅動因素。截至 2025 年底，我們看到商業保險服務的需求呈指數型成長。推理需求的這種指數型成長，標誌著模式上的根本轉變：算力如今已成為驅動 AI 營收的核心資產。

And today, all AI-related revenue models are centered on AI compute. And at the same time, there's a consensus across the industry, as I mentioned, that compute will remain in a shortage of supply for some time to come.

而今天，所有 AI 相關的營收模式都以 AI 算力為核心。同時，正如我提到的，產業內也有共識認為，算力在未來一段時間仍將供不應求。

At the same time, the higher gross margins of mass inference services have also made a major difference if compute was once a cost center, traditionally, compute has now been transformed into a core productive asset whose value generation is positively correlated with revenue.

同時，大規模推理服務較高的毛利率也帶來了重大改變：如果算力過去傳統上是一個成本中心，那麼如今算力已轉變為核心生產性資產，其價值創造與營收呈正相關。

So high-priced computing power remains in short supply across the industry precisely at a time where you have widespread adoption of GPUs across diverse use cases. So pricing models are tending to converge on the most high margin, the most margin generative monetization approaches. So this is driving the pricing models for nearly all GPU-related products.

因此，在各類使用場景廣泛採用 GPU 的同時，產業內高價算力仍然供給短缺。因此，定價模式正趨於收斂到毛利最高、最能產生利潤的變現方式。這正在推動幾乎所有 GPU 相關產品的定價模式。

Moreover, Alibaba, both comprehensive multimodal model capabilities. Our models are state-of-the-art level within the industry, giving us a distinct advantage in realizing the value of that compute power and providing a robust anchor for our pricing strategy. So, when it comes time to price for new customers or to sign -- re-sign contracts with existing customers as they renew, we can adopt more healthy pricing models. And so we expect to see this as a very positive short-term driver for improving margin in the coming year plus.

此外，阿里巴巴具備全面的多模態模型能力。我們的模型在產業內屬於最先進水準，使我們在實現算力價值方面具有明顯優勢，並為我們的定價策略提供堅實的錨點。因此，無論是為新客戶定價，或是在現有客戶續約時簽署——重新簽署合約，我們都能採取更健康的定價模式。因此，我們預期這將成為未來一年以上改善利潤率的一個非常正面的短期驅動因素。

Next, let me talk about the scale effect and networking effects, which are very important long-term growth drivers in AI cloud. For the past couple of years, a lot of people have asked what is the super app for AI. And the answer to that is that the real super application is compute, cloud-based AI compute, because all of these different workloads need to run on a full stack of AI cloud compute, including training, inferencing AI software and agents requiring GPUs, CPUs, storage, databases, virtualization, as well as harness tools among others.

接下來，我想談談規模效應與網路效應，這些是 AI 雲端長期成長非常重要的驅動因素。過去幾年，很多人問 AI 的超級 App 是什麼。答案是，真正的超級應用其實是算力、是雲端化的 AI 算力，因為所有不同的工作負載都需要運行在完整的 AI 雲端算力全栈之上，包括訓練、推理、需要 GPU 的 AI 軟體與代理（agents），以及 CPU、儲存、資料庫、虛擬化，還有各類編排/調度工具等。

So AI cloud is like a super city in which workload is the residents and continually iterating full stock AI cloud services or the urban infrastructure, which in turn attracts more new residents and enhances the stickiness of the existing residents. So this is where you see an extremely powerful network effect and scale effect.

因此，AI 雲端就像一座超級城市：工作負載是居民，而持續迭代的全栈 AI 雲端服務則是城市基礎設施；這反過來會吸引更多新居民，並提升既有居民的黏著度。因此，這裡會出現極其強大的網路效應與規模效應。

Given that we operate the largest number of data centers across any Asian cloud provider, we benefit from the strongest economies of scale. At the same time, the large-scale deployment of our proprietary T-Head AI chips allows us to avoid the high price premiums associated with procuring expensive commercial GPUs and thus avoiding erosion of our gross margins. And with our state-of-the-art performance in our proprietary models, we possess strong pricing power for our compute resources.

鑑於我們在所有亞洲雲端供應商中運營的資料中心數量最多，我們受益於最強的規模經濟。同時，我們自研的平頭哥（T-Head）AI 晶片的大規模部署，使我們能避免採購昂貴商用 GPU 所帶來的高溢價，從而避免毛利率被侵蝕。再加上我們自研模型的最先進表現，我們對算力資源具備強大的定價能力。

So looking ahead from the perspective of industry. Development trends and our own product strength, the long-term revenue growth trend and margin expansion trend are exceptionally strong. And as a result, we're highly confident in our ability to achieve our goal of RMB100 billion in external cloud revenue by 2030. And we have good visibility into achieving gross margin of 20%.

因此，從產業發展趨勢以及我們自身產品實力的角度展望未來，長期營收成長趨勢與利潤率擴張趨勢都異常強勁。而且因此，我們對於在 2030 年達成人民幣 1,000 億元的外部雲端營收目標非常有信心。同時，我們對於實現 20% 的毛利率也有良好的可見度。

Yuan Liao, CITICS.

廖源，中信證券。

(interpreted) Congratulations on the strong quarterly results and especially the progress made in the AI sector. I have a follow-up question on the MaaS business. As Eddie mentioned earlier, ARR as of August has exceeded RMB16 billion. Last quarter, I believe you stated that the target for year-end is to surpass RMB30 billion in MaaS ARR. I am wondering, given the progress to date, do you anticipate making any adjustments to that year-end goal?

（口譯）恭喜本季業績表現強勁，尤其是在 AI 領域取得的進展。我有一個關於 MaaS 業務的追問。如 Eddie 先前提到，截至 8 月的 ARR 已超過人民幣 160 億元。上季我記得你們提到年末目標是 MaaS ARR 超過人民幣 300 億元。我想請問，基於目前的進度，你們是否預期會對這個年末目標做任何調整？

Additionally, within the MaaS business, what are the respective shares of our own proprietary models versus third-party models? As model-related competition intensifies and more open source models emerge, how will these factors possibly affect gross margin and profitability in the MaaS business?

另外，在 MaaS 業務中，我們自研模型與第三方模型的收入占比分別是多少？隨著模型相關競爭加劇、以及更多開源模型出現，這些因素可能會如何影響 MaaS 業務的毛利率與獲利能力？

(interpreted) Thank you for the question. Yes, indeed, growth in Bailian's MaaS business is very rapid. And in -- as of August, we reached RMB16 billion or surpassed RMB16 billion in ARR. So given the current growth momentum as well as the pipeline of new models slated for launch, we remain confident that we will achieve our year-end target of RMB30 billion ARR by the end of the year.

（口譯）謝謝你的提問。是的，百鍊的 MaaS 業務成長確實非常快。截至 8 月，我們的 ARR 已達到人民幣 160 億元或超過人民幣 160 億元。因此，考量目前的成長動能以及即將上線的新模型管線，我們仍然有信心在年底前達成 MaaS ARR 人民幣 300 億元的年末目標。

So on our MaaS platform, our own proprietary model still account for the majority of the revenue. But having said that, revenue from third-party models is also not small. And having said that, perhaps let me talk a little bit about how we see different model capabilities.

在我們的 MaaS 平台上，自研模型仍然貢獻了大部分收入。但即便如此，第三方模型帶來的收入也不小。另外，或許我也談一下我們如何看待不同的模型能力。

A lot of customers tend to need to use or want to use multiple different models in their own AI applications because those different models they can draw and have different characteristics or different capabilities. So having more open source models on platforms like ours like Bailian to provide inferencing is a good thing for us and for Bailian.

很多客戶在自己的 AI 應用中，往往需要使用或希望使用多個不同模型，因為不同模型各有特性或能力。因此，在像我們百鍊這樣的平台上有更多開源模型提供推理服務，對我們、對百鍊而言都是好事。

When it comes to gross margin, the level of gross margin that we can achieve on a platform like Bailian from hosting our own proprietary models versus third-party models is actually very similar. It's highly comparable. We're really developing those proprietary models on the one hand in order to keep creating higher levels of model intelligence and also as part of our ultimate drive to achieve AGI.

就毛利率而言，在像百鍊這樣的平台上，託管我們自研模型與託管第三方模型所能達到的毛利率水準其實非常相近。高度可比。我們一方面持續開發自研模型，是為了不斷提升模型智能水準，同時也是我們最終推動實現 AGI 的一部分。

But simply from the perspective of the mouse business, the level of gross margin from those two kinds of models is actually very comparable. But overall, having a prosperous and flourishing open ecosystem with many of these open-source models on it is highly favorable for a cloud provider like Alibaba Cloud.

但僅從滑鼠業務的角度來看，這兩種模式的毛利率水準其實非常接近。但整體而言，擁有一個繁榮且蓬勃的開放生態系，並在其上承載許多這類開源模型，對像阿里雲這樣的雲端服務供應商是非常有利的。

Alex Yao, JPMorgan.

Alex Yao，摩根大通。

(interpreted) I'd like to come back to Eddie's earlier remarks, he spoke at length about how Alibaba is developing a full stock AI ecosystem.

(口譯)我想回到Eddie先前的發言，他詳細談到阿里巴巴如何在打造一個全棧AI生態系。

My question really is in which layer of that full stack ecosystem do you think value will accrete and monetization will be concentrated? We saw just after it had been released for three months that you open sourced the weight of your flagship model Qwen 3.8 Max. At the same time, your proprietary chips are also proving successful, now serving over 600 external customers. I am wondering if this means that the future value will accrete mainly in the compute layer or perhaps in the orchestration layer and not necessarily in the model layer? Or do you think that value will accrete to different layers in different stages of development of the industry?

我的問題是，在這個全棧生態系的哪一層，您認為價值會累積、變現會更集中？我們看到在發布僅三個月後，你們就將旗艦模型Qwen 3.8 Max的權重開源。同時，你們的自研晶片也證明相當成功，目前已服務超過600家外部客戶。我想請問，這是否意味著未來的價值主要會累積在算力層，或可能在編排層，而不一定是在模型層？或者您認為，在產業不同發展階段，價值會在不同層次累積？

And in the long term, if you think that value and monetization will largely be concentrated in the hardware and compute layers, then how should we think about competition going forward given that it will be a government-led process for allocating a lot of that hardware and compute capacity?

從長期來看，如果您認為價值與變現將主要集中在硬體與算力層，那麼考量到大量硬體與算力產能的配置可能會是一個政府主導的過程，我們應該如何看待未來的競爭格局？

(interpreted) Thanks. That's a very professional question, and really, it's a matter of long-term judgment. So I think it's inherently associated with a high level of uncertainty. But what I can say is that we are investing in the full stack. And what that means is that whichever layer represents the greatest value and no matter how that may shift across layers in different periods of time. All of those layers are part of our ecosystem.

(口譯)謝謝。這是一個非常專業的問題，確實屬於長期判斷的範疇。因此我認為它本質上伴隨著高度不確定性。但我可以說的是，我們正在投資全棧。這意味著無論哪一層代表最大的價值，以及在不同時期價值如何在各層之間轉移，所有這些層次都是我們生態系的一部分。

I guess I can share with you my own short-term view namely in the short-term perspective, I think that most of the value will be in chips and in AI cloud infrastructure. It's a pattern that we can see not just in China but globally across a lot of different companies when a technology is in its early stages and especially when there's a shortage of supply. Lots of the value tends to be concentrated in the infrastructure and in the core hardware. In this case, chips and storage. So in Alibaba's case, we've integrated our compute power, our cloud infrastructure and our AI inference into one core business segment.

我想我可以分享我自己的短期看法：從短期角度來看，我認為大部分價值會在晶片與AI雲端基礎設施上。當一項技術處於早期階段、尤其在供給短缺時，這種模式不僅在中國，在全球許多公司身上都能看到。大量價值往往會集中在基礎設施與核心硬體上。在這個案例中，就是晶片與儲存。因此在阿里巴巴的情況下，我們已將算力、雲端基礎設施與AI推理整合為一個核心業務板塊。

Let me turn next to where the ultimate commercial value will be realized from these models. It's a question around which there's a lot of debate within the industry and indeed, there are different views even inside our own company. So here, I'm just sharing my own personal opinion, but in my personal view, I think that the current monetization model for large language models through APIs is just a short-term approach, a short-term transitional approach and is certainly not the ultimate business model.

接下來我想談談這些模型最終的商業價值將在哪裡實現。這是業界爭論很多的問題，事實上即便在我們公司內部也有不同觀點。因此我在此僅分享個人看法：我認為目前透過API對大型語言模型進行變現的模式，只是一種短期做法、一種短期的過渡性做法，當然也不會是最終的商業模式。

Our company has invested a tremendous amount of compute across our entire platform, but the objective is not simply to be able to generate that kind of short-term API revenue. I think when we get to the stage where we've accomplished AGI or we're close to achieving AGI at that point, the ultimate business model will be delivering actual products, delivering actual results that clients are looking for. It will be conducting the actual R&D that delivers products and that delivers operations.

我們公司在整個平台投入了大量算力，但目標並不只是為了產生那種短期的API收入。我認為當我們達到AGI或接近實現AGI的階段時，最終的商業模式將是交付真正的產品、交付客戶所追求的實際成果。也就是進行真正能產出產品、能支撐營運的研發工作。

So the reason that all these different AI model companies are investing so heavily and engaging in an arms race today is not simply to be able to compete to provide that API-based service. It's because they have to eyes on that ultimate end game where I think that the monetization level will be significantly higher, be much higher than what you see today selling the service through API calls.

因此，當下所有不同的AI模型公司之所以投入如此之重、並展開軍備競賽，並不只是為了競爭提供基於API的服務。而是因為他們都把目光放在我認為最終的終局：屆時的變現水準將顯著更高，遠高於今天透過API呼叫來銷售服務所能看到的水準。

In terms of hardware, I'd like to add a few thoughts regarding our T-Head proprietary chips. I know it's a topic about which we haven't communicated a lot with investors in the past, but the last generation of T-Head chips we've already manufactured over 500,000 of them and shipped. And then the latest generation in August has already been deployed on AI -- Alibaba's AI cloud as supernodes. And I think we're one of the only companies that's able to deploy such proprietary chips, domestic chips at scale.

在硬體方面，我想補充幾點關於我們平頭哥（T-Head）自研晶片的看法。我知道這是我們過去與投資人溝通不多的主題，但上一代平頭哥晶片我們已經量產超過50萬顆並完成出貨。而最新一代在8月已部署於AI——阿里巴巴的AI雲端上，作為超級節點。我認為我們是少數能夠以規模化方式部署此類自研晶片、國產晶片的公司之一。

One thing that's really unique about our T-Head chips, domestically manufactured chips, is that they are designed with GPU architecture as their core technical foundation, and they can very well support both training and inference workloads. There are now already several hundreds of companies that are leveraging these chips via Alibaba Cloud for both inference as well as for model training.

我們平頭哥晶片（國產製造晶片）非常獨特的一點是：其設計以GPU架構作為核心技術基礎，能很好地同時支援訓練與推理工作負載。目前已經有數百家公司透過阿里雲使用這些晶片，既用於推理，也用於模型訓練。

And these span companies across Embodied AI, autonomous driving, as well as large model companies. In terms of our generation two of chips, we are going to start developing them in the second half of this year. We expect them to boast exceptionally high compute power as well as extremely robust interconnection bandwidth, making them fully capable of serving as a direct replacement for existing chips.

這些公司涵蓋具身智能（Embodied AI）、自動駕駛，以及大模型公司等。至於我們第二代晶片，我們將在今年下半年開始研發。我們預期其將具備極高的算力以及極其強健的互連頻寬，使其完全有能力作為現有晶片的直接替代方案。

So I think we're in a really, really unique position in the chip sector, especially when it comes to large scale model training. So I don't think that there's any government-led compute supply allocation scheme that could produce chips with such truly strong competitiveness.

因此我認為我們在晶片領域處於非常、非常獨特的位置，尤其是在大規模模型訓練方面。所以我不認為任何政府主導的算力供給配置方案，能夠產出具備如此強競爭力的晶片。

So I think that the T-Head's future is highly certain as a very key and core component of Alibaba Cloud, and we remain highly confident in our core competitive strength in this area. I've interacted with a lot of different engineers across China. And I can tell you that these chips have a very broad audience with engineers across a wide range of different engineering domains.

因此我認為，平頭哥作為阿里雲非常關鍵且核心的組成部分，其未來具有高度確定性，我們也對在此領域的核心競爭力保持高度信心。我與中國各地許多不同的工程師交流過。我可以告訴你，這些晶片在廣泛的工程領域中，擁有非常廣大的工程師受眾。

So to sum up, I think that our T-Head chips are definitely the best among domestic Chinese chips for supporting both training and inference across a wide range of different industries. We really are number one in the industry.

總結來說，我認為我們的平頭哥晶片在國產中國晶片中，對於支援各行各業的訓練與推理，絕對是最好的。我們確實是業界第一。

Then I think in terms of future production capacity and deployment, we can confidently claim to be at least one of the top two. But in terms of our ability to actually reach customers with AI chips, Alibaba Cloud is the largest player by market share in China's cloud and AI market. I think we have a very strong edge when it comes to channel distribution.

接著在未來產能與部署方面，我們也有信心至少能稱得上前二。但就將AI晶片真正觸達客戶的能力而言，阿里雲在中國雲端與AI市場的市占率是最大的玩家。我認為我們在通路分發方面具有非常強的優勢。

So from this perspective, I am highly confident in the long-term commercial value of T-Head chips.

因此從這個角度來看，我對平頭哥晶片的長期商業價值非常有信心。

Thank you very much. We appreciate your support, and we look forward to updating you on our progress next quarter. Thank you.

非常感謝。感謝各位的支持，我們期待在下個季度向各位更新我們的進展。謝謝。

Thank you. That does conclude our conference for today. Thank you for participating. You may now disconnect.

謝謝。今天的電話會議到此結束。感謝各位參與。您現在可以斷線。

Portions of this transcript that are marked (interpreted) were spoken by an interpreter present on the live call. The interpreter was provided by the company sponsoring this event.

本逐字稿中標註為（口譯）的部分，為現場電話會議中由口譯員所轉述的內容。該口譯員由主辦本活動的公司提供。