Stock LLM Wiki

活動_Amazon_3Q26法說逐字稿_20261029

Amazon(AMZN)— 2026-10-29 3Q26 法說會逐字稿(英文原文+中譯)

日期:2026-10-29(美國時間盤後);與談人:CEO Andy Jassy、CFO Brian Olsavsky、IR VP Dave Fildes 逐字稿為 earnings call 音檔 AI 轉寫,英文原文與中譯並列;專有名詞與數字以公司 press release/10-Q 為準。中譯與英文原文不一致時以英文為準。

校正紀錄

本檔為 AI 語音轉寫,落地前已校正下列轉寫誤植;未列出者為原文。 - 「$104 billion annual revenue run rate」(Justin Post 問題之答覆,00:26:47)→ $184 billion(AWS 單季 460 億美元×4=1,840 億;同場 00:10:29、00:22:23、00:48:57 三處均為 $184 billion;中譯「1040 億」同步改為 1840 億) - 「AWS's Cura」(00:07:57,列舉 Continuum 整合進程式撰寫工具)→ Kiro(AWS 自家 coding 工具,同場 00:08:35、00:27:39 皆提 Kiro for coding;同音誤植) - 「Duck Labs」→ DuckDB Labs(DuckDB 收購案,專有名詞) - 「Inactive grocery customers shop nearly four times more…」→ Grocery customers…(與 00:41:42「grocery customers shop about four times more in non-grocery categories」同句對照,Inactive 為轉寫贅字) - 「Bedrock Day workloads」→ Bedrock workloads(Ross Sandler 問題之答覆) - 00:41:04「approximately half of all units sold within everyday essentials」→「approximately half of all units sold」(與 00:13:21「fresh groceries and everyday essentials … approximately half of all units sold」矛盾,以後者與國泰 call memo 口徑為準;中譯同步) - 00:16:25 中譯「高個位數」→ 高十位數(原文為 high teens,廣告 4Q26 YoY 回升;中譯誤,已改) - 財務交叉驗算:北美 87 億+國際 16 億+AWS 178 億=281 億,與合併營業利益相符;營益率 281/2,019=13.9%;AWS 178/460=38.7%;北美 87/1,129=7.7%、國際 16/431=3.7%;分部營收 1,129+431+460=2,020 億,與 2,019 億差異為四捨五入 - 保留原樣、待確認:00:06:54「Bedrock-managed versions of Kimi, GLM, and SpaceX models」中「SpaceX models」無法由脈絡判定(可能指 SpaceX/xAI 合併後之模型),未改動;00:07:25「AWS Continuum」為產品名,未改動 - 保留原樣:Jassy 於 00:19:48 說明 Prime Day 移轉對整體 YoY 影響約 400bps,廣告單獨影響 850bps,兩者口徑不同非矛盾

Presentation

Operator - Operator

(00:00:00)Thank you for standing by. Good day, everyone, and welcome to the Amazon.com Q3 2026 financial results teleconference. At this time, all participants are in listen-only mode. After the presentation, we will conduct a question-and-answer session. Today's call is being recorded. For opening remarks, I will turn the call over to the Vice President of Investor Relations, Mr. Dave Fildes. Thank you, sir. Please go ahead.

(00:00:00)感謝您的耐心等待。大家好,歡迎參加亞馬遜網路公司 2026 年第三季財務結果電話會議。此時,所有參與者均為聆聽模式。在簡報結束後,我們將進行問答環節。今天的通話正在錄音。現在,我將把通話交給投資者關係副總裁,戴夫·菲爾德斯先生。謝謝您,先生。請繼續。

Dave Fildes - Amazon, Investor Relation

(00:00:28)Hello, and welcome to our Q3 2026 Financial Results Conference Call. Joining us today to answer your questions are Andy Jassy, our CEO, and Brian Olsavsky, our CFO. As you listen to today's conference call, we encourage you to have our press release in front of you, which includes our financial results, metrics, and commentary on the quarter. Please note that, unless otherwise stated, all comparisons in this call will be against our results for the comparable period in 2025. Our comments and responses to your questions reflect management's views as of today, October 29th, 2026, and include forward-looking statements.

(00:00:28)大家好,歡迎參加我們的 2026 年第三季財務結果電話會議。今天與我們一起回答您問題的有我們的首席執行官安迪·賈西和首席財務官布萊恩·奧爾薩夫斯基。在您收聽今天的電話會議時,我們鼓勵您手邊有我們的新聞稿,其中包括我們的財務結果、指標和對本季度的評論。請注意,除非另有說明,否則本次通話中的所有比較將與 2025 年可比期間的結果進行對比。我們的評論和對您問題的回答反映了管理層截至今天,即 2026 年 10 月 29 日的觀點,並包括前瞻性聲明。

Dave Fildes - Amazon, Investor Relation

(00:01:03)Actual results may differ materially. Additional information about factors that could impact our financial results is included in today's press release and our filings with the SEC, including our most recent annual report on Form 10-K and subsequent filings. During this call, we may discuss certain non-GAAP financial measures. In our press release, slides accompanying this webcast, and our SEC filings—each posted on our Investor Relations website—you will find additional disclosures regarding these non-GAAP measures, including reconciliations to comparable GAAP measures.

(00:01:03)實際結果可能會有重大差異。有關可能影響我們財務結果的因素的更多信息已包含在今天的新聞稿和我們向美國證券交易委員會(SEC)提交的文件中,包括我們最近的 10-K 表格年報及後續文件。在本次通話中,我們可能會討論某些非公認會計原則(non-GAAP)財務指標。在我們的新聞稿、隨本次網絡直播的幻燈片以及我們的 SEC 文件中——這些都已在我們的投資者關係網站上發布——您將找到有關這些非 GAAP 指標的更多披露,包括與可比 GAAP 指標的調整。

Dave Fildes - Amazon, Investor Relation

(00:01:35)Our guidance incorporates the order trends we've seen to date and what we believe are appropriate assumptions today. Our results are inherently unpredictable and may be materially affected by many factors, including fluctuations in foreign exchange rates and energy prices,

(00:01:35)我們的指導假設包括我們迄今為止所見的訂單趨勢以及我們今天認為合適的假設。我們的結果本質上是不可預測的,可能會受到許多因素的重大影響,包括外匯匯率和能源價格的波動,

Dave Fildes - Amazon, Investor Relation

(00:01:49)changes in global economic and geopolitical conditions, tariff and trade policies, resource and supply volatility—including for memory chips—and customer demand and spending, including the impact of recessionary fears, inflation, interest rates, regional labor market constraints, world events, the rate of growth of the internet, online commerce, cloud services, new and emerging technologies, and various factors detailed in our SEC filings. Our guidance assumes, among other things, that we do not conclude any additional business acquisitions, restructurings, or legal settlements.

(00:01:49)全球經濟和地緣政治條件的變化,關稅和貿易政策,資源和供應的波動——包括記憶體晶片——以及客戶需求和支出,包括衰退恐懼、通脹、利率、地區勞動市場限制、世界事件、互聯網增長速度、在線商務、雲服務、新興技術以及我們 SEC 文件中詳細列出的各種因素。我們的指導假設包括,除其他事項外,我們不會完成任何額外的商業收購、重組或法律和解。

Dave Fildes - Amazon, Investor Relation

(00:02:21)It is not possible to accurately predict demand for our goods and services; therefore, our actual results could differ materially from our guidance. And now, I'll turn the call over to Andy.

(00:02:21)準確預測我們商品和服務的需求是不可能的;因此,我們的實際結果可能會與我們的指導有重大差異。現在,我將把通話交給安迪。

Andy Jassy - Amazon, CEO

(00:02:30)Thanks, Dave. We're reporting $201.9 billion in revenue, up 12% year over year. Operating income was $28.1 billion, up 61% year over year. Q3 was another strong quarter for Amazon. In AWS, our plan is working. Despite lapping a comparable year-over-year period where we added significant capacity, AWS grew 39.3% year over year, accelerating for the sixth straight quarter—our fastest growth in nearly five years. Backlog is $578 billion, growing triple digits year over year and well diversified. We're installing an AI accelerator about every 14 seconds, and almost as fast as we install capacity, customers are consuming it.

(00:02:30)謝謝,戴夫。我們報告的收入為 2019 億美元,同比增長 12%。營業收入為 281 億美元,同比增長 61%。第三季度對亞馬遜來說又是強勁的一季。在 AWS 方面,我們的計劃正在奏效。儘管與去年同期相比,我們增加了顯著的產能,但 AWS 仍同比增長 39.3%,連續第六個季度加速增長——這是近五年來最快的增長。待處理訂單為 5780 億美元,同比增長三位數,且多樣化良好。我們每 14 秒安裝一個 AI 加速器,幾乎與我們安裝產能的速度相同,客戶正在消耗它。

Andy Jassy - Amazon, CEO

(00:03:14)Our core compute and storage revenue grew at their fastest rates in several years. Our operating margins have long been strong on our core offering and are rapidly improving on the AI side as well, well ahead of where they were on core at a similar stage of evolution. This is despite heightened component costs like memory and significant depreciation from the CapEx we're investing. We continue to accelerate our CapEx payback and have even more conviction than before that the investment we're making in AWS will yield a much larger revenue and free cash flow-generating business than we imagined a couple of years ago, with strong returns on invested capital.

(00:03:14)我們的核心計算和存儲收入以幾年來最快的速度增長。我們的營業利潤率在我們的核心產品上長期以來一直很強,並且在 AI 方面也在迅速改善,遠超過它們在類似發展階段的核心表現。儘管記憶體等元件成本上升以及我們投資的資本支出帶來的重大折舊,我們仍然保持強勁。我們繼續加速資本支出的回報,並比以往更有信心,我們在 AWS 的投資將產生比幾年前想像的更大收入和自由現金流的業務,並且資本回報率強勁。

Andy Jassy - Amazon, CEO

(00:03:51)We have a compelling flywheel spinning in AWS. Borrowing from Jim Collins' flywheel definition in Good to Great: if you do A, you almost can't help but do B; if you do B, you almost can't help but do C; and so on. In AWS, having leading functionality, security, and operational performance drives customers to run their apps and store their data in AWS, which drives AWS AI demand because customers prefer their production inference near their data and apps, which drives core AWS demand since AI uses core for tasks like reinforcement learning, post-training, and tool usage, which drives reinvestment back into our broader AWS functionality.

(00:03:51)我們在 AWS 中擁有一個引人注目的飛輪。借用吉姆·柯林斯在《從優秀到卓越》中的飛輪定義:如果你做 A,你幾乎無法不做 B;如果你做 B,你幾乎無法不做 C;依此類推。在 AWS 中,擁有領先的功能、安全性和運營性能驅使客戶在 AWS 上運行他們的應用程序和存儲數據,這推動了 AWS AI 的需求,因為客戶更喜歡將生產推理靠近他們的數據和應用程序,這推動了核心 AWS 的需求,因為 AI 在強化學習、後期訓練和工具使用等任務中使用核心,這又推動了對我們更廣泛的 AWS 功能的再投資。

Andy Jassy - Amazon, CEO

(00:04:33)This is a different model than primarily hosting GPUs or renting short-term data center capacity. This is a long-term platform strategy that broadly helps customers invent and evolve their experiences, operate stably, and build a reliable partnership they can grow for many years. It also yields very attractive long-term economics for us, regardless of supply scarcity. There are many examples illustrating our progress in AWS, but I'll share a few. On the core side, It's important to remember that the vast majority of global IT spend—about 85%—is still on-premises, and we're in a 10- to 20-year cycle where that will flip to the cloud.

(00:04:33)這是一種不同於主要託管 GPU 或租用短期數據中心容量的模式。這是一種長期平台戰略,廣泛幫助客戶創新和發展他們的體驗,穩定運營,並建立可靠的夥伴關係,讓他們能夠在多年內持續增長。無論供應稀缺與否,這也為我們帶來了非常有吸引力的長期經濟效益。有許多例子可以說明我們在 AWS 中的進展,但我將分享幾個。在核心方面,重要的是要記住,全球 IT 支出的絕大多數——約 85%——仍然是在本地,並且我們正處於一個 10 到 20 年的週期中,這將轉向雲端。

Andy Jassy - Amazon, CEO

(00:05:13)Conventional wisdom held that linchpin workloads like mainframes or storage area networks would never move to the cloud, but these workloads are moving. AWS has built unique cloud substitutes and AI-driven migration capabilities to make this transition much easier. Enterprises benefit from less CapEx, lower operating costs, faster innovation, better developer productivity, and improved data accessibility to more easily leverage AI. Our core business has attractive margins because we've learned to run it very efficiently over the last 20 years.

(00:05:13)傳統觀念認為,關鍵工作負載如大型主機或存儲區網絡永遠不會轉移到雲端,但這些工作負載正在轉移。AWS 已建立獨特的雲替代品和 AI 驅動的遷移能力,使這一過渡變得更加容易。企業受益於較少的資本支出、較低的運營成本、更快的創新、更好的開發者生產力和改善的數據可訪問性,以更輕鬆地利用 AI。我們的核心業務具有吸引人的利潤率,因為我們在過去 20 年中學會了非常高效地運營它。

Andy Jassy - Amazon, CEO

(00:05:46)As our core business grows, we continue reinvesting in both our core and AI functionality. For example, As customers generate increasingly more data and seek to derive value from it through analytics and AI, the complexity of moving data between data stores and analytics engines becomes onerous and costly. With our acquisition of DuckDB Labs, DuckDB will become the universal SQL dialect for AWS storage analytics, simplifying data processing with a modern analytics engine that is fast, easy to use, and ubiquitous, allowing customers to write SQL queries once, and then execute them using any AWS analytics service.

(00:05:46)隨著我們的核心業務增長,我們繼續在核心和 AI 功能上進行再投資。例如,隨著客戶產生越來越多的數據,並尋求通過分析和人工智慧從中獲取價值,數據在數據存儲和分析引擎之間移動的複雜性變得繁重且昂貴。隨著我們收購 DuckDB Labs,DuckDB 將成為 AWS 存儲分析的通用 SQL 方言,簡化數據處理,配備快速、易用且無處不在的現代分析引擎,讓客戶能夠一次編寫 SQL 查詢,然後使用任何 AWS 分析服務執行它們。

Andy Jassy - Amazon, CEO

(00:06:24)For the first time in analytics, customers will incur zero migration or portability costs between query engines. This is a significant advancement, and you can imagine how it will also help agents perform efficient querying. On the AI side, we continue to expand the capabilities of our inference service, Amazon Bedrock, which is experiencing remarkable growth, with Fortune 500 companies spending nearly seven times more on Bedrock year over year. Additionally, total Bedrock token volume has grown nearly 17 times year over year.

(00:06:24)在分析領域,客戶將首次在查詢引擎之間不產生任何遷移或可攜性成本。這是一項重大進展,您可以想像這將如何幫助代理執行高效的查詢。在人工智慧方面,我們繼續擴展我們的推理服務 Amazon Bedrock 的能力,該服務正在經歷顯著增長,財富 500 強公司在 Bedrock 上的支出年增長近七倍。此外,Bedrock 的總代幣交易量年增長近 17 倍。

Andy Jassy - Amazon, CEO

(00:06:54)Bedrock is the fastest-growing multibillion-dollar service in AWS's history. We recently added OpenAI models to Bedrock, and usage is increasing rapidly. We also recently introduced Bedrock-managed versions of Kimi, GLM, and SpaceX models. You cannot discuss AI today without addressing security, which is another reason customers continue to choose Bedrock. We have designed Bedrock so that customers' data remains secure and private. Inputs and outputs are never shared with model providers or used to train base models.

(00:06:54)Bedrock 是 AWS 歷史上增長最快的數十億美元服務。我們最近將 OpenAI 模型添加到 Bedrock 中,使用量正在迅速增加。我們還最近推出了 Bedrock 管理的 Kimi、GLM 和 SpaceX 模型版本。今天談論人工智慧時,您無法不提及安全性,這也是客戶繼續選擇 Bedrock 的另一個原因。我們設計 Bedrock 使客戶的數據保持安全和私密。輸入和輸出從不與模型提供者共享,也不會用於訓練基礎模型。

Andy Jassy - Amazon, CEO

(00:07:25)When customers fine-tune a foundation model, we create a private copy specifically for their use, with data fully encrypted both in transit and at rest. As frontier models become more powerful, their servicing software vulnerabilities emerge at unprecedented rates and were previously undetected. We built AWS Continuum to discover, prioritize, validate, and remediate code vulnerabilities at machine speed. It is already being used by customers such as Capital One, IBM, Infor, Luminate, and Robinhood.

(00:07:25)當客戶微調基礎模型時,我們會為他們創建一個專用的私有副本,數據在傳輸和靜止時均完全加密。隨著前沿模型變得更強大,其服務軟件的漏洞以空前的速度出現,之前未被檢測到。我們建立了 AWS Continuum,以機器速度發現、優先排序、驗證和修復代碼漏洞。它已經被 Capital One、IBM、Infor、Luminate 和 Robinhood 等客戶使用。

Andy Jassy - Amazon, CEO

(00:07:57)In Q3, we announced plans to integrate it into the tools where code is written, including Anthropic's Claude Code, OpenAI's Codex, and AWS's Kiro. Customer interest has exceeded anything we've seen before for a new security service. Agents continue to be deployed at accelerating rates, many of which are built on Bedrock AgentCore's building blocks. Tens of thousands of customers currently use AgentCore, deploying twice as many agents with it in Q3 as in the prior three quarters combined. In addition to building their own purpose-built agents, most companies also use turnkey agentic services, several of which our customers love.

(00:07:57)在第三季度,我們宣佈計劃將其整合到編寫代碼的工具中,包括 Anthropic 的 Claude Code、OpenAI 的 Codex 和 AWS 的 Kiro。客戶的興趣超過了我們之前對新安全服務的任何預期。代理的部署速度持續加快,其中許多是基於 Bedrock AgentCore 的構建塊。目前有數萬名客戶使用 AgentCore,在第三季度部署的代理數量是前三個季度的兩倍。除了構建自己的專用代理外,大多數公司還使用即用型代理服務,其中幾個是我們的客戶非常喜愛的。

Andy Jassy - Amazon, CEO

(00:08:35)These include Kiro for coding, Amazon Connect for contact center agents, and Amazon Transform for migrations. All are iterating quickly and gaining customer momentum. Similarly, Amazon Q, our intelligent AI work companion, is growing rapidly as large enterprises like Mondelez, EchoStar, 3M, Southwest Airlines, and DXC choose Q because it is purpose-built for work without compromising customer data. Finally, our chips business continues to grow at triple-digit percentages year over year, with revenue commitments for Graviton and Trainium each increasing more than 20 times year over year.

(00:08:35)這些包括用於編碼的 Kiro、用於聯絡中心代理的 Amazon Connect 和用於遷移的 Amazon Transform。所有這些都在快速迭代並獲得客戶的支持。同樣,我們的智能 AI 工作夥伴 Amazon Q 正在迅速增長,像 Mondelez、EchoStar、3M、西南航空和 DXC 等大型企業選擇 Q,因為它是專為工作而設計的,並且不會妨礙客戶數據的安全。最後,我們的晶片業務持續以三位數的百分比增長,Graviton 和 Trainium 的收入承諾年增長均超過 20 倍。

Andy Jassy - Amazon, CEO

(00:09:11)AI remains a tailwind for Graviton because post-training reinforcement learning and agent tool use are mostly done on CPUs rather than AI accelerators, and Graviton is the strongest CPU chip, offering up to 30% to 40% better price performance than alternatives. Customers using our custom silicon saved more than $2 billion in Q3 alone, and we have just open-sourced Torch Neuron so developers can run existing PyTorch code, which most already use, on Trainium without rewriting anything. As we launch new generations of chips, demand for prior generations remains strong.

(00:09:11)人工智慧對 Graviton 仍然是一個助力,因為後訓練強化學習和代理工具的使用主要是在 CPU 上進行,而不是在 AI 加速器上,Graviton 是最強的 CPU 晶片,提供比替代品高出 30% 到 40% 的價格性能。使用我們定製硅晶片的客戶在第三季度單獨節省了超過 20 億美元,我們剛剛開源了 Torch Neuron,讓開發者可以在 Trainium 上運行現有的 PyTorch 代碼,而無需重寫任何內容。隨著我們推出新一代晶片,對前幾代晶片的需求仍然強勁。

Andy Jassy - Amazon, CEO

(00:09:49)After our launch of Graviton 5, backlog for prior-generation Graviton instances more than doubled quarter over quarter. Short-term capacity for prior-generation training instances sells at a premium compared to the first half of the year. With training demand so strong, customers continue encouraging us to make it available outside of AWS. We intend to do so late in 2027. Overall demand for AWS capacity remains exceptionally strong. Enterprises are still very early in using inference at scale, yet inference already accounts for well over half of AWS's AI revenue and is growing, alongside significant training revenue.

(00:09:49)在我們推出 Graviton 5 之後,前一代 Graviton 實例的積壓在季度之間增長了兩倍以上。與上半年相比,前一代訓練實例的短期容量以溢價出售。隨著訓練需求如此強勁,客戶持續鼓勵我們在 AWS 之外提供這項服務。我們計劃在 2027 年底這樣做。整體對 AWS 容量的需求仍然異常強勁。企業在大規模使用推理方面仍然非常早期,但推理已經佔據了 AWS 人工智慧收入的一半以上,並且正在增長,與顯著的訓練收入一起增長。

Andy Jassy - Amazon, CEO

(00:10:29)We still expect 2026 cash capex of approximately $220 billion, the majority for AWS, and anticipate spending meaningfully more in 2027, with a clear line of sight to attractive financial returns as outlined on our last call. AWS is a $184 billion annualized revenue run-rate business, yet the opportunity remains extraordinary. With our continued pace of innovation, new core and AI functionality, enterprises wanting to run their core and AI workloads in AWS, and the capacity we add each year, we believe AWS can grow north of 40% in both 2027 and 2028 and become a trillion-dollar revenue business within the next 5 to 10 years.

(00:10:29)我們仍然預計 2026 年的現金資本支出約為 2200 億美元,其中大部分用於 AWS,並預計在 2027 年的支出將顯著增加,並且對於吸引財務回報有清晰的視野,正如我們上次電話會議中所述。AWS 是一個年化收入達 1840 億美元的業務,但機會仍然非常巨大。隨著我們持續的創新步伐、新的核心和人工智慧功能,企業希望在 AWS 上運行其核心和人工智慧工作負載,以及我們每年增加的容量,我們相信 AWS 在 2027 年和 2028 年都能增長超過 40%,並在未來 5 到 10 年內成為一個萬億美元的收入業務。

Andy Jassy - Amazon, CEO

(00:11:14)We are in the very early days of this inflection. Our strategy is working, and I expect it will continue to do so for the foreseeable future. In stores, I think some people underestimate how much progress that business has made and what it has become over the last few years. I am not aware of another retail business approaching Amazon's revenue size—hundreds of billions of dollars—with double-digit operating margins inclusive of ads, growing at this pace. Year to date, units have grown nearly 15% year over year, and that is on an enormous base, having sold more than 30 billion items last year.

(00:11:14)我們正處於這一轉折點的非常早期階段。我們的策略正在奏效,我預計在可預見的未來將繼續如此。在商店中,我認為有些人低估了這項業務在過去幾年中取得的進展及其所成為的樣子。我不知道有哪個零售業務能接近亞馬遜的收入規模——數千億美元——並且在包含廣告的情況下擁有雙位數的營業利潤率,並以這樣的速度增長。截至目前為止,單位數量年增長近 15%,而這是在一個龐大的基數上,去年售出了超過 300 億件商品。

Andy Jassy - Amazon, CEO

(00:11:47)In absolute dollar terms, we have added more than $90 billion in growth sales over the trailing 12 months, growing meaningfully faster than other retail competitors, adding about double the growth sales per quarter compared to our next largest U.S. competitor, and consistently gaining retail market share. Over the past few years, Our North American and international segments have each expanded operating margins by nearly 10 percentage points, enabling us to reinvest in improving the customer experience.

(00:11:47)在絕對美元方面,我們在過去 12 個月中增加了超過 900 億美元的增長銷售,增長速度明顯快於其他零售競爭對手,與我們的下一大美國競爭對手相比,每季度增加的增長銷售約為兩倍,並持續獲得零售市場份額。在過去幾年中,我們的北美和國際業務各自擴大了近 10 個百分點的營業利潤率,使我們能夠重新投資於改善客戶體驗。

Andy Jassy - Amazon, CEO

(00:12:17)Remember, it is still early. The significant majority of retail—nearly 80%—currently occurs in physical stores. That will eventually shift to e-commerce over time. We are helping enable this shift by rapidly building out our efficient and speedy fulfillment network, providing delightful customer experiences while driving significant free cash flow and return on invested capital. Each of our priorities—from faster delivery to cost to serve, groceries and pharmacy, and AI shopping—are force multipliers, and we are firing on all cylinders.

(00:12:17)記住,這仍然是早期階段。零售的絕大多數——近 80%——目前發生在實體商店中。隨著時間的推移,這最終將轉向電子商務。我們正在通過迅速建立高效且快速的履行網絡來幫助促進這一轉變,提供愉悅的客戶體驗,同時推動顯著的自由現金流和資本回報。我們的每個優先事項——從更快的交付到服務成本、雜貨和藥房,以及人工智慧購物——都是增強力量的因素,我們正在全力以赴。

Andy Jassy - Amazon, CEO

(00:12:50)Our delivery speeds continue to improve as we invest in our same-day network, as well as in our Amazon Now and Amazon Prime Air ultra-fast services. We had our fastest Q3 ever globally, delivering more than 1.5 billion items to Prime customers same day or overnight in Q3. We are energized by the customer response to our rapid rollout of Amazon Now worldwide. Amazon Now is now available in over 450 towns and cities, with the number of items ordered through the service up 10 times year over year.

(00:12:50)隨著我們對同日配送網絡的投資,我們的交付速度持續改善,以及對我們的 Amazon Now 和 Amazon Prime Air 超快速服務的投資。我們在全球的第三季度創下了最快的紀錄,為 Prime 客戶在同一天或隔夜交付了超過 15 億件商品。我們對客戶對我們在全球快速推出 Amazon Now 的反應感到振奮。Amazon Now 現在在超過 450 個城鎮和城市提供服務,通過該服務訂購的商品數量同比增長了 10 倍。

Andy Jassy - Amazon, CEO

(00:13:21)We are on track to reduce cost to serve per unit for the fourth consecutive year, enabling us to continue meeting or beating competitors' prices and cost-effectively deliver a broad selection of fresh groceries and everyday essentials. Increasing speed is fueling growth in our grocery business, including perishables and everyday essentials, with sales reaching nearly $250 billion in the trailing 12 months. In Q3... Fresh groceries and everyday essentials grew about three times faster than the rest of our store, representing approximately half of all units sold.

(00:13:21)我們正朝著連續第四年降低每單位服務成本的目標邁進,使我們能夠繼續滿足或超越競爭對手的價格,並以具成本效益的方式提供廣泛的新鮮雜貨和日常必需品。提高速度正在推動我們的雜貨業務增長,包括易腐爛商品和日常必需品,過去 12 個月的銷售額接近 2500 億美元。在第三季度……新鮮雜貨和日常必需品的增長速度約是我們其他商店的三倍,佔所有銷售單位的約一半。

Andy Jassy - Amazon, CEO

(00:13:55)Customers who shop perishables with same-day delivery purchase nearly seven times more everyday essential units than those who do not. Grocery customers shop nearly four times more in non-grocery categories like home and apparel and spend over twice as much on those categories compared to non-grocery customers. Amazon Pharmacy is also growing rapidly. The number of customers using Amazon Pharmacy is up over 60% year over year, with growth accelerating each quarter through 2026. Repeat customers grew more than 50% year over year.

(00:13:55)使用同日配送購買易腐爛商品的客戶,其日常必需品的購買量幾乎是未使用者的七倍。雜貨客戶在非雜貨類別(如家居和服裝)上的購買量幾乎是非雜貨客戶的四倍,並且在這些類別上的花費超過兩倍。Amazon Pharmacy 也在快速增長。使用 Amazon Pharmacy 的客戶數量同比增長超過 60%,並且每個季度的增長速度在 2026 年之前持續加快。重複客戶的增長率同比增長超過 50%。

Andy Jassy - Amazon, CEO

(00:14:26)J.D. Power recently ranked Amazon Pharmacy as the top mail-order pharmacy for the third consecutive year, and we continue to improve it, now offering three-hour prescription delivery in more than 350 cities. Over 400 million customers have used Alexa for shopping, our agentic shopping assistant, in the last 12 months, with customer interactions up 570% year over year. U.S. Customers using Alexa for shopping in the U.S. are spending 50% more per order than those who do not. Alexa researches products, compares options, checks a full year of price history, provides personalized gift recommendations, and tracks deals and price drops on customers' behalf.

(00:14:26)J.D.Power 最近將 Amazon Pharmacy 評為連續第三年最佳郵購藥房,我們也在不斷改進,現在在超過 350 個城市提供三小時處方配送。在過去的 12 個月中,超過 4 億客戶使用 Alexa 進行購物,我們的智能購物助手,客戶互動同比增長 570%。美國。在美國,使用 Alexa 進行購物的客戶每筆訂單的花費比不使用者多 50%。Alexa 會研究產品、比較選項、檢查一整年的價格歷史、提供個性化的禮物建議,並代表客戶跟蹤交易和價格下跌。

Andy Jassy - Amazon, CEO

(00:15:07)While we anticipate working with horizontal agents, we believe Alexa for Shopping will be the primary shopping agent customers use to shop on Amazon. Having a shopping assistant constantly by your side as you browse, get inspired, ask questions on product detail pages, and simply ask to buy something based on your prior purchases—with your history, shipping address, and payment info already saved—is uniquely helpful and convenient for customers. Moving to Amazon Ads, we generated $19.6 billion in revenue in Q3, up 12% year over year.

(00:15:07)雖然我們預期會與水平代理合作,我們相信 Alexa for Shopping 將是客戶在 Amazon 上購物的主要購物代理。在瀏覽時,擁有一個購物助手隨時在您身邊,獲得靈感、在產品詳細頁面上詢問問題,並根據您之前的購買要求購買某樣東西——您的歷史、送貨地址和付款信息已經保存——對客戶來說是獨特的幫助和便利。轉到 Amazon Ads,我們在第三季度創造了 196 億美元的收入,同比增長 12%。

Andy Jassy - Amazon, CEO

(00:15:43)Although there were some timing anomalies this past quarter, we remain pleased with the overall momentum across our ad offerings. I will mention three items impacting the growth rates. First, we shifted Prime Day into Q2 for most countries, moving significant sales and ad revenue that were in Q3 2025 into Q2 2026, which is a tailwind to our Q2 2026 rate and a headwind to Q3 2026. For advertising, that represented an 850 basis point sequential impact on Q3 year-over-year growth. Second, Q2 2026 included the first season of NBA on Prime, which last year's Q2 did not have, while Q3 does not include any NBA content.

(00:15:43)儘管在過去的季度中出現了一些時間異常,但我們對廣告產品的整體勢頭仍然感到滿意。我將提到三個影響增長率的因素。首先,我們將 Prime Day 移至第二季度,對於大多數國家來說,將原本在 2025 年第三季度的重大銷售和廣告收入移至 2026 年第二季度,這對我們 2026 年第二季度的增長率是有利的,但對 2026 年第三季度則是逆風。對於廣告來說,這代表了對第三季度同比增長的 850 個基點的順序影響。第二,2026 年第二季度包括了 NBA 在 Prime 的第一季,而去年的第二季度則沒有,而第三季度不包括任何 NBA 內容。

Andy Jassy - Amazon, CEO

(00:16:25)The NBA season starts in Q4. Third, in Q3 2025, we changed our search engine marketing programs to drive more traffic to Amazon search pages. This provides a better experience for customers, and we also monetized that traffic at a higher rate. That change helped accelerate advertising revenue growth starting a year ago, and we began lapping that impact in Q3. We expect year-over-year growth to return to the high teens in Q4. The natural question is whether agentic shopping from external horizontal agents is meaningfully impacting Amazon Ads.

(00:16:25)NBA 賽季在第四季度開始。第三,在 2025 年第三季度,我們改變了搜索引擎營銷計劃,以吸引更多流量到 Amazon 搜索頁面。這為客戶提供了更好的體驗,我們也以更高的比率將該流量變現。這一變化幫助加速了廣告收入的增長,從一年前開始,我們在第三季度開始重複這一影響。我們預計第四季度的同比增長將回到高十位數(high teens)。自然的問題是,外部水平代理的智能購物是否對 Amazon Ads 產生了實質性影響。

Andy Jassy - Amazon, CEO

(00:16:56)The answer is no. Few customers are shopping this way yet, as the experience is still not great nor more convenient than going directly to a retailer to research or place a one-click order. Q3 was full of innovation from our ads team that we expect will have impact in the coming months. This was headlined by Amazon Ads Agent, which unifies our full advertising offering under a single AI-first platform. Backed by trillions of shopping, browsing, and streaming signals, it supports advertisers at every step with insights, answers, and the ability to optimize and act on their behalf.

(00:16:56)答案是否定的。目前還沒有多少客戶以這種方式購物,因為這種體驗仍然不佳,也不比直接去零售商那裡研究或下單更方便。第三季度我們的廣告團隊充滿了創新,我們預期這些創新將在未來幾個月內產生影響。這一切以 Amazon Ads Agent 為首,該平台將我們的完整廣告產品統一在一個以 AI 為首的平台下。在數萬億的購物、瀏覽和串流信號的支持下,它在每一步都為廣告商提供洞察、答案,以及優化和代表他們行動的能力。

Andy Jassy - Amazon, CEO

(00:17:32)We also launched full-funnel campaigns in Amazon Ads Agent, integrating sponsored ads, display, video, and streaming TV into a single campaign. Advertisers provide initial inputs, including budget and duration, and Amazon's AI plans, executes, and continually optimizes the rest. We continue to see momentum in several other areas. I will mention just a few. Prime Video has emerged as a significant business with a revenue run rate in the tens of billions of dollars and is now profitable on its own.

(00:17:32)我們還在 Amazon Ads Agent 中推出了全漏斗廣告活動,將贊助廣告、展示廣告、視頻和串流電視整合到一個活動中。廣告商提供初步輸入,包括預算和持續時間,然後 Amazon 的 AI 計劃、執行並持續優化其餘部分。我們在其他幾個領域也持續看到勢頭。我將僅提及幾個。Prime Video 已經成為一個重要的業務,年收入運行率達到數百億美元,現在已經獨立盈利。

Andy Jassy - Amazon, CEO

(00:18:03)We have momentum with hit shows like Carrie, The Summer I Turned Pretty, The Boys, Off Campus, Reacher, and Fallout; breakout movies like Project Hail Mary; and compelling live sports, including Thursday Night Football, which is off to its best start ever in 2026, NBA on Prime, and NASCAR, the Masters, and UEFA Champions League Soccer. We recently launched our most powerful lineup of devices yet, featuring 34 new products across Echo, Kindle, Alexa tablets, Fire TV, Ring, and Blink. It's early, but we are off to a very strong start and excited for more customers to experience these products as we approach the holiday season.

(00:18:03)我們在熱門節目方面有勢頭,如《凱莉》、《我變漂亮的夏天》、《男孩們》、《校外》、《瑞奇》和《輻射》;突破性電影如《Hail Mary 計劃》;以及引人入勝的現場體育,包括《星期四夜足球》,在 2026 年的開局表現最佳,NBA 在 Prime 和 NASCAR,《大師賽》和《歐洲冠軍聯賽足球》。我們最近推出了我們迄今為止最強大的設備系列,包含 34 款新產品,涵蓋 Echo、Kindle、Alexa 平板、Fire TV、Ring 和 Blink。雖然還早,但我們已經有了一個非常強勁的開端,並期待更多客戶在假日季節體驗這些產品。

Andy Jassy - Amazon, CEO

(00:18:41)Finally, Amazon LEO remains on track to begin initial service this year. We are entering one of our most exciting periods of the year with both AWS re:Invent and the holidays approaching. I want to thank our teammates for their hard work and innovation in making customers' lives better and easier every day. What we are building at Amazon is unlike anything ever seen before. There has never been a business with a trillion dollars in annual revenue. Amazon is on a path to have two such businesses, in AWS and stores.

(00:18:41)最後,亞馬遜 LEO 仍然按計劃在今年開始初步服務。我們正進入一年中最令人興奮的時期,AWS re:Invent 和假日季節即將來臨。我想感謝我們的團隊成員,感謝他們的辛勤工作和創新,讓客戶的生活每天都變得更好、更輕鬆。我們在亞馬遜所建立的事物是前所未見的。從未有過年收入達到一兆美元的企業。亞馬遜正朝著擁有兩個這樣的業務邁進,分別是 AWS 和商店。

Andy Jassy - Amazon, CEO

(00:19:11)This is rare, and we are energized by what lies ahead. Over to you, Brian.

(00:19:11)這是罕見的,我們對未來的發展感到振奮。現在把時間交給你,Brian。

Brian Olsavsky - Amazon, CFO

(00:19:15)Thank you, Andy. Let's begin with our top-line financial results. Worldwide revenue was $201.9 billion, a 12% increase year over year, or 13% excluding the impact of foreign exchange. Worldwide operating income was $28.1 billion, with an operating margin of 13.9%, our highest ever. Starting with our North America and international segment results. First, let me remind you that the Prime Day sales event shifted this year. It occurred in Q2 for most of our largest countries, including the United States.

(00:19:15)謝謝你,Andy。讓我們開始討論我們的營收財務結果。全球收入為 2019 億美元,同比增長 12%,不計外匯影響則增長 13%。全球營業收入為 281 億美元,營業利潤率為 13.9%,創下歷史新高。首先,我們來看看北美和國際業務的結果。首先,讓我提醒您,今年的 Prime Day 銷售活動有所調整。它在第二季度舉行,涵蓋了我們大多數最大的國家,包括美國。

Brian Olsavsky - Amazon, CFO

(00:19:51)In 2025, Prime Day was entirely in Q3. Excluding this timing change, third quarter 2026 year-over-year growth would be approximately 400 basis points higher. In Q3, we saw strong customer engagement and a continuation of trends from the first half of the year. In the North America segment, third quarter revenue was $112.9 billion, a 6% increase year over year. International segment revenue was $43.1 billion, up 8% year over year, excluding foreign exchange impact. Worldwide paid units grew 12% year over year.

(00:19:51)在 2025 年,Prime Day 完全在第三季度進行。如果不考慮這一時間變化,2026 年第三季度的同比增長將大約高出 400 個基點。在第三季度,我們看到客戶參與度強勁,並延續了上半年趨勢。在北美業務中,第三季度收入為 1129 億美元,同比增長 6%。國際業務收入為 431 億美元,同比增長 8%,不計外匯影響。全球付費單位同比增長 12%。

Brian Olsavsky - Amazon, CFO

(00:20:31)Prime remains a key pillar of our business. Hundreds of millions of Prime members are responding to our combination of vast selection, low prices, and record delivery speeds. Over the past year, monthly orders of fresh groceries placed by Prime members for three same-day delivery have more than tripled. Let's shift to profitability. North America's segment operating income was $8.7 billion, with an operating margin of 7.7%. International segment operating income was $1.6 billion, with an operating margin of 3.7%.

(00:20:31)Prime 仍然是我們業務的關鍵支柱。數以億計的 Prime 會員正在響應我們的廣泛選擇、低價格和創紀錄的配送速度的組合。在過去一年中,Prime 會員下的當日送達新鮮食品的每月訂單數量已增長超過三倍。現在讓我們轉向盈利能力。北美業務的營業收入為 87 億美元,營業利潤率為 7.7%。國際業務的營業收入為 16 億美元,營業利潤率為 3.7%。

Brian Olsavsky - Amazon, CFO

(00:21:05)In our fulfillment network, the team managed higher transportation costs driven by fuel inflation and increased line haul rates. Despite these headwinds, we lowered our overall cost to serve in Q3 through better inventory placement, shorter shipping distances, and package consolidation. Shipping costs increased 11% year over year versus a 12% increase in units. Excluding fuel and line haul inflation, shipping costs grew 8%, approximately 400 basis points slower than unit growth. Robotics and AI are delivering measurable gains in productivity and employee safety, which we expect to compound over time.

(00:21:05)在我們的履行網絡中,團隊管理了更高的運輸成本這是由於燃料通脹和運輸費用上漲所驅動的。儘管面臨這些挑戰,我們通過更好的庫存配置、更短的運輸距離和包裹合併降低了第三季度的整體服務成本。運輸成本同比增長 11%,而單位增長則為 12%。不考慮燃料和運輸費用的通脹,運輸成本增長 8%,大約比單位增長慢 400 個基點。機器人技術和人工智慧在生產力和員工安全方面帶來了可衡量的增益,我們預計這一增益將隨時間累積。

Brian Olsavsky - Amazon, CFO

(00:21:46)In 2026, we doubled our fleet of robotic arms such as Cardinal and Sparrow. We are also developing robotics foundation models that enable faster programming and teach robots increasingly complex physical tasks, aiming to deliver useful robotics at scale. Looking ahead, we see meaningful opportunities to improve productivity across our global fulfillment network and reduce our cost to serve while continuing to deliver products to customers faster. Our operating margin may fluctuate quarter to quarter, but we remain focused on sustained long-term improvement in cost to serve per unit.

(00:21:46)在 2026 年,我們將機器手臂的數量翻倍,例如 Cardinal 和 Sparrow。我們還在開發機器人基礎模型,使編程更快,並教導機器人執行越來越複雜的物理任務,旨在大規模提供有用的機器人技術。展望未來,我們看到在全球履行網絡中改善生產力的重大機會,同時降低我們的服務成本,並繼續更快地向客戶交付產品。我們的營業利潤率可能會季度波動,但我們仍然專注於持續的長期改善每單位的服務成本。

Brian Olsavsky - Amazon, CFO

(00:22:23)Moving to our AWS segment, revenue was $46 billion, growing 39.3% year over year, driven by both core and AI services. We added $3.8 billion in quarter-over-quarter revenue. AWS now has an annualized revenue run rate of $184 billion. Customers continue to migrate to the cloud and scale their use of AWS core services. Increasingly, customers seeking the full benefits of AI are accelerating their transition to the cloud. We see a strong linkage between AI spending and growth in our revenue for core services.

(00:22:23)轉到我們的 AWS 業務,收入為 460 億美元,同比增長 39.3%,這得益於核心和人工智慧服務。我們在季度間增加了 38 億美元的收入。AWS 現在的年化收入運行率為 1840 億美元。客戶繼續向雲端遷移,並擴大對 AWS 核心服務的使用。越來越多的客戶尋求人工智慧的全部好處,正在加速向雲端的轉型。我們看到人工智慧支出與我們核心服務收入增長之間的強烈聯繫。

Brian Olsavsky - Amazon, CFO

(00:23:00)As customers invest in AI, we observe a corresponding increase in core consumption, and we expect this relationship to strengthen. This will occur as more AI workloads move into production, driving additional demand for compute, storage, and networking. AWS operating income was $17.8 billion, with an operating margin of 38.7%, reflecting strong growth and our focus on driving efficiencies. Our investments in custom chips, custom networking, software, and process improvements are helping us optimize capacity and build a more cost-efficient infrastructure.

(00:23:00)隨著客戶對人工智慧的投資,我們觀察到核心消費的相應增長,我們預計這一關係將加強。隨著更多的人工智慧工作負載進入生產,這將推動對計算、存儲和網絡的額外需求。AWS 的營業收入為 178 億美元,營業利潤率為 38.7%,反映出強勁的增長和我們對提高效率的專注。我們在定製晶片、定製網絡、軟件和流程改進方面的投資正在幫助我們優化產能,並建立更具成本效益的基礎設施。

Brian Olsavsky - Amazon, CFO

(00:23:36)As we have long stated, we expect AWS operating margins to fluctuate, partly driven by the level of investment we make at any given time. Our AI offerings are gaining strong traction with customers, with AI revenue growing triple digits year over year. Bedrock has been a significant driver of this growth. Now, turning to our cash capital expenditures, which were $75.9 billion in Q3. This primarily relates to AWS and AI as we invest to support strong customer demand. We continue to make significant investments, especially in AI, as we believe it represents a massive opportunity to drive long-term revenue, free cash flow, and return on invested capital.

(00:23:36)正如我們長期以來所說的,我們預期 AWS 的營運利潤率會波動,部分原因是我們在任何時候的投資水平。我們的 AI 產品在客戶中獲得了強勁的吸引力,AI 收入年增長率達到三位數。Bedrock 是這一增長的重要推動力。現在,轉向我們的現金資本支出,第三季度為 759 億美元。這主要與 AWS 和 AI 有關,因為我們正在投資以支持強勁的客戶需求。我們繼續進行重大投資,特別是在 AI 方面,因為我們相信這代表著推動長期收入、自由現金流和資本回報的巨大機會。

Brian Olsavsky - Amazon, CFO

(00:24:18)I'll conclude with our financial guidance for Q4. Q4 net sales are expected to range between $240 billion and $247 billion. We estimate the year-over-year impact of changes in foreign exchange rates based on current rates, which we expect to be a headwind of approximately 60 basis points for the quarter. Q4 operating income is expected to be between $28.5 billion and $33.5 billion. I'll mention one factor impacting our Q4 guidance. Amazon LEO's commercial service is on track to begin initial service this year, and we expect to start capitalizing certain costs in Q1 of next year, including production and launch expenses.

(00:24:18)我將以我們對第四季度的財務指引作結。第四季度的淨銷售預計在 2400 億到 2470 億美元之間。我們根據當前匯率估計外匯變動的年增影響,我們預計這將對本季度造成約 60 個基點的逆風。第四季度的營運收入預計在 285 億到 335 億美元之間。我將提到一個影響我們第四季度指引的因素。亞馬遜 LEO 的商業服務預計將在今年開始初步服務,我們預計在明年第一季度開始資本化某些成本,包括生產和發射費用。

Brian Olsavsky - Amazon, CFO

(00:25:02)This timing is later than I had predicted in earlier earnings calls. As a result, we will continue to expense the majority of LEO-related costs in Q4. We're ready to serve customers this holiday season and support the millions building with AWS. I want to thank our teams across Amazon for their hard work and customer focus. Our commitment to elevating the customer experience remains the only reliable way to create lasting value for our shareholders. With that, let's move on to your questions.

(00:25:02)這一時間比我在早期的財報電話會議中預測的要晚。因此,我們將繼續在第四季度計入大部分 LEO 相關成本。我們已準備好在這個假日季節為客戶提供服務,並支持數百萬使用 AWS 的客戶。我想感謝亞馬遜各團隊的辛勤工作和對客戶的關注。我們對提升客戶體驗的承諾仍然是為我們的股東創造持久價值的唯一可靠方式。有鑑於此,讓我們進入你的問題。

Operator - Operator

(00:25:36)Thank you. At this time, we will open the call for questions. We ask each caller to please limit themselves to one question. If you would like to ask a question, please press star 1 on your keypad. We ask that when you pose your question, you pick up your handset to provide optimum sound quality. Once again, to initiate a question, please press star then 1 on your touchtone telephone at this time. Please hold while we poll for questions. Thank you. Our first question comes from Justin Post with Bank of America.

(00:25:36)謝謝。此時,我們將開放電話以接受問題。我們請每位來電者限制自己提出一個問題。如果您想提出問題,請在您的鍵盤上按星號 1。我們請您在提出問題時拿起話筒,以提供最佳的音質。再次提醒,要開始提問,請在此時按星號然後按 1。請稍候,我們將進行問題調查。謝謝。我們的第一個問題來自美國銀行的 Justin Post。

Operator - Operator

(00:26:18)Please proceed with your question.

(00:26:18)請繼續提出您的問題。

Justin Post - Bank of America, Analyst

(00:26:22)Great. Thank you.

(00:26:22)很好。謝謝。

Q&A Justin Post - Bank of America, Analyst

(00:26:23)Andy, we see the cloud business accelerating nicely, but quarter-over-quarter growth slowed. I'm wondering if there are any specific factors on the customer side or token costs that drove this, and how you are thinking about capacity additions in Q4? And Brian, could you comment on any impact of fuel costs on your Q4 guidance? Thank you.

(00:26:23)Andy,我們看到雲業務增長良好,但季度間增長放緩。我想知道是否有任何特定的客戶因素或代幣成本驅動了這一點,以及您對第四季度的產能增加有何看法?Brian,您能否評論燃料成本對您第四季度指引的影響?謝謝。

Andy Jassy - Amazon, CEO

(00:26:47)I'll start with the growth question. We're very pleased with the growth. A 39.3% growth rate on a business with a $184 billion annual revenue run rate is quite remarkable. This marks our fastest growth in about five years, with six consecutive quarters of acceleration. And we're adding a significant amount of capacity. The growth is driven by a combination of factors. First, on the AI side, we offer an unusual top-to-bottom stack, whether you're building your own model with SageMaker AI, which is the easiest way to do so, or leveraging one of the leading frontier models with high security and strong price performance through Bedrock.

(00:26:47)我將從增長問題開始。我們對增長感到非常高興。在年化收入達到 1840 億美元的業務中,39.3% 的增長率相當驚人。這標誌著我們五年來最快的增長,並且連續六個季度加速增長。而且我們正在增加大量的產能。增長是由多種因素驅動的。首先,在 AI 方面,我們提供了一個不尋常的從上到下的堆疊,無論您是使用 SageMaker AI 構建自己的模型,這是最簡單的方式,還是通過 Bedrock 利用一個具有高安全性和強大價格性能的領先前沿模型。

Andy Jassy - Amazon, CEO

(00:27:39)Whether building your own agents using capabilities like Strands and Agent Core to build and run them securely and scalably, or using turnkey agents, which most companies do, such as Kiro for coding and Connect for call centers, and Transform for migrations, plus Amazon Q, which is growing rapidly as an AI companion at work, people are adopting these solutions quickly. Additionally, we add a lot of capacity every quarter. However, the amount of capacity added each quarter varies. We add different amounts of capacity each quarter depending on demand and other factors.

(00:27:39)無論是使用 Strands 和 Agent Core 等功能構建自己的代理,安全且可擴展地運行它們,還是使用大多數公司所做的即用型代理,例如用於編碼的 Kiro 和用於呼叫中心的 Connect,以及用於遷移的 Transform,加上迅速增長的 Amazon Q 作為工作中的 AI 伴侶,人們正在迅速採用這些解決方案。此外,我們每個季度都增加大量的產能。然而,每個季度增加的產能數量有所不同。我們根據需求和其他因素每個季度增加不同數量的產能。

Andy Jassy - Amazon, CEO

(00:28:21)But if I look at how much we've added this year and what we have planned over the next, I'll call it, eight quarters or so, which is about as far out as we get—eight to ten quarters. We have a lot of capacity coming as well. And then I layer on top of that the impact of chips. We're very well positioned for what's happening in AI because increasingly, what you're finding right now is that most customers—every customer that uses AI at scale, particularly enterprises— as they're starting to deploy more inference into production, are very eager to save on price performance.

(00:28:21)但是如果我看看我們今年增加了多少,以及我們接下來計劃的內容,我會稱之為,大約八個季度,這是我們能預測的最遠時間——八到十個季度。我們還有很多產能即將到來。然後我在此基礎上加上晶片的影響。我們在人工智慧方面的定位非常好,因為越來越多的客戶——每一位大規模使用人工智慧的客戶,特別是企業——在他們開始將更多推理部署到生產中時,都非常渴望在價格性能上節省開支。

Andy Jassy - Amazon, CEO

(00:29:01)Training with the meaningful advantage it has in price performance versus other alternatives is very compelling to people. And as more people use AI broadly, it drives many core capabilities like reinforcement learning, post-training, and tool calling. We have the leading CPU chip in Graviton, which offers about 30 to 40% better price performance than other x86 providers. These factors are driving very significant growth. The amount of capacity we add each quarter will vary, but at this scale, projecting growth over 40% year over year in both 2027 and 2028 is quite significant.

(00:29:01)在價格性能上與其他選擇相比,訓練所帶來的顯著優勢對人們來說非常有吸引力。隨著越來越多的人廣泛使用人工智慧,它推動了許多核心能力,如強化學習、後訓練和工具調用。我們擁有領先的 CPU 晶片 Graviton,其價格性能比其他 x86 供應商高出約 30% 到 40%。這些因素正在推動非常顯著的增長。我們每季度增加的產能會有所變化,但在這個規模下,預測 2027 年和 2028 年年增長超過 40% 是相當顯著的。

Andy Jassy - Amazon, CEO

(00:29:47)We believe this can be a trillion-dollar revenue business in five to ten years—an extraordinary and unusual opportunity we're very excited to pursue.

(00:29:47)我們相信這在五到十年內可以成為一個萬億美元的收入業務——這是一個非凡且不尋常的機會,我們非常期待追求。

Brian Olsavsky - Amazon, CFO

(00:29:59)Justin, regarding your question about fuel and line haul inflation, I don't have an exact number, but it is higher in Q4 than in Q3. Remember, we instituted a fuel surcharge at the beginning of the year for FBA sellers that recovers some, but not all, of the fuel costs for those shipments. The remaining costs are on retail inbound and outbound shipments. So, it's included there. It's slightly higher than in Q3.

(00:29:59)Justin,關於你提到的燃料和運輸通脹的問題,我沒有確切的數字,但在第四季度比第三季度高。記住,我們在年初為 FBA 賣家實施了燃料附加費,這回收了一部分,但不是全部的運輸成本。剩餘的成本則在零售的進貨和出貨中。所以,這些成本包含在內。它比第三季度略高。

Operator - Operator

(00:30:33)Thank you. Our next question comes from Brian Nowak with Morgan Stanley. Please proceed with your question.

(00:30:33)謝謝。我們的下一個問題來自摩根士丹利的 Brian Nowak。請提出你的問題。

Brian Nowak - Morgan Stanley, Analyst

(00:30:41)Thanks for taking my questions. I have two. Andy, your excitement about AWS is evident. It's rare to hear comments about a trillion dollars of revenue five to ten years from now. That's more than most would suggest. Could you walk us through how you arrive at that trillion-dollar figure? Is it a top-down approach? Or bottom-up? What gives you confidence that you can achieve that level of revenue growth in the next five to ten years? Secondly, for either of you, regarding robotics in 2027, what key milestones are you focused on for the retail business?

(00:30:41)感謝你回答我的問題。我有兩個問題。Andy,你對 AWS 的興奮顯而易見。聽到關於五到十年後一萬億美元收入的評論是很少見的。這比大多數人所建議的要多。你能告訴我們你是如何得出這一萬億美元的數字的嗎?這是自上而下的方法嗎?還是自下而上的?什麼讓你有信心在接下來的五到十年內能實現這樣的收入增長?其次,對於你們兩位,關於 2027 年的機器人技術,你們在零售業務中關注哪些關鍵里程碑?

Andy Jassy - Amazon, CEO

(00:31:21)On the AWS side, Brian, look at the growth rates on the current run rate and what we're projecting for 2027 and 2028. As I've said a few times, we will always— Despite how much capacity we're providing, the large chunks of capacity customers take—hundreds of megawatts or sometimes up to a gigawatt—are reserved well in advance because you can't leave a gigawatt of capacity unused. We have very good visibility into 2027, 2028, and some of 2029 regarding what customers want in those large capacity chunks.

(00:31:21)在 AWS 方面,Brian,看看當前運行率的增長率,以及我們對 2027 年和 2028 年的預測。正如我幾次所說,我們將始終——儘管我們提供了大量的產能,但客戶所需的大量產能——數百兆瓦或有時高達一吉瓦——都是提前預訂的,因為你不能讓一吉瓦的產能閒置。我們對 2027 年、2028 年以及部分 2029 年客戶在這些大容量需求上的需求有非常好的可見性。

Andy Jassy - Amazon, CEO

(00:32:08)We always leave a fair amount of capacity available for on-demand use on EC2 or Bedrock, working closely with customers to ensure plenty of capacity is available in AWS. So, we have very good visibility into what AWS will look like over the next few years. Those growth rates, when you do the math on revenue and the current run rate, lead quickly to escalating numbers. And then— I would add that we don't believe much of the usage we see today will attenuate anytime soon; we're still very early in enterprises using inference pervasively in production applications.

(00:32:08)我們總是保留相當多的產能供 EC2 或 Bedrock 按需使用,並與客戶密切合作,以確保 AWS 中有足夠的產能可用。因此,我們對未來幾年 AWS 的情況有非常好的可見性。當你計算收入和當前運行率的增長率時,這會迅速導致數字的上升。然後——我想補充的是,我們不認為今天看到的使用量會很快減少;我們仍然處於企業在生產應用中廣泛使用推理的早期階段。

Andy Jassy - Amazon, CEO

(00:32:50)It is truly just the beginning. Part of that is enterprises gaining enough skilled people to operate AI, post-train, and fine-tune models for the right applications. Another part is the cost of inference. If we want enterprises to use inference as we believe they will, we must continue working to reduce the cost of inference. That is a major mission for AWS. It's a significant reason we're building our own custom silicon, including Trainium, because we believe it's crucial for customers to use inference more cost-effectively.

(00:32:50)這確實只是個開始。其中一部分是企業獲得足夠的技術人員來操作人工智慧、後訓練和微調模型以適應正確的應用。另一部分是推理的成本。如果我們希望企業如我們所相信的那樣使用推理,我們必須繼續努力降低推理的成本。這是 AWS 的一個主要任務。這是我們建造自定義矽晶片的重大原因之一,包括 Trainium,因為我們相信這對客戶以更具成本效益的方式使用推理至關重要。

Andy Jassy - Amazon, CEO

(00:33:26)The cost of inference has come down substantially over the last decade. Over the past couple of years, and I believe it will decrease meaningfully more. So, just think about what you're seeing. For example, in security alone, today's models are so advanced that they find vulnerabilities humans have not yet discovered. And if you've had access to these models, you are using them both to identify issues and ultimately to assist with patching and deployment. Some believe that over the next year or two, about 20% of the total tokens used may be dedicated solely to security.

(00:33:26)在過去十年中,推理的成本已大幅下降。在過去幾年中,我相信它會顯著降低。所以,想想你所看到的。例如,僅在安全方面,今天的模型已經如此先進,以至於能夠發現人類尚未發現的漏洞。如果您已經接觸到這些模型,您將利用它們來識別問題,最終協助修補和部署。有些人認為,在接下來的一到兩年內,約 20% 的總代幣使用量可能專門用於安全。

Andy Jassy - Amazon, CEO

(00:34:11)This is a relatively new use case, building on the growth we see in coding and the various AI companions emerging in both consumer and workspace environments, such as with Q. These represent just the initial major use cases. Additionally, I believe virtually every modern enterprise production application will be rewritten to be agentically capable. This approach is necessary to move quickly and to properly guard against security risks. There is an immense amount of AI development underway. It's impossible to predict with certainty five to ten years from now, especially at these growth rates, but it's plausible this could become a trillion-dollar business.

(00:34:11)這是一個相對較新的用例,基於我們在編碼和各種 AI 伴侶的增長,這些伴侶在消費者和工作環境中出現,例如 Q。這些僅代表最初的主要用例。此外,我相信幾乎每個現代企業的生產應用程序都將被重寫,以具備代理能力。這種方法對於快速行動和妥善防範安全風險是必要的。目前正在進行大量的 AI 開發。在未來五到十年內,無法確定預測,但在這些增長率下,這可能成為一個萬億美元的業務。

Andy Jassy - Amazon, CEO

(00:35:09)On the robotics front, what's happening right now is quite remarkable. This is especially true at Amazon. We now have over a million robots operating in our fulfillment network. These robots have significantly impacted our cost structure, productivity, and safety; our fulfillment network's safety metrics are at record lows. We've made substantial progress over the past four to five years, with robotics playing a major role. Yet, I would say we are still in the early stages regarding both what we have built and deployed.

(00:35:09)在機器人技術方面,現在發生的事情相當驚人。在亞馬遜尤其如此。我們現在在履行網絡中運行著超過一百萬台機器人。這些機器人對我們的成本結構、生產力和安全性產生了重大影響;我們的履行網絡安全指標達到了歷史低點。在過去四到五年中,我們取得了重大進展,機器人在其中扮演了重要角色。然而,我會說我們在建設和部署方面仍處於早期階段。

Andy Jassy - Amazon, CEO

(00:35:53)And considering what we're currently developing, we will always have many teammates in our fulfillment centers, but many of the repetitive jobs currently done by humans will be handled by robots. We are investing in a variety of additional robotics across different form factors. We also see opportunities to assist with last-mile delivery to customer stores. We have a lot in development. For us, the key metrics for robotics are whether they deliver comparable or better productivity, safety, cost structure, and delivery times.

(00:35:53)而考慮到我們目前正在開發的內容,我們在履行中心始終會有許多隊友,但目前由人類完成的許多重複性工作將由機器人處理。我們正在投資於各種不同形狀的額外機器人技術。我們還看到協助最後一公里配送到客戶商店的機會。我們有很多正在開發中。對我們來說,機器人的關鍵指標是它們是否提供可比或更好的生產力、安全性、成本結構和交付時間。

Andy Jassy - Amazon, CEO

(00:36:36)When we see that combination of economics, we deploy robotics at scale. This is evident in our fulfillment network. Our new fulfillment network designs, including expanded same-day facilities, heavily incorporate robotics. Robotics has had a significant impact, though we believe this is small compared to what will occur over the next five years.

(00:36:36)當我們看到這種經濟組合時,我們會大規模部署機器人。這在我們的履行網絡中顯而易見。我們的新履行網絡設計,包括擴展的當日設施,重度整合了機器人技術。機器人技術產生了重大影響,儘管我們認為這與未來五年內將發生的事情相比仍然很小。

Operator - Operator

(00:37:03)Thank you. Our next question comes from Mark Mahaney of Evercore ISI. Please proceed with your question.

(00:37:03)謝謝。我們的下一個問題來自 Evercore ISI 的 Mark Mahaney。請提出您的問題。

Mark Mahaney - Evercore ISI, Analyst

(00:37:13)Two things. First, thank you for the forward comments on AWS growth. Could you also elaborate on margins? I understand margins fluctuate quarter to quarter, but over the next two years, do you expect the current margin structure to remain relatively stable? Specifically, do you anticipate maintaining operating margins in the high 30s to 40% range for AWS? Secondly, regarding perishable checkouts, I'm very excited about this opportunity, having observed its rollout in my neighborhood and elsewhere. How extensive is your coverage for perishable checkout?

(00:37:13)兩件事。首先,感謝您對 AWS 增長的前瞻性評論。您能否詳細說明一下利潤率?我瞭解利潤率會隨著季度波動,但在接下來的兩年中,您是否預期目前的利潤結構將保持相對穩定?具體來說,您是否預計 AWS 的營業利潤率將保持在高 30% 到 40% 的範圍內?其次,關於易腐爛商品的結帳,我對這個機會感到非常興奮,因為我在我的社區和其他地方觀察到了它的推出。您的易腐爛商品結帳覆蓋範圍有多廣?

Mark Mahaney - Evercore ISI, Analyst

(00:37:46)In other words, how many Prime households do you estimate are covered by perishable checkout? For those who have used it, how much of a boost does it provide per Prime household? Does it increase coverage? Or does it increase spend per Prime household by 20%, 50%, or even 100%, considering the size of the fresh grocery category? Thank you.

(00:37:46)換句話說,您估計有多少 Prime 家庭受到易腐爛商品結帳的覆蓋?對於那些使用過的人來說,這對每個 Prime 家庭提供了多少提升?它是否增加了覆蓋範圍?或者考慮到新鮮食品類別的規模,它是否使每個 Prime 家庭的支出增加了 20%、50% 或甚至 100%?謝謝。

Brian Olsavsky - Amazon, CFO

(00:38:08)Hey Mark, let me begin with the operating margin outlook for AWS. We are not providing guidance on AWS margins, but I can share my perspective. The operating margin was strong this quarter at 38.7%. This profitability is not random; it reflects discipline, efficiency gains, capacity optimization, and careful fixed cost management. As Andy mentioned, it also includes increasing use of custom chips in our cost structure over time. However, there are other factors at play. The rapid growth of AI exerts margin pressure on the overall business because AI margins are lower.

(00:38:08)嘿,Mark,讓我先談談 AWS 的營業利潤展望。我們不會對 AWS 的利潤率提供指導,但我可以分享我的觀點。本季度的營業利潤率為 38.7%,表現強勁。這種盈利能力並非隨機,而是反映了紀律、效率提升、產能優化和謹慎的固定成本管理。正如 Andy 提到的,這也包括隨著時間推移在我們的成本結構中增加使用定製芯片。然而,還有其他因素在起作用。AI 的快速增長對整體業務施加了利潤壓力,因為 AI 的利潤率較低。

Brian Olsavsky - Amazon, CFO

(00:39:04)We are pleased with progress in both core and AI margins; as Andy noted, AI margins are ahead of where core margins were at a similar stage, though still lower, creating a mixed impact. There are also headwinds from component pricing, as previously discussed. While unpredictable, multiple factors will continue to influence margins; we are working hard on cost efficiency and managing investments and new product margins as they scale.

(00:39:04)我們對核心和 AI 利潤率的進展感到滿意;正如 Andy 所提到的,AI 利潤率在相似階段超過了核心利潤率,儘管仍然較低,造成了混合影響。之前討論過的元件價格也帶來了逆風。雖然不可預測,但多種因素將繼續影響利潤率;我們正在努力提高成本效率,並管理投資和新產品利潤率,隨著它們的擴大而調整。

Andy Jassy - Amazon, CEO

(00:39:41)Regarding perishables checkout, Mark, I share your enthusiasm. As you know, we've had a significant grocery business for some time. It has primarily focused on the middle aisles of a grocery store, including consumables, canned goods, pharmaceutical items, and beauty products. We also operate a leading organic grocery store through Whole Foods Market, which is growing faster than the overall grocery segment, both in revenue and profit, with a new format and daily shop concept in urban areas. This initiative is performing very well, and we are in the process of rolling it out over the next few years.

(00:39:41)關於易腐爛商品結帳,Mark,我與您分享同樣的熱情。如您所知,我們的雜貨業務已經有一段時間相當顯著。它主要集中在雜貨店的中間過道,包括消耗品、罐頭食品、藥品和美容產品。我們還通過全食超市運營一家領先的有機雜貨店,該店的增長速度超過整體雜貨市場,無論是在收入還是利潤方面,並在城市地區推出了新格式和每日購物概念。這項舉措表現非常良好,我們正在計劃在未來幾年內推廣它。

Andy Jassy - Amazon, CEO

(00:40:17)However, we've always understood that to serve customers the way they want and the way we aim to, we need to offer a mass-market selection. This effort begins with perishables. When we analyze total household spending, comparing what customers spend on items we could serve versus what they currently spend with us, the largest gap is in grocery. In many shopping trips, grocery shopping starts with perishables. We are very excited because we believe we have found a solution for shipping perishables from our same-day delivery facilities.

(00:40:17)然而,我們一直明白,要以客戶希望的方式和我們的目標來服務客戶,我們需要提供大眾市場的選擇。這項努力始於易腐物品。當我們分析家庭總支出時,對比客戶在我們可以提供的商品上花費的金額與他們目前在我們這裡花費的金額,最大的差距在於雜貨。在許多購物行程中,雜貨購物始於易腐物品。我們非常興奮,因為我們相信我們已經找到了一個解決方案,可以從我們的當日送貨設施運送易腐物品。

Andy Jassy - Amazon, CEO

(00:41:04)Our everyday essentials category has grown about three times faster than the rest of our store. It now accounts for approximately half of all units sold. Customers who purchase perishables through our same-day delivery buy about seven times more everyday essential units than those who do not. So, if the goal is to close the gap between overall consumer spending on items we can provide and what they currently spend with us, perishables play a significant role. We also observe that our grocery customers shop about four times more in non-grocery categories,

(00:41:04)我們的日常必需品類別的增長速度約為我們商店其他部分的三倍。現在,它大約佔所有銷售單位的一半。通過我們的當日送貨購買易腐物品的客戶,購買的日常必需品單位數量約是那些不使用我們服務的客戶的七倍。因此,如果目標是縮小我們可以提供的商品與客戶目前在我們這裡花費的總消費之間的差距,易腐物品扮演著重要角色。我們還觀察到,我們的雜貨客戶在非雜貨類別的購物次數約為四倍,

Andy Jassy - Amazon, CEO

(00:41:42)such as home goods and apparel, and spend about twice as much in these categories compared to customers who do not use us for grocery. These are substantial differences, and we believe the progress we're making in perishables and grocery will ultimately make us more relevant to customers across more of their shopping needs.

(00:41:42)例如家居用品和服裝,並且在這些類別中的花費約是那些不使用我們雜貨服務的客戶的兩倍。這些都是相當大的差異,我們相信我們在易腐物品和雜貨方面所取得的進展,最終將使我們在客戶的更多購物需求中變得更具相關性。

Operator - Operator

(00:42:06)Thank you. Our next question comes from Eric Sheridan. He is with Goldman Sachs. Please proceed with your question.

(00:42:06)謝謝。我們的下一個問題來自埃裡克·謝裡丹。他來自高盛。請提出您的問題。

Eric Sheridan - Goldman Sachs, Analyst

(00:42:15)Thank you for taking my question. I have a two-part question regarding market opportunities within AWS. First, regarding Bedrock, what are your key learnings about the market opportunity and how it might scale in the coming years? Second, concerning the decision to sell Trainium externally, how do you balance the market opportunity for external sales with the need for compute capacity for that opportunity relative to other internal use cases and AWS broadly? Thank you.

(00:42:15)感謝您回答我的問題。我有一個關於 AWS 市場機會的兩部分問題。首先,關於 Bedrock,您對市場機會的主要學習是什麼,以及它在未來幾年可能如何擴展?其次,關於將 Trainium 外部銷售的決定,您如何平衡外部銷售的市場機會與相對於其他內部用例和 AWS 整體的計算能力需求?謝謝。

Andy Jassy - Amazon, CEO

(00:42:52)Yes. Regarding Bedrock, as you may have heard from us multiple times over the past few years, one key lesson from over 20 years of offering technology to customers is that there is no single tool that dominates the market. People often assume that the early leader will be the only tool used, but that has never been the case — not in databases, analytics, or AI frameworks. Initially, TensorFlow was seen as the sole framework, but PyTorch has since become much larger. The same will apply to models. Over the next two to three years, I expect there will be six to eight models that perform similarly well or are sufficient for over 95% of workloads.

(00:42:52)是的。關於 Bedrock,正如您可能在過去幾年中多次聽到的,提供技術給客戶的 20 多年來的一個關鍵教訓是,市場上沒有單一的工具主導。人們經常假設早期的領導者將是唯一使用的工具,但這從來不是事實——無論是在數據庫、分析還是 AI 框架中。最初,TensorFlow 被視為唯一的框架,但 PyTorch 隨後變得更大。模型也將適用於此。在接下來的兩到三年內,我預計將會有六到八個模型表現相似良好,或足以滿足超過 95% 的工作負載。

Andy Jassy - Amazon, CEO

(00:43:51)Some models will excel at reasoning, others at coding, some at visual tasks, some will offer lower latency, and others will be more cost-effective. Most customers will use multiple models, often within the same generative AI applications. Customers consistently tell us they want a service offering a broad selection of leading models, with excellent price-performance, strong security, and access control. They want the flexibility to choose and switch between models as they evolve and as their application requirements change.

(00:43:51)一些模型在推理方面表現出色,其他模型在編碼方面表現優異,一些在視覺任務上表現良好,某些模型將提供更低的延遲,而其他模型則更具成本效益。大多數客戶將使用多個模型,通常在同一生成 AI 應用中。客戶不斷告訴我們,他們希望有一個提供廣泛領先模型的服務,具有優秀的性價比、強大的安全性和訪問控制。他們希望能夠根據模型的演變和應用需求的變化,靈活選擇和切換模型。

Andy Jassy - Amazon, CEO

(00:44:37)This is exactly what we see with Bedrock. People are very enthusiastic about it. We offer both closed models and models like those from Anthropic. We also provide open models, including those from Kimi and others. The Western market is particularly eager for additional Western open models. I believe this demand will be met over time. Overall, the opportunity size for Bedrock is enormous. It could become as large a service as EC2, which is already a major AWS offering. So, we see it as a very significant opportunity.

(00:44:37)這正是我們在 Bedrock 中所看到的。人們對此非常熱情。我們提供封閉模型和像 Anthropic 這樣的模型。我們還提供開放模型,包括來自 Kimi 和其他公司的模型。西方市場對額外的西方開放模型特別渴望。我相信這一需求將隨著時間的推移得到滿足。總體而言,Bedrock 的機會規模是巨大的。它可能成為與 EC2 一樣大的服務,而 EC2 已經是 AWS 的一個主要產品。因此,我們將其視為一個非常重要的機會。

Andy Jassy - Amazon, CEO

(00:45:18)Regarding the balance between chips sold externally and those used internally in our cloud, that's an excellent question. We are actively evaluating this balance as we speak. Currently, only two companies have both a significant cloud business and chips strong enough to sell externally. Therefore, very few customers face this type of challenge at such a scale. At this point, we continue to assess the best approach moving forward.

(00:45:18)關於外部銷售的芯片與我們雲中內部使用的芯片之間的平衡,這是一個很好的問題。我們正在積極評估這種平衡。目前,只有兩家公司擁有既有顯著的雲業務又有足夠強大的芯片可以外部銷售。因此,面對這種挑戰的客戶非常少。在這一點上,我們將繼續評估未來的最佳方法。

Andy Jassy - Amazon, CEO

(00:45:58)Everybody is producing their advanced chips with the same provider. There is only a limited number of chips that this provider can supply each year. Everyone receives an allocation, and you must decide how much of that allocation to sell externally versus how much to use in your cloud. There is a trade-off: if you sell externally, you can recognize revenue immediately as the chips leave your inventory. Conversely, if you use the chips within your cloud, such as in EC2 instances or Bedrock, you will generate more absolute revenue and free cash flow over time, since you utilize them over five to six years, driving revenue for a longer period.

(00:45:58)每個人都在與同一供應商生產其先進的晶片。這個供應商每年能提供的晶片數量有限。每個人都會獲得一個分配額,而你必須決定將多少分配額用於外部銷售,多少用於你的雲端。這是一個權衡:如果你選擇外部銷售,你可以在晶片離開庫存時立即確認收入。相反地,如果你在雲端內部使用這些晶片,例如在 EC2 實例或 Bedrock 中,你將隨著時間的推移產生更多的絕對收入和自由現金流,因為你會在五到六年內使用它們,從而推動更長期的收入。

Andy Jassy - Amazon, CEO

(00:46:49)It requires careful consideration of your strategic priorities. To date, we have taken a different approach in how we use our chips compared to the other provider facing a similar balancing act. As I mentioned earlier, we expect to have a very significant business selling chips externally, starting in late 2027. We will continue to evaluate that balance, likely on an annual basis. Both internal use and external sales will be very significant. Our ability to create price-performance scalable chips will be a major strategic advantage at scale, benefiting both our cost structure and our external chip sales business.

(00:46:49)這需要仔細考慮你的戰略優先事項。到目前為止,我們在使用晶片的方式上採取了與其他面臨類似平衡行為的供應商不同的方法。正如我之前提到的,我們預計到 2027 年底將開始有非常重要的外部晶片銷售業務。我們將繼續評估這種平衡,可能會每年進行一次。內部使用和外部銷售都將非常重要。我們創造價格性能可擴展晶片的能力將在規模上成為一個主要的戰略優勢,這將有利於我們的成本結構和外部晶片銷售業務。

Andy Jassy - Amazon, CEO

(00:47:35)Selling external chips is fundamentally a very different business from our cloud infrastructure business.

(00:47:35)銷售外部晶片從根本上來說是一個與我們的雲端基礎設施業務非常不同的業務。

Operator - Operator

(00:47:44)Thank you. Our next question comes from Doug Anmuth with JP Morgan. Please proceed with your question.

(00:47:44)謝謝。我們的下一個問題來自 JP Morgan 的 Doug Anmuth。請提出你的問題。

Doug Anmuth - JP Morgan, Analyst

(00:47:53)Thanks for taking my questions. When you consider the 40% plus AWS revenue growth projected for each of the next two years and the multi-year path beyond that, can you discuss the CapEx required to support that revenue growth over the next few years and beyond? Also, Andy, any updates on how you are thinking about the duration of returns on servers, networking, and data centers?

(00:47:53)感謝你回答我的問題。當你考慮到未來兩年預計超過 40% 的 AWS 收入增長以及之後的多年度增長路徑時,你能談談支持這種收入增長所需的資本支出嗎?此外,Andy,對於伺服器、網絡和數據中心的回報持續時間,你有什麼更新嗎?

Andy Jassy - Amazon, CEO

(00:48:20)Thanks. I'll take those in reverse order. Regarding breakeven on our infrastructure, as I mentioned in my opening remarks, we are continuing to move that breakeven point earlier. We are consistently shortening the breakeven period. There are many cores involved, which differ somewhat from AI workloads and how these assets are financed. So the breakeven varies depending on the asset, but we continue to make progress, as we historically have been effective at extending useful lives and accelerating breakeven in AWS.

(00:48:20)謝謝。我將按相反的順序回答這些問題。關於我們基礎設施的盈虧平衡,正如我在開場白中提到的,我們正在持續將盈虧平衡點提前。我們不斷縮短盈虧平衡期。涉及的核心數量很多,這些核心與 AI 工作負載及這些資產的融資方式有所不同。因此,盈虧平衡因資產而異,但我們持續取得進展,因為我們歷來在延長使用壽命和加速 AWS 的盈虧平衡方面非常有效。

Andy Jassy - Amazon, CEO

(00:48:57)We are making good progress in this area as well. Regarding the CapEx needed to support the growth we are discussing, it is substantial. We view this as a once-in-a-lifetime opportunity. AI represents the biggest transformation in our lifetime. Considering the ability to grow 39% with a $184 billion annual run rate, and to sustain 40% growth annually over the next two years, ultimately becoming a trillion-dollar-plus business, we will invest to maintain our leadership. We are the current leader in infrastructure and AI.

(00:48:57)我們在這方面也取得了良好的進展。關於支持我們討論的增長所需的資本支出,這是相當可觀的。我們將這視為一次千載難逢的機會。AI 代表了我們一生中最大的變革。考慮到以 1840 億美元的年運行率增長 39%,並在未來兩年內保持每年 40% 的增長,最終成為一個超過一兆美元的業務,我們將投資以維持我們的領導地位。我們是基礎設施和 AI 的當前領導者。

Andy Jassy - Amazon, CEO

(00:49:48)Would it be more convenient if this acceleration and opportunity unfolded over 25 years, allowing CapEx to be spread out? Yes, but that is not the reality or the opportunity we face. We can fund this plan, and it will fundamentally change the size of Amazon. It will permanently transform our revenue. It will permanently transform our free cash flow. We will invest and pursue this to remain the significant leader we are today.

(00:49:48)如果這種加速和機會在 25 年內展開,讓資本支出可以分散,會不會更方便?是的,但這不是我們面臨的現實或機會。我們可以資助這個計劃,這將根本改變亞馬遜的規模。這將永久改變我們的收入。這將永久改變我們的自由現金流。我們將投資並追求這一點,以保持我們今天的顯著領導地位。謝謝。

Operator - Operator

(00:50:30)Thank you. Our final question comes from Ross Sandler with Barclays. Please proceed with your question.

(00:50:30)我們的最後一個問題來自 Barclays 的 Ross Sandler。請提出你的問題。很好。

Ross Sandler - Barclays, Analyst

(00:50:39)Great. I wanted to follow up on the Bedrock question. The token marketplace on Bedrock appears to be one of the hottest AI services currently. Our understanding is that the majority of these workloads run on Trainium. As this becomes a larger part of the overall AWS story, as you mentioned, Andy, how might that impact both revenue growth and especially margins at AWS as you shift from selling bare metal to offering more inference as a service? Could you discuss that? Thank you.

(00:50:39)我想跟進 Bedrock 的問題。Bedrock 上的代幣市場似乎是當前最熱門的 AI 服務之一。我們的理解是,這些工作負載中的大多數運行在 Trainium 上。隨著這成為整體 AWS 故事中更大的一部分,正如你提到的,Andy,這可能會如何影響 AWS 的收入增長,尤其是利潤率,因為你從銷售裸金屬轉向提供更多推理作為服務?你能談談這個嗎?謝謝。是的。

Andy Jassy - Amazon, CEO

(00:51:18)Yes. You are correct that the majority of Bedrock workloads run on Trainium. We do this in large part because we really value The price performance that we are achieving. It allows us to keep prices as low as possible for customers while maintaining the margins we value. When considering margins in the cloud space and comparing different players' cost structures, there are several key factors to consider. The first is their data centers, Specifically, how many they own or long-lease versus how many are colocated or with neocloud providers.

(00:51:18)你是對的,大多數 Bedrock 工作負載運行在 Trainium 上。我們這樣做在很大程度上是因為我們非常重視我們所達成的價格表現。這使我們能夠在保持我們重視的利潤率的同時,為客戶提供盡可能低的價格。在考慮雲端領域的利潤率並比較不同參與者的成本結構時,有幾個關鍵因素需要考慮。第一個是他們的數據中心,具體來說,他們擁有或長期租賃的數量與與新雲供應商共同設置的數量。

Andy Jassy - Amazon, CEO

(00:52:12)Neoclouds typically have operating margins about 3,500 basis points lower than long-leased data centers. Next, consider the hardware on which their services run. If you are using... The same hardware as everyone else, your cost structure will be comparable to theirs. However, having the ability to use different hardware provides an advantage. Most of our capacity is either owned or long-leased, which is, It results in better operating margins compared to those relying heavily on colocation or neocloud solutions.

(00:52:12)新雲的運營利潤率通常比長期租賃的數據中心低約 3500 個基點。接下來,考慮他們服務運行的硬體。如果你使用的是...與其他人相同的硬體,你的成本結構將與他們相當。然而,能夠使用不同的硬體則提供了一個優勢。我們的大部分容量要麼是自有的,要麼是長期租賃的,這就是,與那些高度依賴共同設置或新雲解決方案的公司相比,這會導致更好的運營利潤率。

Andy Jassy - Amazon, CEO

(00:53:06)Running on our own chip, which offers superior price performance and cost efficiency because we own it, gives us structural cost advantages that benefit our customers as we pass those savings on. This also supports the overall economics of AWS.

(00:53:06)在我們自己的晶片上運行,因為我們擁有它,提供了優越的價格表現和成本效率,這使我們在結構上具有成本優勢,並將這些節省轉嫁給客戶。這也支持了 AWS 的整體經濟學。

Dave Fildes - Amazon, Investor Relation

(00:53:32)Thank you for joining the call today and for your questions. A replay will be available on our investor relations website for at least three months. We appreciate your interest in Amazon and look forward to speaking with you again next quarter.

(00:53:32)感謝您今天參加電話會議並提出問題。重播將在我們的投資者關係網站上提供至少三個月。我們感謝您對亞馬遜的關注,期待在下個季度再次與您交談。